diff --git a/projects/3D Virtual Test Track for Autonomous Driving/README.md b/projects/3D Virtual Test Track for Autonomous Driving/README.md deleted file mode 100644 index f2041510..00000000 --- a/projects/3D Virtual Test Track for Autonomous Driving/README.md +++ /dev/null @@ -1,74 +0,0 @@ -**Project 171:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

3D Virtual Test Track for Autonomous Driving

-

Design a 3D virtual environment to test the diverse conditions needed to develop an autonomous vehicle.

-
- -## Motivation - -Autonomous driving will revolutionize transportation and change the way we move around and receive goods and services. Designing a safe and reliable autonomous vehicle is one of the hardest problems that humans have ever tackled. Simulation is critical to the development of such systems, but there are few open scene datasets. By developing a 3D virtual test track available to researchers worldwide you will help accelerate the development of autonomous vehicles. - -## Project Description - -Test tracks such as [Mcity](https://en.wikipedia.org/wiki/Mcity) and [K-City](https://www.imnovation-hub.com/digital-transformation/k-cit-test-bed-fo-driverless-cars/) are used to test physical prototypes for autonomous vehicles. Virtual test tracks can be used at an earlier stage to explore ideas and test algorithms for autonomous driving. RoadRunner is the leading product for building automated driving scenes. Do a literature search to find the common challenging scenarios for autonomous driving and use RoadRunner to build a test track that has the key elements. Export your work into industry-standard formats such as [OpenDRIVE](https://www.asam.net/standards/detail/opendrive/) and [FBX](https://www.autodesk.com/products/fbx/overview). - -Attributes of such a dataset might include: - -- Both highway and urban road elements - -- Range of road grades, banking, and curvatures - -- Traffic control infrastructure (signals, signage, street markings) - -- Occlusions (buildings, signposts, trees, parked vehicles) - -- Complex junctions (offset, oblique, 5+ legs, roundabouts, merges and splits) - -- Restricted lane types (bus only, bike only, gore points) - -- Pedestrian-road interfaces (crosswalks, sidewalks, “pedestrian scramble”) - -- Sensor challenges (weathered lane markings, occluded signage, overpasses, foliage) - -- Unprotected turns - - -Advanced features: - -- Adapt your scene to create variants incorporating the signs and road markings of different countries - -- Add different pavement types and on-road elements including speed bumps and traffic control objects (cones, barriers, etc.) - -- Edit your scene to create a version suitable for left-hand traffic - -## Background Material - -- [RoadRunner](https://www.mathworks.com/products/roadrunner.html) - -- [K-City: Pilot City for Autonomous Vehicles](https://www.youtube.com/watch?v=uts6n8go1Q0) - -## Impact - -Contribute to autonomous vehicle development by creating virtual test scenes that can be used with many simulators across multiple vehicle development programs. - -## Expertise Gained - -Autonomous Vehicles, Automotive, Modeling and Simulation - - -## Project Difficulty - -Bachelor - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/20) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[pfryscak](https://github.com/pfryscak) - -## Project Number - -171 diff --git a/projects/Aggressive Maneuver Stabilization for a Minidrone/README.md b/projects/Aggressive Maneuver Stabilization for a Minidrone/README.md deleted file mode 100644 index 4293e844..00000000 --- a/projects/Aggressive Maneuver Stabilization for a Minidrone/README.md +++ /dev/null @@ -1,71 +0,0 @@ -**Project 230:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Aggressive Maneuver Stabilization for a Minidrone

-

Design a controller to enable a micro aerial vehicle to stabilize in the scenario of an external aggressive disturbance.

-
- -## Motivation - -The Unmanned Aerial Vehicle industry is a growing field with its applications in transportation, delivery, agriculture, and surveillance. The MathWorks tools play a crucial part in designing these systems using the Model-Based Design approach – whether to enable the [Pilotless Flight of Aurora Centaur](https://www.mathworks.com/videos/pilotless-flight-of-aurora-centaur-119494.html) or to develop Airnamics [Unmanned Aerial System for Close-Range Filming](https://www.mathworks.com/company/user_stories/airnamics-develops-unmanned-aerial-system-for-close-range-filming-with-model-based-design.html). - -Performing aggressive maneuvers is a challenging control problem for UAVs that need to be addressed for all applications where agile flying vehicles need to move with high acceleration and pass-through obstacles with a precise pose value that can approach singularity. Moreover, such control strategies will be necessary to overcome the hurdles caused by unexpected external circumstances – a strong gust of wind, relaunching from a failed vehicle landing, an obstacle disturbance in a cluttered space, etc. - - -## Project Description - -Use MATLAB and Simulink to design and implement a non-linear control strategy able to deal with high disturbances, fast input variations, and track complex trajectories using the tools that are used by the aerospace industry. Provide an aggressive input to the minidrone to change its position and orientation and stabilize it to the designated position and orientation. -Suggested Steps: -1. Become familiar with the MATLAB and Simulink using resources listed in the Background Material section below. -2. Install the [Simulink Support Package for Parrot Minidrones](https://www.mathworks.com/matlabcentral/fileexchange/63318-simulink-support-package-for-parrot-minidrones) from MATLAB-Add-Ons. -3. Use the [Parrot Minidrone Hover Model](https://www.mathworks.com/help/supportpkg/parrot/ug/fly-a-parrot-minidrone-using-the-hover-simulink-model.html) as the baseline controller. -4. Improve the altitude estimator and controller for flying over objects by using the pressure sensor along with the altitude sensor -5. Design a controller to enable the vehicle to hover from - - A freehand throw - - A free fall - - An upside-down orientation [1] -6. Provide aggressive inputs to the aerial vehicle using inputs in simulations. Update the controller and state estimator to stabilize the minidrone’s flight. You can use the data from the minidrone’s sensors. - -7. Hardware Deployment: If you have the hardware available with you, deploy your algorithm designed in simulations on the Parrot Mambo Minidrone hardware. Check the Background Material for details. - -Advanced project work: -1. Generate a complex trajectory for maneuver to have the drone follow it – in simulations and deployed on the hardware. -Project variations: -1. Implement a quaternion-based attitude controller and state estimator [2], [3] to enable the drone to perform a 360 degrees flip maneuver - - -## Background Material - -- Getting started self-paced courses - [MATLAB Onramp](https://matlabacademy.mathworks.com/details/matlab-onramp/gettingstarted?s_tid=abt_train_b), [Simulink Onramp](https://www.mathworks.com/learn/tutorials/simulink-onramp.html), [Control Design Onramp](https://www.mathworks.com/learn/tutorials/control-design-onramp-with-simulink.html) -- Deploy to hardware using [Simulink Support Package for Parrot Minidrones](https://www.mathworks.com/help/supportpkg/parrot/) -- Video series on [Drone Simulation and Control](https://www.mathworks.com/videos/series/drone-simulation-and-control.html) that explains the workflow for developing a control system for the Parrot Mambo Minidrone and explains how to deploy the algorithms on the hardware - -Suggested readings: - -[1] Taeyoung Lee, Melvin Leok, and N. Harris McClamroch, " Geometric Tracking Control of a Quadrotor UAV on SE(3) ", 49th IEEE Conference on Decision and Control December 15-17, 2010 Hilton Atlanta Hotel, Atlanta, GA, USA - -[2] Emil Fresk and George Nikolakopoulos, “Full Quaternion Based Attitude Control for a Quadrotor”, 2013 European Control Conference (ECC) July 17-19, 2013, Zürich, Switzerland. - -[3] C. G. Mayhew, R. G. Sanfelice, and A. R. Teel, “Quaternion-based hybrid control for robust global attitude tracking,” IEEE Transactions on Automatic control, vol. 56, no. 11, pp. 2555–2566, 2011. - -## Impact - -Contribute to advancements in aerial vehicle control in contracted spaces with unforeseen environment conditions. - -## Expertise Gained - -Autonomous Vehicles, Drones, Robotics, Aerospace, Low-cost Hardware, Modeling and Simulation, State Estimation, UAV, Control - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/63) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -230 diff --git a/projects/Aggressive Maneuver Stabilization for a Minidrone/student submissions/project-230 b/projects/Aggressive Maneuver Stabilization for a Minidrone/student submissions/project-230 deleted file mode 160000 index 12abf351..00000000 --- a/projects/Aggressive Maneuver Stabilization for a Minidrone/student submissions/project-230 +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 12abf351488a404a101d2e8656958348ad7a4388 diff --git a/projects/Applying Machine Learning for the Development of Physical Sensor Models in Game Engine Environment/README.md b/projects/Applying Machine Learning for the Development of Physical Sensor Models in Game Engine Environment/README.md deleted file mode 100644 index 92db7993..00000000 --- a/projects/Applying Machine Learning for the Development of Physical Sensor Models in Game Engine Environment/README.md +++ /dev/null @@ -1,59 +0,0 @@ -**Project 149:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Applying Machine Learning for the Development of Physical Sensor Models in Game Engine Environment

-

Realistic synthetic sensor data will soon eliminate the need of collecting tons of real data for machine learning algorithms. Accelerate this transition by creating a real-time camera distortion model.

-
- -## Motivation - -Deep Learning technology has now been adopted in almost every domain with results that have reached and even surpassed the level of accuracy of conventional techniques. However, to reach such a high level of performances, a huge amount of data is needed. -The development of learning-based detection and classification algorithms for autonomous system applications requires sensor’s data previously collected to use during the training process. For example, one of the challenges in developing algorithms for advanced driver assistance systems (ADAS) is recording sensor signals (e.g. image, video, point cloud, etc.) and labelling them with ground truth data. MathWorks has developed several virtual sensors to generate synthetic sensor data using game engine. Machine learning can be used even in this case to expedite and expand the virtual sensor development. -The objective of this project is to automate development of new Game Engine Integration Component and Automated Driving Toolbox™ (ADT) sensors, and refine models of the existing ADT sensors by applying machine learning methods. - - -## Project Description - -This project aims to implement a deep learning-based approach to distort, in real-time, synthetic images with the objective to simulate a stream of camera data. -An implementation of the un-distortion algorithm for an ADAS monocular camera is well known and is available as part of the Computer Vision System Toolbox™ (CVT). (https://www.mathworks.com/help/vision/ug/camera-calibration.html) -Implementation of a distortion algorithm is quite challenging because it requires solving cubic or sextic level equations for every pixel of the image, making it unsuitable for real-time applications. The objective of this project is to adopt a machine learning technique (an example could be GANs, i.e. Generative Adversarial Networks) for developing a model of the physical camera with distortion able to output data in real-time. - -Suggested steps: - -1. From the Unreal Game Engine, collect undistorted images using the Unreal Engine Scenario Simulation in Simulink® (https://www.mathworks.com/help/driving/unreal-engine-scenario-simulation.html) -2. Obtain a second set of images by distorting the previously collected ones by solving the cubic equations from the CVT (https://www.mathworks.com/help/vision/ug/camera-calibration.html) for undistorted pixels (x, y). Both sets of images will be necessary to train your Neural Network. -3. Train the Generative Adversarial Network or your preferred network architecture using the Deep Learning Toolbox™ (https://www.mathworks.com/help/deeplearning/ug/train-generative-adversarial-network.html) -4. Deploy the model into a simulated environment, test the correctness of the output, and process time against the distortion method based on cubic equations. - - -## Background Material - -- [Simulation 3D Camera](https://www.mathworks.com/help/driving/ref/simulation3dcamera.html) (the distortion model used in this block works only for low distortion lens) -- [Computer Vision Toolbox](https://www.mathworks.com/help/vision/index.html?s_tid=CRUX_lftnav) -- [Deep Learning Toolbox](https://www.mathworks.com/products/deep-learning.html) -- [How to Design and Train Generative Adversarial Networks (GANs)](https://www.mathworks.com/videos/how-to-design-and-train-generative-adversarial-networks-gans-1583904310687.html) - -## Impact - -Reduce development efforts of autonomous vehicles and robots. - -## Expertise Gained - -Artificial Intelligence, Autonomous Driving, Computer Vision, Deep Learning, Machine Learning, Modeling and Simulation, Neural Networks - -## Project Difficulty - -Master’s level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/15) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[iklimchy](https://github.com/iklimchy) - -## Project Number - -149 diff --git a/projects/Augmented Reality for Architecture/README.md b/projects/Augmented Reality for Architecture/README.md deleted file mode 100644 index 5c66ee42..00000000 --- a/projects/Augmented Reality for Architecture/README.md +++ /dev/null @@ -1,84 +0,0 @@ -**Project 240:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Augmented Reality for Architecture

-

Develop an augmented reality system to enhance a photo or video of a 2D architectural floor plan printed on paper with a virtual 3D representation of the structure.

-
- -## Motivation - -Augmented reality (AR) combines the real world with computer generated content, often in an interactive way. Newer cell phones, VR headsets and even glasses now have some AR capabilities, and AR has been used for gaming, design, art, utility, and more. As companies strive to improve their sensors, display and computing hardware capabilities, AR will become increasingly commonplace and find new and exciting uses in the world. - -## Project Description - -This project aims to bring to life architectural drawing using the [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html). A live video stream of a 2D floor plan drawn or printed on paper will be processed and augmented with a 3D representation of the structure. - -Suggested steps: - -- Print or draw a floor plan consisting of line segments that represent walls onto a flat sheet of paper. - -- Record a video of the floor plan, moving the camera to view it from different angles. - -- Get the camera calibration parameters using the [Camera Calibrator App](https://www.mathworks.com/help/vision/ref/cameracalibrator-app.html). - -- Determine the pose of the floor plan using either known features (e.g. [AprilTag](https://www.mathworks.com/help/vision/ref/readapriltag.html)) or [detected features](https://www.mathworks.com/help/vision/feature-detection-and-extraction.html). - -- Detect the relevant floor plan features (e.g., lines for walls). - -- Augment the video with a 3D representation of the features and correct the visualization to match the perspective from the current pose of the camera. - -- Try running your algorithm in real-time on an incoming video feed from a webcam. - -Advanced work: - -- Use existing video features to estimate pose, rather than requiring a known added feature such as the AprilTag. - -- Include additional information in the floor plan, e.g., markers for windows, doors, furniture, colors, etc. and augment the 3D representation to show it. - -- Render features that obscure others with transparency. - -- Automatically adjust the colors of the rendered features to better match the lighting of the environment. - -- Automatically measure and display the lengths of walls with a scaling factor. - -- Compile and run in real-time on a cell phone or other mobile device. - -## Background Material - -Suggested readings: - -[1] Marco Schumann et al. [Evaluation of augmented reality supported approaches for product design and production processes.](https://www.sciencedirect.com/science/article/pii/S2212827120314402) Procedia CIRP 2021. - -[2] Georgiou, T., Liu, Y., Chen, W. et al. [A survey of traditional and deep learning-based feature descriptors for high dimensional data in computer vision](https://link.springer.com/article/10.1007/s13735-019-00183-w/?tag=dvside-21#citeas). Int J Multimed Info Retr 2020. - -Useful links: - -[Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html) - -[Feature Detection and Extraction](https://www.mathworks.com/help/vision/feature-detection-and-extraction.html) - -[Augmented Reality using AprilTag markers](https://www.mathworks.com/help/vision/ug/augmented-reality-using-apriltag-markers.html) - -[houhglines (Line detection in an image)](https://www.mathworks.com/help/images/ref/houghlines.html) - -## Impact - -Develop a proof-of-concept augmented reality system to aid in architectural design. - -## Expertise Gained - -Computer Vision, Image Processing, Sensor Fusion and Tracking - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/76) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -240 diff --git a/projects/Automatically Segment and Label Objects in Video/README.md b/projects/Automatically Segment and Label Objects in Video/README.md deleted file mode 100644 index 8ed837a6..00000000 --- a/projects/Automatically Segment and Label Objects in Video/README.md +++ /dev/null @@ -1,69 +0,0 @@ -**Project 203:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Automatically Segment and Label Objects in Video

-

Implement algorithms to automatically label data for deep learning model training

-
- -## Motivation - -The dream of delivering self-driving vehicles is the most exciting technical challenge of our lifetime. However, the early promise of achieving full autonomy (level 5) has been slowed down by the tremendous amount of labelled data required to build and validate robust AI algorithms needed to unlock level 5 autonomy. We desperately need better labeling tools that automatically label objects with minimal human intervention. Help build these labeling algorithms to accelerate our progress towards self-driving vehicles. - -## Project Description - -Using MathWorks [Computer Vision Toolbox ™](https://www.mathworks.com/products/computer-vision.html) and [Deep Learning Toolbox ™](https://www.mathworks.com/products/deep-learning.html), design, implement, and test an algorithm to automatically segment and label the same object across a sequence of video frames with minimal human intervention. This process is often referred to as video object segmentation or video instance segmentation. Demonstrate the effectiveness of your algorithm compared to manual labeling by integrating the algorithm into the [Video Labeler app](https://www.mathworks.com/help/vision/ref/videolabeler-app.html) and evaluating the time savings gained using automation. - -Suggested steps: - -1. Get familiar with the [Video Laber app]((https://www.mathworks.com/help/vision/ref/videolabeler-app.html)) and how to [create an automation Algorithm for labeling](https://au.mathworks.com/help/vision/ug/create-automation-algorithm-for-labeling.html). -2. Review state-of-the-art techniques for video object segmentation. -3. Identify and evaluate the effectiveness of algorithms from [[1]](#yao) for label automation. -4. Collect and prepare data required for training the algorithms, if needed. -5. Integrate these algorithms into the video Labeler app. -6. Measure the effectiveness of your labeling algorithm compared to manual labeling. - -Initial set of objects for segmentation: vehicles, pedestrians, cyclists, lane markers, drivable path, curbs, walkways. - - -## Background Material - -- [Video object segmentation](https://paperswithcode.com/task/video-object-segmentation) -- [Video instance segmentation](https://paperswithcode.com/task/video-instance-segmentation) -- [Getting started with the Video Labeler](https://au.mathworks.com/help/vision/ug/get-started-with-the-video-labeler.html) -- [Create an automation Algorithm for labeling](https://au.mathworks.com/help/vision/ug/create-automation-algorithm-for-labeling.html) -- [How to Use Custom Automation Algorithms for Data Labeling](https://www.youtube.com/watch?v=Y36D1fJZkT0) -- [Using Ground Truth for Object Detection](https://www.mathworks.com/matlabcentral/fileexchange/69180-using-ground-truth-for-object-detection?s_eid=PSM_15028) - -Suggested readings: - -[1] Yao, Rui, et al. "Video object segmentation and tracking: A survey." ACM Transactions on Intelligent Systems and Technology (TIST) 11.4 (2020): 1-47. - -[2] Caelles, Sergi, et al. "The 2019 davis challenge on vos: Unsupervised multi-object segmentation." arXiv preprint arXiv:1905.00737 (2019). - -[3] Yang, Linjie, Yuchen Fan, and Ning Xu. "Video instance segmentation." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019. - - -## Impact - -Accelerate the development of robust AI algorithms for self-driving vehicles. - -## Expertise Gained - -Artificial Intelligence, Computer Vision, Deep Learning, Machine Learning - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/33) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[mwbpatel](https://github.com/mwbpatel) - -## Project Number - -203 diff --git a/projects/Autonomous Navigation for Vehicles in Rough Terrain/README.md b/projects/Autonomous Navigation for Vehicles in Rough Terrain/README.md deleted file mode 100644 index d4a8cd58..00000000 --- a/projects/Autonomous Navigation for Vehicles in Rough Terrain/README.md +++ /dev/null @@ -1,83 +0,0 @@ -**Project 209:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Autonomous Navigation for Vehicles in Rough Terrain

-

Design and implement a motion planning algorithm for off-road vehicles on rough terrain.

-
- -## Motivation - -Automating a vehicle for off-road conditions poses different sets of challenges when compared to those tackled for on-road autonomous cars. The lack of road rules and lane structure presents the vehicle with unknown challenges it needs to deal with reactively. Various industries (Agriculture, Construction, Mining, Planetary exploration) are looking to automate tasks where humans are involved, especially in mundane tasks (alt: repetitive actions) and in some cases hazardous operational situations. -A lot of these applications require a human driver to move the vehicle to accomplish the tasks. Automation of this motion can help free up humans for more sophisticated jobs and move them away from harm's way. It can also help with faster scaling and quick deployments to newer areas. - -## Project Description - -Demo a robot/vehicle (AMR, front loader, excavator, curiosity mars rover) working in a cluttered field (off-road) moving from point A to point B. The field should have minor bumps and ditches (less than 0.25m from ground plane for 1m radius). - -Suggested steps: -1. Start with [Execute Tasks for a Warehouse Robot example]( https://www.mathworks.com/help/robotics/ug/execute-tasks-for-a-warehouse-robot.html). -2. Look at [A* Path Planning and Obstacle Avoidance in a Warehouse]( https://www.mathworks.com/help/robotics/ug/a-star-path-planning-and-obstacle-avoidance.html) to learn about replacing planner and connecting to Gazebo using co-simulation ([Perform Co-Simulation between Simulink and Gazebo]( https://www.mathworks.com/help/robotics/ug/perform-co-simulation-between-simulink-and-gazebo.html)) -3. Choose an application area (e.g. agriculture) and corresponding type of vehicle -4. Replace warehouse in above examples with a scenario of chosen application. Realistic scenes close the gap between simulation and real-world, use/construct a world with an uneven field and scattered obstacles -5. Pick/create an algorithm for motion planning which ensures stability of the vehicle. i.e. takes care of roll, pitch, elevation constraints forced by the terrain -6. Implement the chosen/created algorithm as a MATLAB function and replace A* or PRM in the above examples -7. Add sensor to the vehicle for sensing the environment, such as Lidar or Camera. Read sensor data from gazebo [Perform Co-Simulation between Simulink and Gazebo ](https://www.mathworks.com/help/robotics/ug/perform-co-simulation-between-simulink-and-gazebo.html) -8. Create map using sensor data (Easy: [Insert lidar pointcloud to 3D map](https://www.mathworks.com/help/nav/ref/occupancymap3d.insertpointcloud.html). Advanced: see example section below) -9. Integrate modules -10. Demo the vehicle using the pure pursuit algorithm to follow the path/trajectory from point A to point B using Simulink and Gazebo - -Project Variations: -1. Different domain (mining, construction, etc.) and different vehicles (front loaders, digging machines, combines (agriculture harvesters) -2. Different path following controllers such as model predictive controller (MPC) - -Advanced research work: -1. Bring in uncertainty handling to the planning algorithms -2. Deploying on to a platform and demonstrating advantage of simulation - - -## Background Material - -Here are some links to background material that you can use as a starting point for your project. - -Example: -- [Custom planning infrastructure](https://www.mathworks.com/help/nav/ref/nav.statevalidator-class.html#mw_e4f7cedb-14ed-440b-b5ed-5d9902e5f02f) -- [Co-simulation with Gazebo](https://www.mathworks.com/help/robotics/ug/perform-co-simulation-between-simulink-and-gazebo.html) - - [Simulate Mars Rover with Gazebo (Video)](https://www.youtube.com/watch?v=CqVXXirYJaM) - - [Different Gazebo worlds](https://clearpathrobotics.com/blog/2020/07/clearpath-robots-get-new-gazebo-simulation-environments/) -- [Popular planners](http://www.cs.cmu.edu/~maxim/classes/robotplanning_grad/lectures/RRT_16782_fall20.pdf) - - [Comparison Table](https://www.mathworks.com/help/nav/ug/choose-path-planning-algorithms-for-navigation.html) -- SLAM with Computer Vision Toolbox - - [Structure From Motion](https://www.mathworks.com/help/vision/ug/structure-from-motion-from-multiple-views.html) - - [Stereo SLAM](https://www.mathworks.com/help/vision/ug/stereo-visual-simultaneous-localization-mapping.html) - - [Monocular Visual SLAM](https://www.mathworks.com/help/vision/ug/monocular-visual-simultaneous-localization-and-mapping.html). -- SLAM with Lidar Toolbox - - [Map from Lidar data](https://www.mathworks.com/help/vision/ug/build-a-map-from-lidar-data-using-slam.html) - - [Map using segmented Lidar data](https://www.mathworks.com/help/lidar/ug/build-a-map-and-localize-using-segment-matching.html) - -Suggested readings: -- [1] Pivtoraiko, M. and A. Kelly. “Efficient Constrained Path Planning via Search in State Lattices.” (2005). -- [2] Howard, T. M. and A. Kelly. “Optimal Rough Terrain Trajectory Generation for Wheeled Mobile Robots.” The International Journal of Robotics Research 26 (2007): 141 - 166. - -## Impact - -Expand the frontiers of off-road exploration and navigation using mobile robots for precision agriculture, firefighting, search and rescue, and planetary exploration. - -## Expertise Gained - -Autonomous Vehicles, Computer Vision, Robotics, Image Processing, Mobile Robots, SLAM, UGV, Optimization - - -## Project Difficulty - -Bachelor, Master's, Doctoral -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/40) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[lgavshin](https://github.com/lgavshin) - -## Project Number - -209 diff --git a/projects/Autonomous Navigation for Vehicles in Rough Terrain/student submissions/Autonomous-Nav-Rough-Terrain b/projects/Autonomous Navigation for Vehicles in Rough Terrain/student submissions/Autonomous-Nav-Rough-Terrain deleted file mode 160000 index 417b5269..00000000 --- a/projects/Autonomous Navigation for Vehicles in Rough Terrain/student submissions/Autonomous-Nav-Rough-Terrain +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 417b52694045695007a0b74f411cbe57b5af6f0a diff --git a/projects/Autonomous Vehicle localization/README.md b/projects/Autonomous Vehicle localization/README.md deleted file mode 100644 index 1a260235..00000000 --- a/projects/Autonomous Vehicle localization/README.md +++ /dev/null @@ -1,66 +0,0 @@ -**Project 20:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Autonomous Vehicle Localization Using Onboard Sensors and HD Geolocated Maps

-

Revolutionize the current transportation system by improving autonomous vehicles localization for level 5 automation.

-
- -## Motivation - -Autonomous vehicles are revolutionizing the way the current transportation system works and many companies are investing on this mega trend technology to secure a share in this market. Researchers and engineers are combining efforts to achieve a full driving automation (Level 5) system that is safe and comfortable for the passengers. Localization is a key component of an autonomous vehicle to enable autonomous driving by processing sensors data combined with high definition maps for accurate results. - -## Project Description - -Using a simulation platform from MathWorks Automated Driving Toolbox™ (ADT), either Driving Scenario or Unreal-based simulator, -develop a demonstration integrating information from cameras, radar, Lidar, INS sensors as well as HERE HD map data to localize the vehicle -with high precision. Use anything at your disposal in the Automated Driving Toolbox and Navigation Toolbox™. -The following features might be useful while building a prototype of a localization approach: -- Point cloud processing capabilities in the Computer Vision toolbox™ (CVT) -- Visual Simultaneous Localization and Mapping (SLAM), Structure from motion (SFM), Visual Odometry (VO) frameworks from the CVT -- Driving scenario generation capabilities in ADT -- Sensors from ADT: radar, camera, lidar -- Synthetic sensors from Navigation Toolbox: IMU, GPS, INS - -The project as stated would have a very large scope and complexity. It would require that you add constraints to make it implementable in a realistic period of time. - -Suggested high-level steps: - -1. Build a simulation scenario with ego-vehicle and sensors using the Unreal Engine-based simulator in ADT. This simulation should be done in Simulink. - -2. Use data from simulated lidar, INS, and camera sensors to design algorithms for accurate estimation of the car’s pose potentially using SLAM. Optionally, involve use of High-definition maps (from HERE) to further enhance your localization algorithms (https://www.mathworks.com/help/driving/ref/herehdlmreader.html). - -3. Compare your results against ground truth derived from Unreal simulator. - -## Background Material - -- [Automated Driving Toolbox](https://www.mathworks.com/help/driving/) -- [Computer Vision Toolbox](https://www.mathworks.com/help/vision/) -- [Navigation Toolbox](https://www.mathworks.com/help/nav/) -- [Unreal Engine](https://www.unrealengine.com) -- [Unreal Engine Scenario Simulation](https://www.mathworks.com/help/driving/unreal-engine-scenario-simulation.html) - -## Impact - -Contribute to the change of automobile industry, and transportation system. - -## Expertise Gained - -Autonomous Driving, Computer Vision, Robotics, SLAM, State Estimation, Sensor Fusion and Tracking - -## Project Difficulty - -Master’s level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/3) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[thewitek](https://github.com/thewitek) - -## Project Number - -20 - diff --git a/projects/Battery Pack Design Automation/README.md b/projects/Battery Pack Design Automation/README.md deleted file mode 100644 index b7506ff4..00000000 --- a/projects/Battery Pack Design Automation/README.md +++ /dev/null @@ -1,72 +0,0 @@ -**Project 142:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Battery Pack Design Automation

-

Reduce the effort required to properly develop a battery pack and contribute to the global transition to zero-emission energy source.

-
- -## Motivation - -Batteries are everywhere, from your cell phone to car, and are becoming more and more common each passing day. -In seeing the potential to transform our key automotive, industrial, and robotics area due to increasing energy density the need arises to better automate the design of battery pack systems with customer specification driving the pack design based off of voltage, power, energy, and thermal key attributes. -Battery pack automation is a challenging problem because of the complexity of the system, downstream impacts on safety critical design, and in the end a detailed optimization problem. - - -## Project Description - -Work with the Powertrain Blockset™ product to automate the battery pack design using MATLAB® and Simulink® with the key characteristics being electrical, cooling, and mass. -The non-linear parameters will be derived using data fit optimization techniques such as Optimization Toolbox and Simulink Design Optimization. -Finally, a workflow that demonstrates battery pack design optimization using an FTP75 and other drive cycles will be developed. - -Suggested steps: -1. Use Lithium Ion battery technologies. -2. Perform a literature search prior to starting the work. -3. Create a 3RC Lithium Cell model with temperature and SOC as input factors and a thermal connection. https://www.mathworks.com/help/autoblks/ref/equivalentcircuitbattery.html -4. Fit the 3RC Lithium Cell using the Generate Parameter Data for Equivalent Circuit Battery Block: https://www.mathworks.com/help/autoblks/ug/generate-parameter-data-for-estimations-circuit-battery-block.html. -5. Develop a tool that will automatically assemble the Lithium Ion Cell block into modules and packs as part of a Simulink model. The tool should take as an input desired pack voltage, power, energy, module size, thermal connectivity for conduction and convection. Note, for thermal connectivity consider a cube module that has possible connections on all 6 sides. -6. Using the tool developed in 5, determine the optimal size of the battery pack that takes into account range, cost, volume, cooling, and mass constraints. For example, one optimal problem statement would be to maximize range while reducing mass and cost. Another optimal problem would be just to maximize range. The Powertrain Blockset EV reference application can be used as a system model. https://www.mathworks.com/help/autoblks/ug/explore-the-electric-vehicle-reference-application.html - -Advanced project work: - -Extend this work to Solid State Batteries. - - -## Background Material - -- [Powertrain Blockset](https://www.mathworks.com/products/powertrain.html#tradeoff) -- [Powertrain Block set Examples](https://www.mathworks.com/help/autoblks/examples.html?s_tid=CRUX_topnav) - -Suggested readings: - -- [1] Ahmed, R., J. Gazzarri, R. Jackey, S. Onori, S. Habibi, et al. "Model-Based Parameter Identification of Healthy and Aged Li-ion Batteries for Electric Vehicle Applications." SAE International Journal of Alternative Powertrains. doi:10.4271/2015-01-0252, 4(2):2015. -- [2] Gazzarri, J., N. Shrivastava, R. Jackey, and C. Borghesani. "Battery Pack Modeling, Simulation, and Deployment on a Multicore Real Time Target." SAE International Journal of Aerospace. doi:10.4271/2014-01-2217, 7(2):2014. -- [3] Huria, T., M. Ceraolo, J. Gazzarri, and R. Jackey. "High fidelity electrical model with thermal dependence for characterization and simulation of high power lithium battery cells." IEEE® International Electric Vehicle Conference. March 2012, pp. 1–8. -- [4] Huria, T., M. Ceraolo, J. Gazzarri, and R. Jackey. "Simplified Extended Kalman Filter Observer for SOC Estimation of Commercial Power-Oriented LFP Lithium Battery Cells." SAE Technical Paper 2013-01-1544. doi:10.4271/2013-01-1544, 2013. -- [5] Jackey, R. "A Simple, Effective Lead-Acid Battery Modeling Process for Electrical System Component Selection." SAE Technical Paper 2007-01-0778. doi:10.4271/2007-01-0778, 2007. -- [6] Jackey, R., G. Plett, and M. Klein. "Parameterization of a Battery Simulation Model Using Numerical Optimization Methods." SAE Technical Paper 2009-01-1381. doi:10.4271/2009-01-1381, 2009. -- [7] Jackey, R., M. Saginaw, T. Huria, M. Ceraolo, P. Sanghvi, and J. Gazzarri. "Battery Model Parameter Estimation Using a Layered Technique: An Example Using a Lithium Iron Phosphate Cell." SAE Technical Paper 2013-01-1547. Warrendale, PA: SAE International, 2013. - -## Impact - -Contribute to the global transition to zero-emission energy source. - -## Expertise Gained - -Sustainability and Renewable Energy, Control, Electrification, Optimization, Parallel Computing - -## Project Difficulty - -Master’s level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/13) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -142 - -## Proposed By - -[kgrand-mw](https://github.com/kgrand-mw) diff --git a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/DeepLearningModel.png b/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/DeepLearningModel.png deleted file mode 100644 index 72a66697..00000000 Binary files a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/DeepLearningModel.png and /dev/null differ diff --git a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/README.md b/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/README.md deleted file mode 100644 index b7105f7f..00000000 --- a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/README.md +++ /dev/null @@ -1,108 +0,0 @@ -**Project 202:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques

-

Leverage a deep learning approach to extract behavioral models of mixed-signal systems from measurement data and circuit simulation.

-
- -## Motivation - -Phase-locked loops (PLLs) are the heart of many electronic systems. PLL designs can be large and complex. They have a unique property that makes their analysis and simulation even just for their deterministic behaviour quite difficult. -System designers in a chip design company would like to be able to get estimates of the key performance metrics of a PLL (like lock time, phase noise, operating frequency) through less time-consuming simulation and without going too deep into the nitty gritty details of the building blocks. They need to make design decisions and choose a PLL architecture that will meet the specification given to them. They need to accomplish this without having to design all the PLL components in detail. Once this top-down design process is completed the system engineer would be simulating the PLL in detail and verify that the design indeed meets the specifications. - To summarize, an effective behavioural level model for a PLL will be quite effective in getting system engineer’s job done in a timely manner. Deep learning techniques can be helpful in achieving that goal. - -## Project Description - -Based on the discussion above, the key requirements for a deep learning based behavioral level model for a PLL should be as follows: -1. Model needs to be abstract enough so that results can be computed in a reasonable amount of time (deep learning models should be fast). -2. The model needs to be generic enough so that it should estimate the performance metrics even when detailed design of the building blocks is not available. -3. Model needs to capture non-ideal effects of the PLL so that we can use to estimate phase noise (a key metric in PLL design). - -To fulfill the above requirements we need to study various deep learning techniques available and pick the one that is going to be helpful in meeting the above 3 key requirements. Based on the PLL design that are prevalent (block diagram representation of a PLL is shown below) in the industry following are the key input parameters that a system designer might want to play with in choosing a PLL architecture that meets specifications: - -| ![diagram](blockDiagramPLL.png) | -|:--:| -| ***Figure1**: PLL block diagram* | - -1. Input clock frequency -2. Phase Frequency Detector (PFD): - - Dead-band compensation -3. Charge pump: - - Input current - - Leakage current -4. Loop filter: - - R and C values (number of components vary depending on which order filter has been used) -5. Voltage-controlled oscillator (VCO): - - Voltage sensitivity - - Phase noise -6. Clock divider value - -And following are the target metrics that a PLL system designer is interested in: -1. Lock time -2. Phase noise -3. Operating frequency - -Suggested steps: -1. Perform literature research prior to starting the work. One needs to familiarize oneself with the basics of PLL design and various Deep learning techniques available. -2. Generate training data to be fed to the deep learning model. One will be required to setup a [PLL model](https://www.mathworks.com/help/msblks/phase-locked-loop.html) using [Mixed-Signal Blockset™](https://www.mathworks.com/help/msblks/index.html) and setup Simulation-In-Loop to gather training data. - -| ![simulationLoop](SimulationInTheLoop.png) | -|:--:| -| ***Figure2**: PLL simualtion in loop* | - -3. Once the training data has been collected the modeling of the selected deep learning model (CNN, RNN, GAN etc..) will be the next step using [Deep Learning Toolbox™](https://www.mathworks.com/help/deeplearning/index.html?searchHighlight=deep%20learning%20toolbox&s_tid=srchtitle). - -| | -|:--:| -| ***Figure3**: Deep learning model* | - -4. Once the model has been trained the next step will be to test it using some test data that we can generate using our PLL model in Simulink®. - -Project variations: -1. Apply above steps for different mixed-signal systems like ADCs, DACs or CDRs. - -Advanced project work: -1. Create a deep learning model that takes in the target metrics as input and provides you the probable input parameter values that may meet the specification of the desired PLL. -2. Estimate sensitivity of the different parameters of the behavioral model, especially related to silicon technology. - - -## Background Material - -- [Mixed-Signal Blockset](https://www.mathworks.com/help/msblks/index.html?s_tid=srchtitle) -- [Deep Learning Toolbox](https://www.mathworks.com/help/deeplearning/index.html?searchHighlight=deep%20learning%20toolbox&s_tid=srchtitle) -- [Machine Learning for Electronic Design Automation](https://www.mathworks.com/videos/machine-learning-for-electronic-design-automation-1544592067829.html) -- [Center For Advanced Electronics Through Machine Learning](https://c3ps.gatech.edu/center-advanced-electronics-through-machine-learning-caeml) - -Suggested readings: - -[1] Razavi, Behzad. RF Microelectronics. Upper Saddle River, NJ: Prentice Hall PTR, 1998. - -[2] Banerjee, Dean. PLL Performance, Simulation and Design. Indianapolis, IN: Dog Ear Publishing, 2006. - -[3] B. Khailany et al., "Accelerating Chip Design With Machine Learning," in IEEE Micro, vol. 40, no. 6, pp. 23-32, 1 Nov.-Dec. 2020, doi: 10.1109/MM.2020.3026231. - - -## Impact - -Accelerate mixed-signal design and analysis thereby reducing Time-To-Market for semiconductor companies. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Machine Learning, Modeling and Simulation, Neural Networks, RF and Mixed Signal, Optimization, Signal Processing - -## Project Difficulty - -Master's, Doctoral - -## Proposed By - -[pragatikt](https://github.com/pragatikt) - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/32) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -202 diff --git a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/SimulationInTheLoop.png b/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/SimulationInTheLoop.png deleted file mode 100644 index 00542cd2..00000000 Binary files a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/SimulationInTheLoop.png and /dev/null differ diff --git a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/blockDiagramPLL.png b/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/blockDiagramPLL.png deleted file mode 100644 index 07c725a8..00000000 Binary files a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/blockDiagramPLL.png and /dev/null differ diff --git a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/student submissions/PLL-modelling b/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/student submissions/PLL-modelling deleted file mode 160000 index b963768d..00000000 --- a/projects/Behavioral Modelling of Phase-Locked Loop using Deep Learning Techniques/student submissions/PLL-modelling +++ /dev/null @@ -1 +0,0 @@ -Subproject commit b963768db5d631aea7f7edf3dca2c65acd71ee0b diff --git a/projects/Build a wireless communications link with software defined radio/README.md b/projects/Build a wireless communications link with software defined radio/README.md deleted file mode 100644 index 8d383fa6..00000000 --- a/projects/Build a wireless communications link with software defined radio/README.md +++ /dev/null @@ -1,93 +0,0 @@ -**Project 162:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Build a Wireless Communications Link with Software-Defined Radio

-

Gain practical experience in wireless communication by designing inexpensive software-designed radios.

-
- -## Motivation - -The world has gone wireless! So many bits flying through the air :-). We all communicate primarily wirelessly now, from our smartphones, to our cars and robots, even our appliances and smoke alarms. -Learn how to transmit digital data through the air using MATLAB, RF signals, and software-defined radio (SDR). -Industry needs engineers with wireless experience due to use of wireless signals in so many different applications. This project will give you practical experience transmitting -and receiving wireless signals with inexpensive software-defined radios. You will learn about propagation loss, synchronization loss, and many other practical aspects -of wireless data transmission. - - -## Project Description - -Using MATLAB or Simulink as a simulation platform, build a unidirectional wireless communications link. Transmit and receive over-the-air (OTA) signals with a suitable -software-defined radio such as a USRP or ADALM-Pluto. Use [Communications Toolbox™](https://www.mathworks.com/products/communications.html) capabilities to implement modulation, error control coding, and synchronization. -Choose your design criterion: maximize transfer speed, maximize reliability, or maximize confidentiality/security, or you can pick your own optimization criteria. - -Suggested steps: - -1. Select source data and break down into bundles of bits. - -2. Design OTA transmission scheme, including modulation type and optionally, forward error correction. - -3. Model a channel in software, with all the impairments expected in the actual link. - -4. Build corresponding receiver. Receiver will have to accomplish carrier frequency synchronization, timing synchronization, and frame synchronization. - -5. Decode the source bits. - -6. When the link works as expected in software, remove the channel model and transmit and receive using SDR hardware. - -7. Continue from there to optimize your design for selected criteria such as throughput, or reliability, or security, or your own design criterion. - -Project variations: - -1. Start with a single-carrier modulation scheme that requires no carrier synchronization, like DPSK. - -2. Heavily oversample the signal, so that timing synchronization is straightforward. - -3. Once that link is built and verified, reduce the oversampling so that a timing sync loop is required. - -4. Once the timing loop is verified, use a modulation scheme that does require carrier synchronization, like PSK or QAM. - -5. If you prefer to start with an OFDM system, consider using the Schmidl-Cox algorithm ([[1]](#schmidl)). - -Advanced project work: - -Develop an ad hoc network with multiple transceiver nodes and OFDM as the underlying PHY. See [Packetized Modem with Data Link Layer](https://www.mathworks.com/help/comm/ug/packetized-modem-with-data-link-layer.html) for reference. - -## Background Material - -- [Software-Defined Radio (SDR)](https://www.mathworks.com/discovery/sdr.html) -- [QPSK Transmitter and Receiver - MATLAB](https://www.mathworks.com/help/comm/ug/qpsk-transmitter-and-receiver.html) -- [QPSK Transmitter and Receiver - Simulink](https://www.mathworks.com/help/comm/ug/qpsk-transmitter-and-receiver-in-simulink.html) -- [QPSK Transmitter with USRP Hardware - MATLAB](https://www.mathworks.com/help/supportpkg/usrpradio/ug/qpsk-transmitter-with-usrp-r-hardware.html) -- [QPSK Receiver with USRP Hardware - MATLAB](https://www.mathworks.com/help/supportpkg/usrpradio/ug/qpsk-receiver-with-usrp-r-hardware.html) -- [QPSK Transmitter with USRP Hardware - Simulink](https://www.mathworks.com/help/supportpkg/usrpradio/ug/qpsk-transmitter-with-usrp-r-hardware-1.html) -- [QPSK Receiver with USRP Hardware - Simulink](https://www.mathworks.com/help/supportpkg/usrpradio/ug/qpsk-receiver-with-usrp-r-hardware-1.html) - -Suggested readings: -[1] Timothy M. Schmidl and Donald C. Cox, "Robust Frequency and Timing Synchronization for OFDM", IEEE Transactions on Communications, Vol. 45, No. 12, December 1997. - -## Impact - -Develop your own expertise in wireless technology and drive this megatrend forward, in industry and society. - -## Expertise Gained - -5G, Low-Cost Hardware, Modeling and Simulation, Signal Processing, Software-Defined Radio, Wireless Communication - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/18) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[mw-mmclernon](https://github.com/mw-mmclernon) - - -## Project Number - -162 - diff --git a/projects/Carbon Neutrality/README.md b/projects/Carbon Neutrality/README.md deleted file mode 100644 index 1769d0ec..00000000 --- a/projects/Carbon Neutrality/README.md +++ /dev/null @@ -1,57 +0,0 @@ -**Project 242:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Carbon Neutrality

-

Build a CO2 emission model from historical data and create a plan to achieve carbon neutrality in the future.

-
- -## Motivation - -We are threatened by the global climate changes. In 2023, the global climate will be abnormal, and many countries are already experiencing continuous extreme high temperature and many natural disasters at the same time. Excessive carbon dioxide emissions are the main cause of global climate change, which will bring about a series of problems such as melting glaciers, rising sea levels, high temperature and heat waves, and destruction of the ecological environment. Global realization of carbon neutrality is the only way to go under the background of climate warming, and it is also an inevitable choice for the sustainable development of the earth. Achieving carbon neutrality will realize the energy revolution and industrial transformation of transnational, cross-border, and global collaboration. In this context, analyzing and studying the carbon dioxide emissions of various countries (or regions) is of great significance to understand the trends of carbon emissions in various countries, formulate reasonable international cooperation programs, and achieve carbon neutrality on a global scale. - -## Project Description - -Suggested steps: - -1. Become familiar with the MATLAB based Econometrics, statistical learning examples listed in Background Material section below, -2. Understand the [dataset](https://zenodo.org/record/5569235#.Y9fx40HMJhG) and data cleaning -3. According to the data properties, select and build the time series model from the [Econometric Toolbox](https://www.mathworks.com/products/econometrics.html) on carbon dioxide emissions of the selected countries and predict the future values, -4. Suppose that we are convinced that carbon neutrality is attainable by 2050. Design the projected path of CO2 emissions in the next three decades. Use any of the methodologies from [Curve Fitting Toolbox™](https://www.mathworks.com/products/curvefitting.html) or from [Econometric Toolbox](https://www.mathworks.com/products/econometrics.html) (such as State space model). -5. Apply the mapping toolbox to build the movies to reflect the changes on global CO2 emissions. - -Advanced project work: - -1. Build an interactive app on presenting the time series model on CO2 emissions -2. Integrate with the deep learning or machine learning models on time series data analysis -3. Combine the data from different countries, construct a multidimensional model or design the statistics test to illustrate the importance of the international cooperation - - -## Background Material - -1. State-Space Models: https://www.mathworks.com/help/econ/state-space-models.html -2. Econometrics Toolbox: https://www.mathworks.com/products/econometrics.html -3. The Paris Agreement: https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement -4. [The Global Carbon Project's fossil CO2 emissions dataset](https://zenodo.org/record/5569235#.Y9fx40HMJhG) -5. Time Series: Modeling, Computation, and Inference 2nd edition, Raquel Prado, Marco A.R. Ferreia, Mike West, ISBN9781498747059 - - -## Impact - -Set up a strategy for carbon neutrality and consolidate the international collaboration. - -## Expertise Gained - -Computational Finance, Sustainability and Renewable Energy, Modeling and Simulation, Machine Learning - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/77) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -242 \ No newline at end of file diff --git a/projects/Change Detection in Hyperspectral Imagery/README.md b/projects/Change Detection in Hyperspectral Imagery/README.md deleted file mode 100644 index a1b063e6..00000000 --- a/projects/Change Detection in Hyperspectral Imagery/README.md +++ /dev/null @@ -1,80 +0,0 @@ -**Project 210:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Change Detection in Hyperspectral Imagery

-

Develop an efficient method for detecting small changes on Earth surface using hyperspectral images.

-
- -## Motivation - -The Earth’s surface undergoes constant changes over time due to several influences including ones from the increasing human population. Hence, an analysis of the changes is necessary for better management of natural resources and preventing disasters. Change detection is the procedure of analyzing the changes between two Hyperspectral images of the same topographical zone taken at two different times. -Hyperspectral imaging (HSI) is used in distinctive surveillance and monitoring applications like spread of woodland fires, diagnosing crop yield issues, and environment monitoring amongst others. HSI integrates imaging and spectroscopic techniques into one system and captures a set of monochromatic images containing many contiguous wavelengths that go beyond what the human eye can see. Each wavelength captures a different property from the scene such as color, intensity, material property, presence of water vapor, vegetation, etc. The hyperspectral image can then be processed and analyzed to infer and reason about the contents of the scene as well as biological and chemical processes. - - -## Project Description - -Using [Image Processing Toolbox™ Hyperspectral Imaging Library](https://www.mathworks.com/matlabcentral/fileexchange/76796-image-processing-toolbox-hyperspectral-imaging-library) develop a change detection method based on spectral signature analysis. Change detection methods identify anomalous changes between scenes by suppressing background and accentuating changed regions. Certain changes such as vegetation and illumination as well as mis-registration make the change detection process difficult. - -The developed technique should -1. addresses the problem of multiple changes -2. utilize the spectral information at each pixel -3. resolves the ambiguity at mixed pixels, pixels corresponding to regions that have multiple materials - -Links to sample datasets for demonstrating change detection are provided below. A sequence of suggested steps for starting with the downloaded data and generating the change detection mask are also presented below. - -Sample Datasets: -1. [ONERA satellite change detection](https://ieee-dataport.org/open-access/oscd-onera-satellite-change-detection) -2. [Hyperspectral change detection dataset](https://citius.usc.es/investigacion/datasets/hyperspectral-change-detection-dataset) -3. [GETNET](https://drive.google.com/file/d/1cWy6KqE0rymSk5-ytqr7wM1yLMKLukfP/view) - - -Suggested steps: -1. Download a suitable dataset either from the list above or other sources -2. Calibrate or correct atmospheric errors in the hyperspectral data (there are many functions for calibration and correction in MathWorks Hyperspectral Imaging Library) -3. Pseudo-binary change detection to extract initial changes, see reference [1,3] for more details -4. Automatically cluster the changed endmembers based on hierarchical analysis and spectral matching, see reference [4] for more details -5. Generate final change detection map - -Project variations: -Use deep learning methods to train an end-member detector from hyperspectral data. Leverage this network to segment the hyperspectral data and generate the change detection map. - - -## Background Material - -Hyperspectral imaging libraryExamples demonstrating how to use the hyperspectral library for relevant classification needs: -1. [Hyperspectral image classification](https://www.mathworks.com/help/images/classify-hyperspectral-image-using-sam-metric.html) -2. [Hyperspectral image classification using deep learning](https://www.mathworks.com/help/images/hyperspectral-image-classification-using-deep-learning.html) - -References: - -[1] S. Liu, L. Bruzzone, F. Bovolo and P. Du, “Hierarchical change detection in multitemporal hyperspectral images,” Geoscience and Remote Sensing, IEEE Transactions on, vol.53, no.1, pp:244–260, 2015. - -[2] Mahdi Hasanlou & Seyd Teymoor Seydi (2018): Hyperspectral change detection: an experimental comparative study, International Journal of Remote Sensing. - -[3] Huifu Zhuang, Zhixiang Tan, Kazhong Deng & Guobiao Yao (2018) Change detection in multispectral images based on multiband structural information, Remote Sensing Letters, 9:12, 1167-1176. - -[4] S. Liu, L. Bruzzone, F. Bovolo and P. Du, "Unsupervised hierarchical spectral analysis for change detection in hyperspectral images," 2012 4th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), Shanghai, China, 2012. - -[5] Seyd Teymoor Seydi & Mahdi Hasanlou (2017) A new land-cover match-based change detection for hyperspectral imagery, European Journal of Remote Sensing, 50:1, 517-533. - -## Impact - -Revolutionize the management of natural resources, monitoring, and preventing of disasters, going beyond what is visible to the naked eye. - -## Expertise Gained - -Computer Vision, Image Processing, Deep Learning - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/42) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -210 diff --git a/projects/Classify Object Behavior to Enhance the Safety of Autonomous Vehicles/README.md b/projects/Classify Object Behavior to Enhance the Safety of Autonomous Vehicles/README.md deleted file mode 100644 index 1c45adcf..00000000 --- a/projects/Classify Object Behavior to Enhance the Safety of Autonomous Vehicles/README.md +++ /dev/null @@ -1,81 +0,0 @@ -**Project 221:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Classify Object Behavior to Enhance the Safety of Autonomous Vehicles

-

Automatically classify behavior of tracked objects to enhance the safety of autonomous systems.

-
- -## Motivation - -Autonomous vehicles will transform transportation and change how we move around and receive goods. The safety of those systems is paramount. To be able to operate safely in a complex environment, the autonomous vehicle uses sensors to detect and track objects in its vicinity, for example pedestrians, bicyclists, and other vehicles. However, estimating the motion of objects around the autonomous vehicle is insufficient. The next level is for the autonomous vehicle to classify the behavior of the objects, whether safe or unsafe, with respect to it. - -Consider the following examples: -1. A safe driver that follows the rules of the road vs. an aggressive driver, such as changing lanes aggressively at various speeds that might pose a risk to the autonomous vehicle. -2. A pedestrian that walks safely by the side of the road vs. a young child chasing a ball that might suddenly start crossing the road. -Classifying the behavior of tracked objects can help the autonomous vehicle to predict the motion of these objects and plan accordingly. - - -## Project Description - -Use the [Automated Driving Toolbox™](https://www.mathworks.com/products/automated-driving.html) to simulate realistic scenarios that contain vehicles, pedestrians, and roads. Use the [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/products/statistics.html), [the Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html), [the Reinforcement Learning Toolbox™](https://www.mathworks.com/products/reinforcement-learning.html) or other toolboxes to learn the behaviors of safe and risky actors in the scenario. - -Suggested steps: -1. Create scenario sets for training: - 1. Identify a type of scenario, e.g., highway driving, pedestrian crossing the road, etc. - 2. Define parameters and characteristics of safe and risky objects in a scenario. - 3. Create a set of scenarios with objects (vehicles, pedestrians) exhibiting safe and risky behaviors and trajectories. - 4. Collect and label object motion and trajectories. This forms your ground truth. -2. Train learning algorithms to classify between safe and risky object behaviors in a scenario. - -Advanced project work: - -Use the scenario to simulate sensor data coming from the autonomous vehicle sensors. Use the [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html) to track the vehicles and pedestrians in the scene. Use the learned behaviors to classify safe and risky objects to test the ability of your trained algorithm to classify the behavior of tracked objects. Use the following steps: -1. Simulate autonomous vehicle sensors to collect sensor data. -2. Configure a tracking system to estimate the motion of the actors in the scenario. -3. Apply the learning algorithms you trained in the first part of the project to classify the behaviors of tracked objects. -4. Assess the robustness of your classifier to errors introduced by the sensing and tracking. - -Project variations: -1. Extend this work to autonomous aerial vehicles. -2. Extend this work to environments where humans and robots work together: manufacturing, warehouses, etc. -3. Extend this work to off-road conditions, e.g., agricultural, mining scenarios, etc. - - -## Background Material - -- [Examples on how to generate scenarios using Driving Scenario Designer and Unreal Engine](https://www.mathworks.com/help/driving/examples.html?category=scenario-simulation) -- [Driving Scenario Designer](https://www.mathworks.com/help/driving/ref/drivingscenariodesigner-app.html) -- [Unreal Engine Scenario Simulation](https://www.mathworks.com/help/driving/unreal-engine-scenario-simulation.html) -- [RoadRunner](https://www.mathworks.com/products/roadrunner.html) - -Suggested readings -- [Spatiotemporal Relationship Reasoning for Pedestrian Intent Prediction](https://stip.stanford.edu/) -- [Predicting Future Movements of Pedestrians and Autonomous Vehicles](https://www.gislounge.com/predicting-future-movements-of-pedestrians-and-autonomous-vehicles/) - - - -## Impact - -Make autonomous vehicles safer by classifying behaviors of objects around them. - -## Expertise Gained - -Artificial Intelligence, Autonomous Vehicles, Robotics, Drones, Deep Learning, Explainable AI, Machine Learning, Mobile Robots, Neural Networks, Reinforcement Learning, Sensor Fusion and Tracking, UAV, UGV, Automotive - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/53) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed by - -[Eladki](https://github.com/eladki) - -## Project Number - -221 diff --git a/projects/Classify RF Signals Using AI/README.md b/projects/Classify RF Signals Using AI/README.md deleted file mode 100644 index 548230f5..00000000 --- a/projects/Classify RF Signals Using AI/README.md +++ /dev/null @@ -1,75 +0,0 @@ -**Project 245:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Classify RF Signals Using AI

-

Use deep learning to classify wireless signals and perform real-world testing with software defined radios.

-
- -## Motivation - -AI is now becoming a mainstream technology, and is being used to develop new drugs, combat human trafficking, and help diagnose cancer. Some of these same techniques can be used to classify RF signals propagating through the air. Classifying RF signals is an important problem, because our spectrum is getting more crowded every year. Cellular, WiFi, Bluetooth, ZigBee, UWB, SatCom, radar, GPS, IoT – interference scenarios are cropping up all the time, and we need to know who is interfering with whom. AI can help. - -## Project Description - -Use a deep neural network to classify wireless signals that may interfere with one another, like 5G, LTE, WiFi, Bluetooth, and ZigBee. Perform real-world testing by transmitting and receiving those signals with software defined radios. - -Using the [Spectrum Sensing with Deep Learning to Identify 5G and LTE Signals]( https://www.mathworks.com/help/comm/ug/spectrum-sensing-with-deep-learning-to-identify-5g-and-lte-signals.html) example as a starting point, use transfer learning to adapt that example’s AI network to classify real-world signals that may interfere with one another. Examples include: -- Bluetooth and WiFi -- Radar and 5G -- ZigBee and Bluetooth -- ZigBee and WiFi -- Try your own! - -Suggested steps: - -1. Use MATLAB’s [Wireless Waveform Generator App](https://www.mathworks.com/help/comm/ref/wirelesswaveformgenerator-app.html) to generate two of the signals above. -2. Frequency translate one or both signals so that they do not overlap in frequency. Use MATLAB’s [Multiband Combiner](https://www.mathworks.com/help/comm/ref/comm.multibandcombiner-system-object.html) for this. -3. Follow the steps taken in the [Spectrum Sensing with Deep Learning to Identify 5G and LTE Signals](https://www.mathworks.com/help/comm/ug/spectrum-sensing-with-deep-learning-to-identify-5g-and-lte-signals.html) example: - 1. Create a data set to be used for training and validation. Create signals with a variety of impairments. - 2. Train your AI network on that generated data set. Use the existing network in the example and perform transfer learning. Learn more about transfer learning with these links: - 1. [Get Started with Transfer Learning](https://www.mathworks.com/help/deeplearning/gs/get-started-with-transfer-learning.html) - 2. [Transfer Learning with Deep Network Designer](https://www.mathworks.com/help/deeplearning/ug/transfer-learning-with-deep-network-designer.html) - 3. Validate your AI network with a small percentage of your generated data. - 4. Test using over-the-air signals, captured with an SDR. You can also use an SDR to transmit the waveform, or capture existing signals from nearby cell towers, WiFi routers, etc. The following radios could serve well for this purpose: - - 1. [USRP B2xx, N2xx, or N3xx](https://www.mathworks.com/hardware-support/usrp.html) - 2. [ADALM-PLUTO](https://www.mathworks.com/hardware-support/adalm-pluto-radio.html) - -Consider the following possibilities for advanced project work. Or perhaps even better, imagine some of your own possibilities and try them out! -- Train and test with signals that overlap in frequency. -- Filter one or both signals so that only a portion of their bandwidth is used to train the network. -- Improve the speed of the network. -- Simplify the network with pruning and quantization and still achieve the same classification performance as the original example. -- Change the AI network design using the [Deep Network Designer App](https://www.mathworks.com/help/deeplearning/deep-network-designer-app.html). Try to improve the classification accuracy. -- Use public data sets with the existing network or a modified one. One such data set can be found at [Wireless Intelligence](https://wireless-intelligence.com/#/home). -- Test the network in other bands, with other signals. Use the Worldwide Frequency Allocation Chart, the US Spectrum Allocation Table, or another country-specific frequency allocation chart to determine which signals might interfere with one another. -- Generate C/C++ code from the network and deploy it to an embedded processor. -- Generate HDL code using [Deep Learning HDL Toolbox](https://www.mathworks.com/products/deep-learning-hdl.html) and deploy it to an FPGA. - - -## Background Material - -- [Spectrum Sensing with Deep Learning to Identify 5G and LTE Signals](https://www.mathworks.com/help/comm/ug/spectrum-sensing-with-deep-learning-to-identify-5g-and-lte-signals.html) -- [Deep Learning in MATLAB](https://www.mathworks.com/help/deeplearning/ug/deep-learning-in-matlab.html) - - -## Impact - -Help to mitigate the ever-increasing RF interference problem in the developed world. - -## Expertise Gained - -5G, Artificial Intelligence, Deep Learning, Image Processing, Machine Learning, Neural Networks, Software-defined Radio, Wireless Communication - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/81) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -245 diff --git a/projects/Coastline Prediction using Existing Climate Change Models/README.md b/projects/Coastline Prediction using Existing Climate Change Models/README.md deleted file mode 100644 index c6967d0c..00000000 --- a/projects/Coastline Prediction using Existing Climate Change Models/README.md +++ /dev/null @@ -1,60 +0,0 @@ -**Project 229:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Coastline Prediction using Existing Climate Change Models

-

-
- -## Motivation - -Climate change poses a significant risk to coastal areas as rising global temperatures produce rising sea levels. -According to [NOAA]( https://oceanservice.noaa.gov/facts/sealevel.html), average global sea level rose 2.6 inches between 1993 and 2014. -Rising sea levels will produce more frequent inland flooding and pose a significant risk to the houses, roads, bridges, and other infrastructure located in many coastal cities. -Climate change models use many factors to generate region-specific sea level change and are essential to help communities assess the potential impact of rising sea levels, with applications for insurance, zoning, permitting, and potential re-location. - -## Project Description - -Work with [Mapping Toolbox™](https://www.mathworks.com/products/mapping.html) to develop a function to calculate and visualize coastlines due to rising sea levels. Write an example to connect to a climate data server and visualize predicted coastlines at specific years in the future. - -Suggested steps: -- Identify a coastal region of interest and find corresponding coastline and 3-D elevation data for the region. Find elevation data that is 10 meters or better resolution. Data sources vary for country and region; for example, a good source of elevation data for the United States is the [USGS National Map](https://apps.nationalmap.gov/). -- Develop a MATLAB function using Mapping Toolbox to import the map data and produce a new coastline definition given a change in sea level. Visualize the changing coastline on a map. -- Find a climate data server to obtain predicted sea levels for the coastal region. Climate data servers provide various data sets generated from climate models. A couple of climate data servers to consider include [NOAA Digital Coast]( https://coast.noaa.gov/digitalcoast/) and [Copernicus Climate Data Store](https://cds.climate.copernicus.eu/). See [Visualizations of Northern Hemisphere Sea Ice Concentration]( https://www.mathworks.com/matlabcentral/fileexchange/77542-visualizations-of-northern-hemisphere-sea-ice-concentration) for an example illustrating how to obtain and use data from the Copernicus Climate Data Store. -- Write an example that uses the climate data server’s predicted sea level and uses the function developed above to visualize predicted future coastlines, such as for 100 years into the future. - -Project variations: -- Analyze and visualize different or additional impacts associated with rising sea levels, such as new flood zones or population displacement. - -Advanced project work: -- Develop an app interface to enable the user to enter inputs like a sea level change value or future year and visualize the impact on a map. -- Extend the app to automate all access to the map data required for the calculations, including elevation data and the climate data. Use the automated access to enable the user to select an arbitrary coastline region to analyze instead of a pre-defined region. - -## Background Material - -- [Geographic Plots](https://www.mathworks.com/help/matlab/geographic-plots.html) to get started visualizing data on maps. -- [Import and Export](https://www.mathworks.com/help/map/file-import-and-export.html) for supported map data formats in Mapping Toolbox. -- [Visualize Aircraft Line-of-Sight Over Terrain](https://www.mathworks.com/help/map/visualize-aircraft-line-of-sight-over-terrain.html) for an example of importing and visualizing terrain data. -- [Sea Level Rise Viewer](https://coast.noaa.gov/digitalcoast/tools/slr.html) for an example web application of rising sea levels from NOAA. - - -## Impact - -Assess and plan for the potential impact of climate change. - -## Expertise Gained - -Sustainability and Renewable Energy, Modeling and Simulation - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/60) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -229 diff --git a/projects/Coastline Prediction using Existing Climate Change Models/student submissions/Climate-Change-Map b/projects/Coastline Prediction using Existing Climate Change Models/student submissions/Climate-Change-Map deleted file mode 160000 index 7c96f6bd..00000000 --- a/projects/Coastline Prediction using Existing Climate Change Models/student submissions/Climate-Change-Map +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 7c96f6bd179885af3decafc684d85a5b2c04da8a diff --git a/projects/Control, Modeling, Design, and Simulation of Modern HVAC Systems/README.md b/projects/Control, Modeling, Design, and Simulation of Modern HVAC Systems/README.md deleted file mode 100644 index 4cf90668..00000000 --- a/projects/Control, Modeling, Design, and Simulation of Modern HVAC Systems/README.md +++ /dev/null @@ -1,52 +0,0 @@ -**Project 195:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Control, Modeling, Design, and Simulation of Modern HVAC Systems

-

Model a modern HVAC system and design a controller to improve heating, cooling, ventilation, air quality, pressure, humidity, and energy efficiency.

-
- -## Motivation - -Society needs to create buildings and homes that are energy efficient and healthy for the inhabitants. Achieving this is challenging and requires devices to control air temperature, air humidity, and air quality while being energy efficient. The challenge is developing the systems and controllers to balance these many competing goals. Modeling, simulation, and control is key to developing optimized systems. - -## Project Description - -The goal is to develop a model of a modern HVAC system which is quite complex. The overriding development philosophy is to start simple and add detail. - -Suggested steps: -1. Get familiar with the two [Simscape™](https://www.mathworks.com/products/simscape.html)-based models of HVAC systems, provided in the background material, and explore results through simulation. -2. Search and learn various topics in HVAC systems. Some starting terms are heat pumps, energy recovery ventilators, forced hot water systems, air handlers, de-humidifiers, and humidifiers. This will give you an overview of some components of a HVAC system. -3. Build a simple model of a room, cold outside environment, and a heating device. Develop a simple controller in Simulink and Stateflow to control the temperature. -4. Add complexity to your model to include a varying environment and a cooling system. -5. Add further complexity to control all aspects of the room including temperature, air quality, humidity, and pressure. -6. Add complexity to the room to include a house with multiple rooms. - -## Background Material - -There are various Simscape based models of HVAC systems. Learn how these work and expand them to what you want to model and simulate. -1. [House Heating System](https://www.mathworks.com/help/physmod/simscape/ug/house-heating-system.html) -2. [Vehicle HVAC System](https://www.mathworks.com/help/physmod/simscape/ug/vehicle-hvac-system.html) -3. [Building and HVAC Simulation in MATLAB/Simulink](https://www.matlabexpo.com/content/dam/mathworks/mathworks-dot-com/images/events/matlabexpo/de/2017/gebaude-und-anlagensimulation-mit-matlab-und-simulink-am-beispiel-des-ffg-projekts-saluh.pdf). Provides good background material. -4. Simscape based paper – “Thermal Dynamic Modeling and Simulation of a Heating System for a Multi-Zone Office Building Equipped with Demand Controlled Ventilation Using MATLAB/Simulink” ([PDF](https://networked-embedded.de/es/index.php/staff-details/obermaisser.html?file=files%2Fpublications%2Fpapers%2Frp024_A223.pdf)). - -## Impact - -Contribute to the design and control of modern homes and buildings to preserve energy and healthy living environments. - -## Expertise Gained - -Sustainability and Renewable Energy, Modeling and Simulation, Electrification, Control - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/26) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -195 diff --git a/projects/Deep Image Prior for Inverse Problems in Imaging/README.md b/projects/Deep Image Prior for Inverse Problems in Imaging/README.md deleted file mode 100644 index b5ec5497..00000000 --- a/projects/Deep Image Prior for Inverse Problems in Imaging/README.md +++ /dev/null @@ -1,79 +0,0 @@ -**Project 244:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Deep Image Prior for Inverse Problems in Imaging

-

Use the Deep Image Prior to solve inverse problems in imaging.

-
- -## Motivation - -Inverse problems in imaging arise when we wish to recover the input image to a system that transforms or corrupts it, given some measurement of its output. Examples of this include: denoising, deblurring, super-resolution, in-painting, X-Ray tomography and many more. Typically, these problems are ill-posed and extra information is required to successfully recover the input from the measurement. This extra information is usually about the structure of the input image – it should usually belong to a class of `natural images’. Inverse problems in imaging are at the core of many industries; from medical devices to astronomical imaging to defense. - -The Deep Image Prior is a way to use a deep neural network to promote natural image structure in the solution to an inverse problem. It also has one enticing, and surprising, feature: no training, pre-training, or training data, is required. - -## Project Description - -Use the [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) to implement the Deep Image Prior and use it solve one or more inverse problems in imaging. - -The suggested steps are as follows: - -- Learn about the [Deep Image Prior](https://dmitryulyanov.github.io/deep_image_prior). - -- Choose one or more inverse problems to solve and implement the forward operator, e.g.: - - - Denoising -> forward operator adds noise to the input. - - - Deblurring -> forward operator blurs (convolves) the input. - - - Super-resolution -> forward operator downsizes the input image. - - - In-painting -> forward operator zeroes some areas in the input image. - -- Select a network architecture (e.g., [U-Net](https://www.mathworks.com/help/vision/ref/unetlayers.html)) or implement the exact network defined in the Deep Image Prior paper. - -- Optimize the network weights to minimize the appropriate cost function to recover the input image given a noisy vector as a fixed input. - -Advanced extensions: - -- Consider inverse problems involving a more complex or worse-posed forward model, for example: - - - X-Ray tomography -> forward operator is the [Radon transform](https://www.mathworks.com/help/images/ref/radon.html) of the input image. - - - Compressed sensing -> forward operator is a random matrix with fewer rows than columns. - - - Some combination of any of the suggest items above, or others (e.g., deblurring + denoising). - -- Test out different network architectures, different inputs, etc. - -## Background Material - -[What is an inverse problem](https://tristanvanleeuwen.github.io/IP_and_Im_Lectures/what_is.html) - -[Deep Image Prior](https://dmitryulyanov.github.io/deep_image_prior) and [Paper](https://sites.skoltech.ru/app/data/uploads/sites/25/2018/04/deep_image_prior.pdf) - -[Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) - -[U-Net in MATLAB](https://www.mathworks.com/help/vision/ref/unetlayers.html) - -[Custom training loop in MATLAB](https://www.mathworks.com/help/deeplearning/ug/define-custom-training-loops-loss-functions-and-networks.html) - -## Impact - -Implement the Deep Image Prior to provide high-quality solutions to inverse problems in imaging that are ubiquitous in industry. - -## Expertise Gained - -Artificial Intelligence, Computer Vision, Deep Learning, Image Processing, Machine Learning, Neural Networks, Optimization, Signal Processing - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/80) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -244 diff --git a/projects/Deep Learning for UAV Infrastructure Inspection/README.md b/projects/Deep Learning for UAV Infrastructure Inspection/README.md deleted file mode 100644 index b1c74c00..00000000 --- a/projects/Deep Learning for UAV Infrastructure Inspection/README.md +++ /dev/null @@ -1,66 +0,0 @@ -**Project 187:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Deep Learning for UAV Infrastructure Inspection

-

Automate the process of infrastructure inspection using unmanned aerial vehicles and deep learning.

-
- -## Motivation - -Infrastructure inspection (power lines, oil/gas pipelines, bridges, buildings, etc.) is a vital and safety-critical task, but the process can be dangerous and requires a lot of manual work. UAVs provide a tantalizing opportunity to automate this process, detect infrastructure faults early, and reduce risk to human inspectors. - -Detecting infrastructure faults in images captured from the drone is still mostly a manual process and requires a trained operator on-site. The next big challenge is to add intelligence to UAVs to recognize problems themselves through advanced computer vision and machine learning techniques. - -## Project Description - -Work with [UAV Toolbox](https://www.mathworks.com/products/uav.html), [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html?s_tid=srchtitle), and [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html) to train a deep neural network to recognize infrastructure faults. Use simulation to rapidly train your network. Then use transfer learning to enhance detection performance on real-world camera data. - -**Suggested Steps**: -1. Pick an application area and what kind of fault you want to detect, e.g., cracks in bridges, leaks in oil pipelines, etc. -2. Use Unreal Engine to construct a representative environment that you can use to simulate the chosen fault. You can check the [Unreal Marketplace](https://unrealengine.com/marketplace/en-US/store) to see if you can find relevant assets and scenes as a starting point. -3. Use [UAV Toolbox](https://www.mathworks.com/products/uav.html) to capture a large set of camera images (color images) and annotated images (images with object labels) from the Unreal simulation. You can either fly the UAV along a fixed trajectory or manually change the pose of the UAV to capture images from many different orientations. See the background material below for good example starting points. -4. When you save images, capture both normal scenarios (no infrastructure faults) as well as scenarios with known problems. Automate this data collection as much as possible. -5. Use Deep Learning Toolbox and Computer Vision Toolbox to train a deep learning network to detect the infrastructure fault. -6. Verify in simulated test scenarios that your trained network can detect faults. Calculate the precision and efficiency of your network. - -**Advanced project work**: -* Go from simulated to real images. Use online data sets for infrastructure faults or collect data from a physical UAV. This dataset can be much smaller than the simulated dataset. Use [transfer learning](https://www.mathworks.com/help/deeplearning/gs/get-started-with-transfer-learning.html) to ensure your network works well on real-world data. -* Deploy your trained network to a physical drone and run it during the UAV flight. For example, you can deploy the network to an NVIDIA Jetson with MathWorks tools. See the [GPU Coder™ Support Package for NVIDIA GPUs](https://www.mathworks.com/matlabcentral/fileexchange/68644-gpu-coder-support-package-for-nvidia-gpus?s_tid=srchtitle). -* Extend your network to also use data from other sensors, e.g., lidar, to increase the recognition performance. See the examples in [Lidar Toolbox™](https://www.mathworks.com/products/lidar.html). - - -## Background Material - -* [Simulate Simple Flight Scenario in Unreal Engine](https://www.mathworks.com/help/uav/ug/simulate-a-simple-flight-scenario-and-sensor-in-3d-environment.html) -* [Depth and Semantic Segmentation Visualization Using Unreal Engine](https://www.mathworks.com/help/uav/ug/depth-and-semantic-visual-with-ue4.html) -* [Create Simple Image Classification Network Using Deep Network Designer](https://www.mathworks.com/help/deeplearning/gs/create-simple-image-classification-network-using-deep-network-designer.html) -* [Bridge Crack Dataset](https://github.com/maweifei/Bridge_Crack_Image_Data) -* [Road Crack Image Database](https://github.com/cuilimeng/CrackForest-dataset) -* [Deploying a Deep Learning Network on NVIDIA Jetson Using GPU Coder](https://www.mathworks.com/videos/deploying-a-deep-learning-network-on-nvidia-jetson-using-gpu-coder-1506357891312.html) - - -## Impact - -Enhance safety and speed of infrastructure inspection across a wide range of industries. - -## Expertise Gained - -Computer Vision, Drones, Artificial Intelligence, Robotics, UAV, SLAM, Deep Learning - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/21) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[mw-rpillat](https://github.com/mw-rpillat) - -## Project Number - -187 diff --git a/projects/Deep Learning for UAV Infrastructure Inspection/student submissions/DL_for_UAV_Infrastructure_Inspection b/projects/Deep Learning for UAV Infrastructure Inspection/student submissions/DL_for_UAV_Infrastructure_Inspection deleted file mode 160000 index ed16d283..00000000 --- a/projects/Deep Learning for UAV Infrastructure Inspection/student submissions/DL_for_UAV_Infrastructure_Inspection +++ /dev/null @@ -1 +0,0 @@ -Subproject commit ed16d283419a9e3fc9ede4d093d4e788ccdd53dc diff --git a/projects/Digital Twin and Predictive Maintenance of Pneumatic Systems/README.md b/projects/Digital Twin and Predictive Maintenance of Pneumatic Systems/README.md deleted file mode 100644 index f4b7bb64..00000000 --- a/projects/Digital Twin and Predictive Maintenance of Pneumatic Systems/README.md +++ /dev/null @@ -1,86 +0,0 @@ -**Project 215:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Digital Twin and Predictive Maintenance of Pneumatic Systems

-

Predict faults in pneumatic systems using simulation and AI/machine learning.

-
- -## Motivation - -Pneumatic systems make use of compressed gas or pressurized air to create motion. They are widely used for different applications including processes like drilling, packing, assembly systems, and also in air brakes for heavy vehicles. A typical pneumatic system consists of several mechanical, thermal, and electrical components like compressor, filter, regulator, lubricator, pipes, directional control valves, PLCs, plunger, actuators, and heat exchanger. -This complex system can develop several kinds of faults over time, which are difficult to predict and diagnose. These include problems like air leakage, air choke, damaged filter, faulty valves, etc. All this can lead to production downtime and other business losses. Predictive Maintenance techniques using machine learning, provide promising possibilities to make data-driven decisions and take required maintenance, inventory planning, and repair actions in advance. However, training robust predictive maintenance algorithms requires a lot of sensor data that can effectively represent different scenarios of the system operation. Acquiring this data can be challenging and expensive in a real setting, especially data from faulty or degraded operations. -One promising solution to this problem is to generate synthetic training data from a system simulation, which can represent a variety of operating conditions and fault states. The simulation can even be tuned to a real system in operation as a Digital Twin, allowing for machine-specific predictions and various what-if scenarios. - -## Project Description - -Work with [Simscape™](https://www.mathworks.com/products/simscape.html) to develop a simulation model of a pneumatic system by parameterizing it for normal and faulty behaviors. Generate synthetic sensor data by running simulations under different conditions. The synthetic data should then be used to train predictive models using [Predictive Maintenance Toolbox™](https://www.mathworks.com/products/predictive-maintenance.html) and [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/products/statistics.html) for finding anomalous behavior, classifying fault type, and estimating remaining useful life. Finally, test the accuracy of predictive maintenance models on real data or unseen synthetic data. -Suggested Steps: -1. Do a literature survey of pneumatic systems being used for different applications, including the study of commonly occurring faults, component degradation, environmental conditions, and installed sensors that can help identify potential faults and anomalies. -2. Develop a [Multiphysics model of a commonly used pneumatic system in any application setting, using Simscape™](https://www.mathworks.com/help/predmaint/ug/generate-and-use-simulated-data-ensemble.html) including different mechanical, thermal, and electrical components. The model should be detailed enough to be able to incorporate commonly occurring faulty behaviors under different environmental conditions. -3. Generate synthetic sensor data from the model representing different system behaviors showing the normal operation, continuous degradation, and faulty operation, based on the literature survey. Parallelize the simulations using [Parallel Computing Toolbox™](https://www.mathworks.com/products/parallel-computing.html). -4. Develop predictive models using Statistics and Machine Learning Toolbox™ and Predictive Maintenance Toolbox™, from the synthetic data to: - a. Find anomalous behavior by applying different unsupervised machine learning techniques and compare - the results. - b. Classify faults with supervised machine learning techniques - c. Estimate remaining useful life before failure occurs. -5. Reflect upon the effectiveness and limitations of the proposed methodology. Also, comment on the utility of Simulation Methods and Digital Twins in Predictive Maintenance Applications. - -Project variations: - -Create a Digital Twin by tuning the simulation model parameters based on real data using Simulink Design Optimization™. The simulated asset will now act as Digital Twin of the real system in operation, allowing for machine-specific predictions and various what-if scenarios. - - -Advanced project work: -1. Find or measure real data to apply the predictive models and evaluate the results. You may refer to the dataset on [Air pressure system failures in Scania trucks](https://www.kaggle.com/uciml/aps-failure-at-scania-trucks-data-set). -2. Prototype Operations Optimization in real time, by deploying the Predictive Maintenance algorithm and Simscape model on cyber-physical embedded devices and cloud services, using Industrial IoT workflow concepts. Use the following pointers for inspiration: - a. Generation of raw sensor signals: Use a real machine or, run the Simscape Model in real time on Speedgoat computer or Raspberry Pi to generate sensor data and send it to Raspberry Pi ‘Edge device’ - b. Feature Extraction on Edge device: Perform feature extraction from sensor data on the Raspberry Pi ‘Edge device’ and stream the feature data to Thingspeak Cloud based service - c. Predictive Models running on cloud: Run your predictive models on Thingspeak to make predictions in real- - time about anomalous behavior, fault-type and remaining useful life. - - -## Background Material - -- [What is Predictive Maintenance?]( https://www.mathworks.com/discovery/predictive-maintenance-matlab.html) -- [Predictive Maintenance Toolbox](https://www.mathworks.com/products/predictive-maintenance.html) -- [Predictive Maintenance Video Series](https://www.mathworks.com/videos/series/predictive-maintenance-tech-talk-series.html) -- [Simscape](https://www.mathworks.com/products/simscape.html) -- [Digital Twin](https://www.mathworks.com/discovery/digital-twin.html) -- [Digital Twin for Industrial IoT]( https://www.mathworks.com/content/dam/mathworks/mathworks-dot-com/images/events/matlabexpo/online/2020/matlab-expo-2020-digital-twins-iiot.pdf) -- [Simulink Design Optimization](https://www.mathworks.com/products/sl-design-optimization.html) -- [Deploying Predictive Maintenance Algorithms to the Cloud and Edge](https://www.mathworks.com/company/newsletters/articles/deploying-predictive-maintenance-algorithms-to-the-cloud-and-edge.html) -- [Predictive Maintenance of a Duct Fan Using ThingSpeak and MATLAB](https://www.mathworks.com/videos/predictive-maintenance-of-a-duct-fan-using-thingspeak-and-matlab-1542018024279.html) -- [Parallel Computing Toolbox](https://www.mathworks.com/products/parallel-computing.html) - -Suggested readings: -- Wolfgang Gauchel*, Thilo Streichert, Yannick Wilhelm. 2020. Predictive Maintenance with a Minimum of Sensors using Pneumatic Clamps As An Example. 12th International Fluid Power Conference, Dresden -- P. Aivaliotis, K. Georgoulias & G. Chryssolouris (2019) The use of Digital Twin for predictive maintenance in manufacturing, International Journal of Computer Integrated Manufacturing, 32:11, 1067-1080, DOI: 10.1080/0951192X.2019.1686173 -- Aivaliotis, P., K. Georgoulias, Z. Arkouli, and S. Makris. 2019. “Methodology for Enabling Digital Twin Using Advanced Physics-based Modelling in Predictive Maintenance.” Procedia CIRP 81: 417–422. doi:10.1016/j.procir.2019.03.072. -- Axel Eriksson (2010). Detecting Leakages in the Pneumatic System of Heavy Vehicles Modelling Using Simulink - - -## Impact - -Improve efficiency and reliability of industrial processes - - -## Expertise Gained - -Artificial Intelligence, Industry 4.0, Cyber-Physical Systems, Digital Twins, Embedded AI, Health Monitoring, IoT, Machine Learning, Modeling and Simulation - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/46) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[rohit4849](https://github.com/rohit4849) - -## Project Number - -215 diff --git a/projects/Disturbance Rejection Control for PMSM Motors/README.md b/projects/Disturbance Rejection Control for PMSM Motors/README.md deleted file mode 100644 index 3914dbf4..00000000 --- a/projects/Disturbance Rejection Control for PMSM Motors/README.md +++ /dev/null @@ -1,74 +0,0 @@ -**Project 207:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Disturbance Rejection Control for PMSM Motors

-

Implement Active Disturbance Rejection Control (ADRC) algorithm for closed-loop speed control system for a Permanent Magnet Synchronous Motors (PMSM).

-
- -## Motivation - -Electric motors are found in appliances, industry automation, process control, cars and everywhere. -Permanent magnet synchronous motors (PMSM) are special type of brushless motors that offer advantages like high efficiency, high torque to weight ratio, high performance in both high and low speed of operation, and low maintenance over other motors. -Control algorithms are the key challenges in the motor control industry. Speed control of PSMS motors is commonly achieved by employing Proportional-Integral (PI) controllers. However, in certain applications (like in treadmills, electric vehicles, etc.), the unmodeled highly nonlinear dynamics of disturbances, due to sudden and frequent load variations, makes the use of PI controllers unsuited. Active Disturbance Rejection Control (ADRC) algorithm is a suitable strategy as it offers better dynamic performance for any sudden changes in the load. - -## Project Description - -Work with Motor Control Blockset™ product to implement an Active Disturbance Rejection Control (ADRC) for PMSM motor using MATLAB® and Simulink®. In Motor Control Blockset, you can find [reference examples for motor control with PI](https://www.mathworks.com/help/mcb/gs/field-oriented-control-acim-using-quadrature-encoder.html). Implement an Active Disturbance Rejection Control (ADRC) for motor control to reject the disturbances and compare the controller performance with conventional PI control for the sudden load changes. You can refer applications like treadmill or electric vehicle where sudden changes in the load is expected. - -Suggested steps: - -1. Understand the [reference examples for closed-loop speed control](https://www.mathworks.com/help/mcb/gs/field-oriented-control-acim-using-quadrature-encoder.html) in Motor Control Blockset. Simulate the example models with different motor loads for better understanding the dynamics of the system. -2. In reference example, add the load dynamics of a treadmill or similar application and observe the controller performance. -3. Implement ADRC for motor control in Simulink®. You can also try different control algorithms like reinforcement learning or Model predictive control for disturbance rejection control. -4. Simulate and ensure the controller meets the control gains and motor spins and tracks the reference speed meeting the control objective. Verify the controller performance for treadmill load dynamics. -5. Compare the ADRC controller performance with PI controller. It is likely to observe how disturbances are handled better in ADRC compared to PI controller. - -Project variations: - -Explore control algorithms other than ADRC like reinforcement learning, Model predictive control etc., and compare its performance in disturbance rejection with a conventional PI controller. - -Advanced project work: - -Extend this simulation work in hardware with small motor kit to experience the ADRC algorithm. You can refer the example [Control PMSM Loaded with Dual Motor (Dyno)](https://www.mathworks.com/help/mcb/gs/dual-motor-dyno-control-for-pmsm.html?searchHighlight=Control%20PMSM%20Loaded%20with%20Dual%20Motor&s_tid=srchtitle) for implementing the speed control and simulate the load characteristics in the motor coupled to the primary motor. - - -## Background Material - -Examples: -- [Motor Control Blockset Examples](https://www.mathworks.com/help/mcb/examples.html?s_tid=CRUX_topnav) -- [Surface-mount PMSM](https://www.mathworks.com/help/mcb/ref/surfacemountpmsm.html) -- [Estimate control gains](https://www.mathworks.com/help/mcb/gs/estimate-control-gains-from-motor-parameters.html) - -Suggested readings: - -[1] Danyun Lin; Wenguang Luo; Hao Zhang, “Active disturbance rejection controller of BLDCM in electric vehicle”, 2011 International Conference on Electrical Machines and Systems. - -[2] Zhiqiang Gao, Cleveland State University, U.S.A.; Bao-Zhu Guo, University of the Witwatersrand, South Africa, “Active Disturbance Rejection Control”, A Pre-Conference Tutorial at the 2014 IFAC World Congress. - -[3] Gernot Herbst, A Simulative Study on Active Disturbance Rejection Control (ADRC) as a Control Tool for Practitioners. - - -## Impact - -Improve the customer experience with advanced control strategies to handle the sudden changes in the motor load. - -## Expertise Gained - -Artificial Intelligence, Electrification, Control, Modeling and Simulation, Reinforcement Learning - - -## Project Difficulty - -Master's, Bachelor, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/38) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[AnanthKumarS](https://github.com/AnanthKumarS) - -## Project Number - -207 diff --git a/projects/Electrification of Aircraft/README.md b/projects/Electrification of Aircraft/README.md deleted file mode 100644 index 44cf4f61..00000000 --- a/projects/Electrification of Aircraft/README.md +++ /dev/null @@ -1,81 +0,0 @@ -**Project 200:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Electrification of Aircraft

-

Evaluate electric aircraft energy requirements, power distribution options, and other electrical technologies.

-
- -## Motivation - -Energy used for transport is a significant percentage of each person’s energy utilization. For example, in the UK, it is estimated that on average 32% of a person’s energy consumption is used for transport [Sustainable Energy – without the hot air - Current consumption](http://www.withouthotair.com/c19/page_116.shtml). - -Electrification of automotive vehicles is well underway, but there is still much to be done to electrify flight. The required energy density required for aircraft often comes from sources of combustion which are not sustainable or environmentally friendly. Electrification of aircraft provides opportunity for propulsion energy to come from clean and renewable green energy sources. To succeed it is essential that you provide clear, quantitative, data-driven conclusions underpinned by use of rigorous mathematical models and time-domain simulations. - -## Project Description - -You will use an existing MATLAB®, Simulink®, and [Simscape™](https://www.mathworks.com/products/simscape.html) representation of an all-electric aircraft as the basis for this project. You will extend this by building models of a variety of aircraft electrical energy storage and equipment which permit a thorough evaluation of energy consumption. Use the models to provide informed, data-driven, comparisons, and recommendations as to the most promising electrical configurations and technologies. - -Suggested steps: -1. Become familiar with existing electric aircraft models (links below) and use these as the basis for your project. -2. Project variations: Choose one of the following project ideas: - - Build or integrate a model of an energy storage system. Consider weight, size, and efficiency of one of: - - Hydrogen - - Fuel Cell - - Battery - - Other novel sources? - - Use the model to compare advantages of different distribution systems: - - AC - - DC - - Mixed AC/DC - - Build or integrate a more detailed and representative model of one of these loads: - - Propulsion - - Sensors and electrical actuators - - HVAC - - Galley/Hotel - - Infotainment - - Other areas? -3. Calculate expected efficiency and power requirements for a variety of typical flights -4. Write up data-drive recommendations to influence each of: - 1. Individuals – should technologies you investigated influence flight purchasing decisions by passengers? - 2. Industry – should the technology you investigated be further developed and why? - 3. Government – shape government policy to direct investment and multiply the benefits - -Advanced project work: -- Pick additional item/items from the project variations above. -- Parameterize the aircraft for multiple configurations: a variety of passenger capacities and multiple geographic locations. -- For comparative purposes, build one model of conventional - - Propulsion, or - - Actuation - -## Background Material - -- [Electric Aircraft Model in Simscape]( https://www.mathworks.com/matlabcentral/fileexchange/64991-electric-aircraft-model-in-simscape) -- [More-Electric-Aircraft-in-Simscape]( https://www.mathworks.com/matlabcentral/fileexchange/75289-more-electric-aircraft-in-simscape) -- [MathWorks News & Notes – Power Electronics for a More Electric Aircraft](https://uk.mathworks.com/content/dam/mathworks/tag-team/Objects/m/92984v00_NN2016_Fullbook.pdf) -- [Electrical Component Analysis for Hybrid and Electric Aircraft]( https://www.mathworks.com/help/aeroblks/Electrical-Component-Analysis-Hybrid-and-Electric-Aircraft.html) -- [Simscape Electrical Examples](https://www.mathworks.com/help/physmod/sps/examples.html) - -## Impact - -Contribute to the global transition to zero-emission energy sources by electrification of flight. - -## Expertise Gained - -Sustainability and Renewable Energy, Digital Twins, Electrification, Modeling and Simulation, Zero-fuel Aircraft - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/30) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[Andrew-Bennett-MW](https://github.com/Andrew-Bennett-MW) - -## Project Number - -200 diff --git a/projects/Electrification of Household Heating/README.md b/projects/Electrification of Household Heating/README.md deleted file mode 100644 index cc1b9e6e..00000000 --- a/projects/Electrification of Household Heating/README.md +++ /dev/null @@ -1,85 +0,0 @@ -**Project 201:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Electrification of Household Heating

-

Build and evaluate an electrical household heating system to help minimize human environmental impact and help to reduce climate change.

-
- -## Motivation - -In many temperate parts of the world, energy used for heating is a significant percentage of each person’s energy utilization. For example, in the UK, it is estimated that on average 32% of a person’s energy consumption is used for heating, see [Sustainable Energy – without the hot air - Current consumption](http://www.withouthotair.com/c19/page_116.shtml). - -Energy used for heating often comes from combustion of coal, gas, or oil which is not sustainable or environmentally friendly. Electrification of heating systems provides opportunity for heating energy to come from clean and renewable green energy sources. - - -## Project Description - -You will use MATLAB®, Simulink®, and [Simscape™](https://www.mathworks.com/products/simscape.html) to build a thermal model of a house which includes conduction, convection, radiation, and thermal inertia to permit time-domain simulation. Next, build a model of an electrical domestic heating system to permit thorough evaluation of energy consumption. Then import data from appropriate sources to validate and verify the model. Use the model to provide informed, data-driven, comparisons, and recommendations as to the most promising electrical heating technology. - -Suggested steps: -1. Become familiar with existing Simscape thermal models of a domestic house (links below) and use these as the basis for your project. -2. Investigate refrigeration examples referenced below to transform them into heating examples. -3. Project variations: Use Simscape to build a simulation model of one type of electrical domestic heating system: - - Electrical storage radiators, or - - Air-sourced heat-pump (with electric motor driving pump), or - - Ground-sourced heat pump (with electric motor driving pump), or - - Any other promising technologies discovered from literature review. -4. Calculate expected efficiency of the technology. -5. Integrate heating system into model of house and provide concrete daily and annual efficiency results. -6. Write up data-driven recommendations to influence each of: - 1. Individuals – should that type be purchased and what are the benefits? - 2. Industry – should that technology be further developed and why? - 3. Government – shape government policy to direct investment and multiply the benefits - -Advanced project work: -- Build models of more than one type of electrical domestic heating system. -- For comparative purposes, add models of other domestic heating sources, for example: - - Gas boiler - - Oil burner - - Conventional air conditioning unit - - Other renewable or nonrenewable options -- Parameterize the house for multiple configurations: a variety of building materials and multiple geographic locations. How much energy storage would be required to meet daily and annual demand in different parts of the world? -- Extend your model to simulate an off-grid system – how many solar panels or wind-turbines would be required to make a household self-sustainable? - - -## Background Material - -- [House Heating System - Basic model](https://www.mathworks.com/help/physmod/simscape/ug/house-heating-system.html) -- [House Heating System – More detailed model](https://www.mathworks.com/help/physmod/hydro/ug/house-heating-system.html) -- [Two-Phase Fluid Refrigeration](https://www.mathworks.com/help/physmod/simscape/ug/two-phase-fluid-refrigeration.html) -- [Residential Refrigeration Unit](https://www.mathworks.com/help/physmod/hydro/ug/residential-refrigeration-unit.html) -- [Sustainable Energy – without the hot air](https://www.withouthotair.com/) - - [Smarter heating](https://www.withouthotair.com/c21/page_140.shtml) - - [Heating II](http://www.withouthotair.com/cE/page_289.shtml) -- [Simscape](https://www.mathworks.com/help/physmod/simscape/index.html) -- [Simscape Examples](https://www.mathworks.com/help/physmod/simscape/examples.html) -- [Simscape Fluids Examples](https://www.mathworks.com/help/physmod/hydro/index.html) -- [Simscape Electrical Examples](https://www.mathworks.com/help/physmod/sps/examples.html) - - -## Impact - -Contribute to the global transition to zero-emission energy sources by electrification of household heating. - -## Expertise Gained - -Sustainability and Renewable Energy, Electrification, Digital Twins, Electrification, Modeling and Simulation - - -## Project Difficulty - -Master's - -## Proposed By - -[Andrew-Bennett-MW](https://github.com/Andrew-Bennett-MW) - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/31) to ask/answer questions, comment, or share your ideas for solutions for this project. - - -## Project Number - -201 diff --git a/projects/Energy-Optimal Trajectory Planning for Multirotor Drones/README.md b/projects/Energy-Optimal Trajectory Planning for Multirotor Drones/README.md deleted file mode 100644 index 08b028f6..00000000 --- a/projects/Energy-Optimal Trajectory Planning for Multirotor Drones/README.md +++ /dev/null @@ -1,59 +0,0 @@ -**Project 237:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Energy-Optimal Trajectory Planning for Multirotor Drones

-

Develop a trajectory planning for multirotor drones that minimizes energy consumption.

-
- -## Motivation - -Electric multi-rotor drones are becoming popular vehicles for variety of applications such as delivery, inspection, aerial photography, mapping and surveying, etc. Moreover, many companies are working on developing and deploying drones for passenger transport in urban areas commonly referred to as Urban Air Mobility (UAM). However, the onboard electric power source still provides a very limited operation time representing the main impediment for long distance and/or high payload rides. Thus, it is crucial to deliver, to a battery powered unmanned drone, a set of operations, including trajectory generation and following, to minimize the energy consumption increasing the success of the mission. - -## Project Description - -Develop an optimization-based trajectory planning framework for a multirotor drone that generates a trajectory between initial and final points with optimal energy consumption. The framework utilizes the dynamics of the quadrotor and the battery to generate the optimal trajectory. You may use an existing MATLAB® or Simulink ® representation of a quadcopter model as the basis for your project. You will extend this by developing an optimization-based trajectory planning. - -Suggested steps: - -1. Create a MATLAB or Simulink model of a multi-rotor drone, including the vehicle's airframe, motor, battery, and controller. You can leverage the [UAV Toolbox](https://www.mathworks.com/products/uav.html) or the already existing [quadcopter model](https://www.mathworks.com/matlabcentral/fileexchange/63580-quadcopter-drone-model-in-simscape?s_tid=srchtitle) in [Simscape Multibody™](https://www.mathworks.com/products/simscape-multibody.html) -2. Develop an objective function that takes into account the energy consumption of the drone (use the relation between battery and motor parameters, refer [1]). -3. Formulate a set of constraints that considers the quadrotor dynamics, motor dynamics, battery dynamics, initial and final states, and the state of charge ([SoC](https://www.mathworks.com/solutions/power-electronics-control/battery-state-of-charge.html)) of the battery. -4. Solve the constrained optimization problem using the Optimization Toolbox to get the reference angular speeds of the motors. (If you want to use the Simulink model reference example, learn how to use the Optimization Toolbox with Simulink) -Advanced project work -1. Include the arriving time in the objective function. -2. Include the battery state of health (SoH) as a constraint - - -## Background Material - -- [UAV Toolbox Examples](https://www.mathworks.com/help/uav/examples.html?category=planning-and-control&s_tid=CRUX_topnav) -- [Quadcopter Drone Model in Simscape](https://www.mathworks.com/matlabcentral/fileexchange/63580-quadcopter-drone-model-in-simscape?s_tid=srchtitle) -- [Four Bar Linkage Optimization in Simscape](https://www.mathworks.com/matlabcentral/fileexchange/62371-four-bar-linkage-optimization-in-simscape?s_tid=srchtitle) -- [Simscape Battery™](https://www.mathworks.com/products/simscape-battery.html) - -Suggested Reading: - -[1] Schacht-Rodríguez, R., Ponsart, J. C., García-Beltrán, C. D., Astorga-Zaragoza, C. M., Theilliol, D., & Zhang, Y. (2018). Path planning generation algorithm for a class of uav multirotor based on state of health of lithium polymer battery. Journal of Intelligent & Robotic Systems, 91(1), 115-131. - - -## Impact - -Increase mission time of multirotor drones. - -## Expertise Gained - -Drones, Robotics, Autonomous Vehicles, Electrification, Modeling and Simulation, Optimization, UAV - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/73) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -237 diff --git a/projects/Face Detection and Human Tracking Robot/README.md b/projects/Face Detection and Human Tracking Robot/README.md deleted file mode 100644 index 08c7cbfc..00000000 --- a/projects/Face Detection and Human Tracking Robot/README.md +++ /dev/null @@ -1,64 +0,0 @@ -**Project 214:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Face Detection and Human Tracking Robot

-

Design and implement a real time autonomous human tracking robot using low-cost hardware.

-
- -## Motivation - -Human-robot interaction is important in many computer vision applications, including activity recognition, automotive safety, smart home security applications and surveillance. -The task of a robot is to be a useful assistant to help people with their work. The robot must be able to interact with humans and to communicate well. For this condition, human face tracking system becomes the main requirement for vision system in this type of robots. Face detection can increase the Robot's ability for Human-Robot Interaction. Detecting and tracking human automatically with a sensor or an algorithm is a challenging problem due to the wide variety of positions, complexity of the system, and in the end a detailed optimization problem. This process involves extracting, selecting the best features and then tracking. In this project, you will explore, design and test to finding the optimal algorithm for face detection and tracking. - - -## Project Description - -Design and implement a low cost, user friendly, and real time autonomous human tracking robot using Android device and Deep Learning technology. The face detection is done by the Android device while the real-time control is done by either Arduino or Raspberry Pi based controller. -The face detection algorithm will be designed using [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html) and [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html). Work with the Simulink based Hardware support packages to deploy this algorithm on to hardware. An Android device will be used to run the computation intensive face detection algorithm which will then pass the information to Arduino or Raspberry Pi to control the movements of Robot. Finally, a workflow that demonstrates Deep Learning based Face detection and tracking will be developed. - -Suggested steps -1. Develop an AI based face detection and tracking application for an Android device using the Simulink support package for Android. Use the below examples as a reference for getting started with face detection using the Android device -a. [Detect and Track Face on Android Device](https://www.mathworks.com/help/supportpkg/android/ref/detect-and-track-face-on-an-android-device.html) -2. Design a robot based on Arduino or Raspberry Pi using the Simulink support packages for Arduino or Raspberry Pi. Use the below examples as a reference for getting started with creating a robot using Arduino/Raspberry Pi hardware -a. [Control a Raspberry Pi powered robot with MATLAB and Simulink](https://www.mathworks.com/matlabcentral/fileexchange/47376-control-a-raspberry-pi-powered-robot-with-matlab-and-simulink) -3. The Android application for face detection and recognition should be able to communicate the position of the human with respect to the robot on which it is mounted. Use the below examples as a reference for getting started with creating a robot using Arduino/Raspberry Pi hardware -a. [Control Raspberry Pi from your Android Device using Simulink](https:\www.mathworks.com\matlabcentral\fileexchange\59204-control-raspberry-pi-from-your-android-device-using-simulink) -4. After getting the coordinates of the human, the robot should move towards him/her and stop at a pre-defined distance. -5. The entire system should now track and move towards the human as an when they change their location. - - -## Background Material - -- [Computer Vision Toolbox](https://www.mathworks.com/products/computer-vision.html) -- [Deep Learning Toolbox](https://www.mathworks.com/products/deep-learning.html) -- [Deep Learning Toolbox Examples](https://www.mathworks.com/help/deeplearning/examples.html) -- [Simulink Support Package for Android Devices](https://www.mathworks.com/help/supportpkg/android/) -- [Simulink Support Package for Arduino Hardware](https://www.mathworks.com/hardware-support/arduino-simulink.html) -- [Simulink Support Package for Raspberry Pi Hardware](https://www.mathworks.com/hardware-support/raspberry-pi-simulink.html) - - -## Impact - -Leverage mobile technology and deep learning to advance face detection algorithms for impacting human safety and security. - -## Expertise Gained - -Artificial Intelligence, Computer Vision, Robotics, Deep Learning, Embedded AI, Human-Robot Interaction, Mobile Robots, Modeling and Simulation, Machine Learning, Low-cost Hardware, Image Processing, Control - - -## Project Difficulty - -Bachelor - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/45) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[dryouwu](https://github.com/dryouwu) - -## Project Number - -214 diff --git a/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-Car b/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-Car deleted file mode 160000 index bce5d270..00000000 --- a/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-Car +++ /dev/null @@ -1 +0,0 @@ -Subproject commit bce5d270f7d57c87bb1b4db35ecfa380734fe2a9 diff --git a/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-and-Human-Tracking-Robot b/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-and-Human-Tracking-Robot deleted file mode 160000 index 21e43432..00000000 --- a/projects/Face Detection and Human Tracking Robot/student submissions/Face-Detection-and-Human-Tracking-Robot +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 21e4343257935c66dcb48cde90f7efec671a34e2 diff --git a/projects/Flight Controller Design and Hardware Deployment/README.md b/projects/Flight Controller Design and Hardware Deployment/README.md deleted file mode 100644 index b47f40a3..00000000 --- a/projects/Flight Controller Design and Hardware Deployment/README.md +++ /dev/null @@ -1,84 +0,0 @@ -**Project 217:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Flight Controller Design and Hardware Deployment

-

Build a mini drone and use the PX4 Hardware Support package to design the flight controller using Simulink.

-
- -## Motivation - -Unmanned aerial vehicles (UAVs) are being used across a large variety of industries. Of the many industries, they are set to revolutionize transportation, delivery, agricolture, and surveillance. -MATLAB® and Simulink® are playing an important role by making it easy for customers to design UAVs for their applications. Simulink enables users to not only design and simulate algorithms but also quickly deploy these algorithms on a real UAV and fly the UAV. The end goal of this project is to show how a user can move from design to deployment for UAVs. - -## Project Description - -Work with [UAV Toolbox Support package for PX4® Autopilots](https://www.mathworks.com/help/supportpkg/px4/index.html) that enables you to leverage algorithms and peripherals from the PX4 middleware and deploy Simulink models to the PX4 Autopilot. Design a flight controller in Simulink and deploy it to a PX4 Autopilot based UAV and fly the UAV. - -Suggested steps: - -1. Use the [QAV250](https://shop.holybro.com/pixhawk-4-mini-qav250-kit_p1125.html) kit with the popular QAV250 frame to build a mini-UAV -2. Use the [Pixhawk 4 mini]( https://docs.px4.io/master/en/flight_controller/pixhawk4_mini.html) as the Autopilot for the UAV -3. Establish External mode communication via [radio](https://shop.holybro.com/transceiver-telemetry-radio-v3_p1103.html) between the Autopilot and PC. -4. Use the [PX4 Simulink blocks]( https://www.mathworks.com/help/supportpkg/px4/referencelist.html?type=block&listtype=cat&category=index&blocktype=all&capability=&s_tid=CRUX_topnav) to use the flight estimator from the PX4 middleware -5. Design the flight controller using Simulink and deploy it to PX4 Autopilot. Use External mode to tune parameters and monitor signals - -Project variations: - -Build a UAV plant model in Gazebo and use PX4 host target feature to design the flight controller in Simulink and deploy it as executable on the host and perform a Software in the loop simulation with Gazebo plant. - -1. Establish a connection between Gazebo and Simulink PX4 host target -2. Use PX4 host target for deployment -3. Use the PX4 Simulink blocks to use the flight estimator from the PX4 middleware -4. Design the flight controller using Simulink and deploy it as a PX4 host target. Use External mode to tune parameters and monitor signals - -Advanced project work: - -Setup AirSim that provides physical and virtual simulations, design flight controller in Simulink and run Software in loop using PX4 host target -Reuse a UAV plant model from [Airsim](https://docs.px4.io/master/en/simulation/airsim.html) and use [PX4 host targetPX4 host target](https://www.mathworks.com/help/supportpkg/px4/ug/deployment-using-px4hosttarget-jmavsim.html) feature to design the flight controller in Simulink and deploy it as executable on the host and perform a Software in the loop simulation with Airsim. - -1. Establish a connection between Airsim and Simulink PX4 host target -2. Use PX4 host target for deployment -3. Use the PX4 Simulink blocks to use the flight estimator from the PX4 middleware -4. Design the flight controller using Simulink and deploy it as a PX4 host target. Use External mode to tune parameters and monitor signals - -## Background Material - -Products: -- [UAV Toolbox Support Package for PX4 Autopilots](https://www.mathworks.com/help/supportpkg/px4/index.html?s_tid=CRUX_lftnav) - -- [UAV Toolbox](https://www.mathworks.com/help/uav/getstarted.html) - -- [Simulink Support Package for Parrot Mini drones](https://www.mathworks.com/help/supportpkg/parrot/?s_tid=srchbrcm) - -Examples: -- [Attitude Control with PX4 HSP on Host target](https://www.mathworks.com/help/supportpkg/px4/ref/attitude-control-px4-external-input.html) - -- [Position Tracking with PX4 HSP on Host target](https://www.mathworks.com/help/supportpkg/px4/ref/position-tracking-example.html) - -- [Accessing PX4 middleware using uORB blocks](https://www.mathworks.com/help/supportpkg/px4/ref/getting-started-uorb-blocks.html) - -- [Fly a Parrot mini drone using Simulink HSP](https://www.mathworks.com/help/supportpkg/parrot/ref/color-detection-and-landing-parrot-example.html) - - -## Impact - -Expedite UAV design and assembly with model-based design - - -## Expertise Gained - -Drones, Autonomous Vehicles, Control, Low-cost Hardware, UAV - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/48) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -217 diff --git a/projects/Green Hydrogen Production/README.md b/projects/Green Hydrogen Production/README.md deleted file mode 100644 index e0c11527..00000000 --- a/projects/Green Hydrogen Production/README.md +++ /dev/null @@ -1,58 +0,0 @@ -**Project 204:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Green Hydrogen Production

-

Develop a model of a reversible fuel-cell integrated into a renewable-energy microgrid structure.

-
- -## Motivation - -In the U.S., the main sources of greenhouse gas emissions come from transportation, electricity generation and industrial processes. Green hydrogen is produced from the electrolysis of water, powered by renewable energy sources, and is being considered as a pathway to decarbonize energy dense sectors such as aviation, and industrial manufacturing. Most hydrogen produced today is from reforming of natural gas, but this method releases greenhouse gas into the atmosphere. A key challenge is how to produce green hydrogen economically and in a repeatable way, given the variable nature of renewable energy and the need to also provide electricity to the grid. - -## Project Description - -You will develop a reversible fuel-cell model using [Simscape™]( https://www.mathworks.com/products/simscape.html), which has the ability to act as an electricity generator or electrolyzer. You will integrate this model into a renewable-energy microgrid structure and explore various operational profiles and system configurations that permit a thorough evaluation of energy and sizing requirements to meet certain hydrogen production goals. You will use the models to provide informed, data-driven comparisons and recommendations as to the most promising configurations and supporting technologies. - -Suggested steps: -1. Become familiar with existing models (links below) and use these as the basis for your project – in particular, build the model out to be a three-phase representation and integrate solar power and energy storage. -2. Create a reversible fuel-cell model using the Simscape Language and integrate into the model developed in 1. - -Project variations: -1. Build models of different fidelity levels, in order to evaluate scenarios ranging from longer-term system response over days or weeks to short-term behavior over micro-seconds to seconds to evaluate the impact of power electronic switching through to thermal response. -2. Calculate expected efficiency and power requirements for a number of operational scenarios. -3. Write up data-drive recommendations to influence each of: - 1. Industry – should the technology you investigated be further developed and why? - 2. Government – shape government policy to direct investment and leverage the benefits of clean technologies. - -## Background Material - -- [Solar Power Inverter](https://www.mathworks.com/help/physmod/sps/ug/solar-power-inverter.html?searchHighlight=solar%20power&s_tid=srchtitle) -- [Battery Pack Design Solution for Battery EVs in Simscape](https://www.mathworks.com/matlabcentral/fileexchange/82330-battery-pack-design-solution-for-battery-evs-in-simscape?s_tid=srchtitle) -- [Single-Phase Grid-Connected Solar Photovoltaic System](https://www.mathworks.com/help/physmod/sps/ug/single-phase-grid-connected-in-pv-system.html) -- [PEM Fuel Cell System](https://www.mathworks.com/help/physmod/simscape/ug/pem-fuel-cell-system.html?searchHighlight=fuel%20cell&s_tid=srchtitle) - -## Impact - -Contribute to the global transition to zero-emission energy sources through the production of hydrogen from clean sources. - -## Expertise Gained - -Sustainability and Renewable Energy, Electrification, Digital Twins, Modeling and Simulation - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/35) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[gdudgeon]( https://github.com/gdudgeon) - -## Project Number - -204 diff --git a/projects/Human Motion Recognition Using IMUs/README.md b/projects/Human Motion Recognition Using IMUs/README.md deleted file mode 100644 index b043cdb9..00000000 --- a/projects/Human Motion Recognition Using IMUs/README.md +++ /dev/null @@ -1,60 +0,0 @@ -**Project 232:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Human Motion Recognition Using IMUs

-

Use Deep Learning and Inertial Measurement Units (IMU) data to recognize human activities and gestures.

-
- -## Motivation - -IMUs are everywhere. We all have an IMU in our phones, and some in some of our watches. -With modern Machine Learning and Deep Learning techniques we can use body-worn IMUs to detect if someone is sleeping, walking, or running. Recent work has taken this further to determine full body pose and classify complex activities such as gestures. - -Motion recognition has obvious applications to consumer electronics in wearables, but also is useful in the medical and manufacturing sectors. With these techniques we can determine if a person’s gait is deteriorating because of a medical issue, or if a factory worker is moving in a way likely to cause an injury. - - -## Project Description - -This project focuses on simulating and detecting different categories of human motion using an IMU data along with Machine Learning and Deep Learning techniques. - -Suggested steps: -1. Record signals from an IMU strapped to the body. You can do this using the [Arduino Support Package for MATLAB](https://www.mathworks.com/hardware-support/arduino-matlab.html) or with the MATLAB Mobile App ([IoS](https://apps.apple.com/us/app/matlab-mobile/id370976661), [Android](https://play.google.com/store/apps/details?id=com.mathworks.matlabmobile&hl=en_US&gl=US)) on your phone. Be sure to follow all safety precautions if attaching electronics to the body. -2. Alternatively, use an open-source dataset for body worn IMUs. There are several public datasets for Human Activity Recognition (HAR) including -a. [HARTH Dataset](https://github.com/ntnu-ai-lab/harth-ml-experiments) -b. [Human Activity Recognition Using Smartphones Data Set](https://archive.ics.uci.edu/ml/datasets/human+activity+recognition+using+smartphones) -3. Use [Statistics and Machine Learning Toolbox]() and/or [Deep Learning Toolbox]() to classify categories of human motion in the data. Refer to [Human Activity Learning](https://www.mathworks.com/matlabcentral/mlc-downloads/downloads/submissions/50232/versions/9/previews/MachineLearningMadeEasy_FEX/HumanActivity/html/Human_Activity_Learning.html) for an example. Use this as a starting point and see if you can improve the accuracy of the classification using other datasets or a different technique. Test your classifier with data captured using MATLAB Mobile. -Project Variations: -1. Use several imuSensors worn on different parts of the body (or alternatively, use a dataset with several body worn IMUs). Use Deep Learning and Machine Learning to reconstruct more complex motions. Refer to [IMU-based Human Motion Capture Systems](https://ps.is.mpg.de/research_projects/imu-mocap) for ideas. - - -Advanced project work: -1. Use the imuSensor (in the Navigation Toolbox and Sensor Fusion and Tracking Toolbox) to recreate your recorded signals. Build a Human Motion Model (as in the diagram below) to drive the imuSensor object to mimic the IMU signals you used to train your classification algorithm in the first part of the project. To build the Human Motion Model, consider using the [OpenSim](https://simtk.org/projects/opensim/) modeling framework, or alternatively train an AI network to produce the desired imuSensor input . -2. Run your AI classification algorithms from the first part of the project on real hardware. Use this hardware to control MATLAB using the activities you’re AI algorithm can recognize. For example, [in this video](https://www.youtube.com/watch?v=RlomRYsP7Rg>) gestures are used to control MATLAB. - -## Background Material - -- Navigation Toolbox: [Introduction to Simulating IMU Measurements](https://www.mathworks.com/help/nav/ug/introduction-to-simulating-imu-measurements.html) -- [UC Irvine Human Activity Recognition Dataset](https://archive.ics.uci.edu/ml/datasets/human+activity+recognition+using+smartphones) -- [OpenSim project](https://simtk.org/projects/opensim/) - - -## Impact - -Enable the next generation of wearable electronic devices with motion recognition. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Embedded AI, Neural Networks, Signal Processing - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/66) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -232 diff --git a/projects/Human Motion Recognition Using IMUs/README.md.backup b/projects/Human Motion Recognition Using IMUs/README.md.backup deleted file mode 100644 index 18506a5f..00000000 --- a/projects/Human Motion Recognition Using IMUs/README.md.backup +++ /dev/null @@ -1,57 +0,0 @@ -**Project 232:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Human Motion Recognition Using IMUs

-

Use Deep Learning and Inertial Measurement Units (IMU) data to recognize human activities and gestures.

-
- -## Motivation - -IMUs are everywhere. We all have an IMU in our phones, and some in some of our watches. -With modern Machine Learning and Deep Learning techniques we can use body-worn IMUs to detect if someone is sleeping, walking, or running. Recent work has taken this further to determine full body pose and classify complex activities such as gestures. - -Motion recognition has obvious applications to consumer electronics in wearables, but also is useful in the medical and manufacturing sectors. With these techniques we can determine if a person’s gait is deteriorating because of a medical issue, or if a factory worker is moving in a way likely to cause an injury. - - -## Project Description - -This project focuses on simulating and detecting different categories of human motion using an IMU data along with Machine Learning and Deep Learning techniques. - -Suggested steps: -1. Record signals from an IMU strapped to the body. You can do this using the [Arduino Support Package for MATLAB](https://www.mathworks.com/hardware-support/arduino-matlab.html) or with the MATLAB Mobile App ([IoS] (https://apps.apple.com/us/app/matlab-mobile/id370976661), [Android](https://play.google.com/store/apps/details?id=com.mathworks.matlabmobile&hl=en_US&gl=US)) on your phone. Be sure to follow all safety precautions if attaching electronics to the body. -2. Alternatively, use an open-source dataset for body worn IMUs. There are several public datasets for Human Activity Recognition (HAR) including -a. [HARTH Dataset](https://github.com/ntnu-ai-lab/harth-ml-experiments) -b. [Human Activity Recognition Using Smartphones Data Set](https://archive.ics.uci.edu/ml/datasets/human+activity+recognition+using+smartphones) -3. Use [Statistics and Machine Learning Toolbox]() and/or [Deep Learning Toolbox]() to classify categories of human motion in the data. Refer to [Human Activity Learning](https://www.mathworks.com/matlabcentral/mlc-downloads/downloads/submissions/50232/versions/9/previews/MachineLearningMadeEasy_FEX/HumanActivity/html/Human_Activity_Learning.html) for an example. Use this as a starting point and see if you can improve the accuracy of the classification using other datasets or a different technique. Test your classifier with data captured using MATLAB Mobile. -Project Variations: -1. Use several imuSensors worn on different parts of the body (or alternatively, use a dataset with several body worn IMUs). Use Deep Learning and Machine Learning to reconstruct more complex motions. Refer to [IMU-based Human Motion Capture Systems](https://ps.is.mpg.de/research_projects/imu-mocap) for ideas. - - -Advanced project work: -1. Use the imuSensor (in the Navigation Toolbox and Sensor Fusion and Tracking Toolbox) to recreate your recorded signals. Build a Human Motion Model (as in the diagram below) to drive the imuSensor object to mimic the IMU signals you used to train your classification algorithm in the first part of the project. To build the Human Motion Model, consider using the [OpenSim](https://simtk.org/projects/opensim/) modeling framework, or alternatively train an AI network to produce the desired imuSensor input . -2. Run your AI classification algorithms from the first part of the project on real hardware. Use this hardware to control MATLAB using the activities you’re AI algorithm can recognize. For example, [in this video](< https://www.youtube.com/watch?v=RlomRYsP7Rg>) gestures are used to control MATLAB. - -## Background Material - -- Navigation Toolbox: [Introduction to Simulating IMU Measurements](https://www.mathworks.com/help/nav/ug/introduction-to-simulating-imu-measurements.html) -- [UC Irvine Human Activity Recognition Dataset](https://archive.ics.uci.edu/ml/datasets/human+activity+recognition+using+smartphones) -- [OpenSim project](https://simtk.org/projects/opensim/) - - -## Impact - -Enable the next generation of wearable electronic devices with motion recognition. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Embedded AI, Neural Networks, Signal Processing - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Number - -232 \ No newline at end of file diff --git a/projects/Improve the Accuracy of Satellite Navigation Systems/README.md b/projects/Improve the Accuracy of Satellite Navigation Systems/README.md deleted file mode 100644 index 2dceb66f..00000000 --- a/projects/Improve the Accuracy of Satellite Navigation Systems/README.md +++ /dev/null @@ -1,82 +0,0 @@ -**Project 192:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Improve the Accuracy of Satellite Navigation Systems

-

Improve the accuracy of satellite navigation systems by using non-binary LDPC codes

-
- -## Motivation - -In order to improve the accuracy of satellite navigation messages received at low Signal to Noise Ratio (SNR), non-binary low density parity check (LDPC) codes have been proposed in global navigation satellite systems. -The LDPC codes are capacity approaching codes and now supersede Turbo codes. Satellite navigation systems use smaller navigational messages and regular LDPC codes used in a general wireless communication system, -are not useful as they need longer message lengths. The navigation standards such as [Beidou](https://en.wikipedia.org/wiki/BeiDou) use smaller length non-binary or M-ary LDPC codes for forward error correction in navigation messages with M-bit code words. The structure of such LDPC codes is different from the regular LDPC codes that we use in other wireless standards and, encoder and decoding algorithms designed for them do not work. Other proprietary navigation systems are also proposing the non-binary LDPC codes as the forward error correction schemes. The structure of such LDPC codes is different from the regular LDPC codes that we use in other wireless standards and, encoder and decoding algorithms designed for them do not work. There is a need to develop the efficient encoder and decoder algorithms for non-binary LDPC codes used in satellite navigation systems for improving the position estimation accuracy. - -## Project Description - -Use [Communications Toolbox™](https://www.mathworks.com/help/comm/) to implement the non-binary LDPC encoder and decoder functions and benchmark the bit error rate (BER) performance in Additive White Gaussian Noise (AWGN) channel. -Implement an LDPC decoder to process the soft Log Likelihood Ratios (LLR) values at the receiver using iterative algorithms. -The basic navigation frame used in Beidou is as shown in the figure below. - -| ![navigationMessage](navigationMessage.png) | -|:--:| -| ***Figure1**: Navigation Message* | - -Each frame before error correction encoding has a length of 288 bits, containing PRN (6 bits), Message Type (Mestype, 6 bits), Seconds Of Week (SOW, 18 bits), message data (234 bits), and CRC check bits (24 bits). 64-ary LDPC(96, 48) encoding is applied on the navigation message and the resultant encoded frame length becomes 576 bits or 96 codewords (6-bit). The LDPC check matrix as described in [1], is a sparse matrix Ｈ48,96 of 48 rows and 96 columns defined in GF(26) domain with the primitive polynomial being p(x) = 1 + x + x6. The encoded data is BPSK modulated and 24 preamble symbols are prepended to form the transmitted frame. - -| ![transmittedMessage](transmittedMessage.png) | -|:--:| -| ***Figure2**: Transmitted message symbols* | - -At the receiver, assuming the perfect time and frequency synchronization, demodulated symbols are passed through 64-ary LDPC decoder using extended min-sum algorithm [1], implemented in GF(26) domain to extract the 48 codewords (288 bits) of Navigation message. - -Suggested steps: -- Implement 64-ary LDPC encoder in MATLAB following the steps given in Annex [1] using GF arithmetic from Communications Toolbox. -- Test this encoder data using the reference values provided in [1]. -- Implement a corresponding 64-ary LDPC decoder following the steps given in Annex [1]. -- Form the navigation message as shown in Figure 1 and pass it through 64-array LDPC encoder and perform BPSK modulation using [BPSK modulator](https://in.mathworks.com/help/comm/ref/comm.bpskmodulator-system-object.html) on the encoded data and pass through AWGN channel. -- Run a bit error rate (BER) simulation to benchmark the performance with standard provided results by replacing LDPC encoder and decoder functions in the [Communications Toolbox example](https://www.mathworks.com/help/comm/gs/accelerating-ber-simulations-using-the-parallel-computing-toolbox.html) with 64-ary LDPC encoder and decoder. - -Advanced project work: - -Profile the MATLAB code using [MATLAB profiler](https://in.mathworks.com/help/matlab/matlab_prog/profiling-for-improving-performance.html) to improve the speed of execution by comparing it with that of binary LDPC code from Communications Toolbox. -Implement M-ary LDPC encoder and decoder to support other message lengths defined in Beidou standard. - - - -## Background Material - -- Binary LDPC [encoder](https://www.mathworks.com/help/comm/ref/comm.ldpcencoder-system-object.html) and [decoder](https://www.mathworks.com/help/comm/ref/comm.ldpcdecoder-system-object.html) in [Communications Toolbox](https://www.mathworks.com/help/comm/) -- [Performance evaluation of binary LDPC coder in AWGN channel in Communications Toolbox](https://www.mathworks.com/help/comm/ref/comm.ldpcdecoder-system-object.html#mw_201f2d2d-1059-4774-8e70-4f1a9e0a7cdf) -- [Accelerating BER Simulations Using the Parallel Computing Toolbox](https://www.mathworks.com/help/comm/gs/accelerating-ber-simulations-using-the-parallel-computing-toolbox.html) -- [Profile Your Code to Improve Performance](https://www.mathworks.com/help/matlab/matlab_prog/profiling-for-improving-performance.html) - -Suggested readings: - -[1] [BeiDou Navigation Satellite System Signal In Space Interface Control, Aug 2017 (Sections 6.2.2.2, 6.2.2.3, 6.2.2.4, and Annex)](http://en.beidou.gov.cn/SYSTEMS/ICD/201806/P020180608522414961797.pdf) - -[2] Lin, S., & Costello, D. J. (1983). Error control coding: Fundamentals and applications. Englewood Cliffs, N.J: Prentice-Hall. - - -## Impact - -Accelerate the development of modern satellite navigation receivers. - -## Expertise Gained - -5G, GNSS, Wireless Communication - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/23) to ask/answer questions, comment, or share your ideas for solutions for this project. -## Proposed By -[nkchavali](https://github.com/nkchavali) - -## Project Number - -192 diff --git a/projects/Improve the Accuracy of Satellite Navigation Systems/navigationMessage.png b/projects/Improve the Accuracy of Satellite Navigation Systems/navigationMessage.png deleted file mode 100644 index b402c2d4..00000000 Binary files a/projects/Improve the Accuracy of Satellite Navigation Systems/navigationMessage.png and /dev/null differ diff --git a/projects/Improve the Accuracy of Satellite Navigation Systems/transmittedMessage.png b/projects/Improve the Accuracy of Satellite Navigation Systems/transmittedMessage.png deleted file mode 100644 index 187ae1d1..00000000 Binary files a/projects/Improve the Accuracy of Satellite Navigation Systems/transmittedMessage.png and /dev/null differ diff --git a/projects/Intelligent Fan Air Cooling System/README.md b/projects/Intelligent Fan Air Cooling System/README.md deleted file mode 100644 index 1576792b..00000000 --- a/projects/Intelligent Fan Air Cooling System/README.md +++ /dev/null @@ -1,88 +0,0 @@ -**Project 161:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Intelligent Fan Air Cooling System

-

Design an intelligent fan cooling system to moderate temperatures in a building to eliminate or reduce the need for air conditioning systems.

-
- -## Motivation - -More than 90% of all homes in United States and Japan have air conditioners [1]. By 2050, 2/3 of the world’s households could have an air conditioner. Air conditioners use about 6% of all the electricity produced in the United States, at an annual cost of about $29 billion to homeowners [2]. As a result, roughly 117 million metric tons of carbon dioxide are released into the air each year from the US alone. - -Extractor or two-way fan systems have the potential to reduce energy over more complex conditioner systems. The cooler evening and night-time air can be exchanged with warmer interior air. Your task is to design an intelligent control system that will maximize the time within the desired temperature range using models for airflow and capacitive building effects. - -## Project Description - -Night-time temperatures are generally lower than daytime temperatures. Model a simple fan mechanism that pumps-in or extracts air based on temperature differential between inside and outside. The cooler night-time air can be used to bring down the temperature of air and capacitive elements within the house. Over the course of the day, the air and the cooler structure will keep temperatures lower, until it crosses the evening or night-time temperatures. - -Develop a simulation model and an intelligent control mechanism for a simple fan system. The fan system could be: -- A [twin fan system](https://en.wikipedia.org/wiki/Window_fan), or -- An extractor fan (e.g., [window](https://en.wikipedia.org/wiki/Window_fan) or [whole-house](https://en.wikipedia.org/wiki/Whole-house_fan)) - -Use temperature measurements from outside and inside the building and model building/room capacity as well as seasonal weather patterns to optimally maintain a temperature inside the building within a defined comfort range. - -Create simulation models for the thermal behavior of a building, fan system, and control system. - -**Part 1:** Use Simulink and Simscape to build a dynamic heat and airflow simulation model for a building that includes: - -- Convective heat transfer: airflow between rooms and to the outside (via drafts, etc.) -- Conductive heat transfer: via walls, windows, doors and capacitive elements within the house. - -This could be a simple model (e.g., [3], [4]) or a more complex model (e.g., [4]). - -**Part 2:** Devise and model a one or two-way fan cooling device and its influence on the room temperature. Model air flow using a physics model. Consider calibrating the parameters with measured data. - -**Part 3:** Develop a control system to maintain a temperature as close as possible to a specified goal or within a desired range based on temperature readings from inside and outside of the building at a specified location in the building. - - -**Advanced project work:** - -Modeling and control -- Use predicted weather temperatures to design a control system that will maximize the time within the desired temperature range. -- Consider additional temperature measurement points for improved control. -- Develop a compressor-based air conditioner model using physics model or actual measured data for comparison of energy usage. -- Compare effectiveness and energy use vs a traditional air conditioning system or a two-way vs extractor fan system. -- Consider what a user can prescribe: preferred temperature, or temperate minimums and maximums that they can accept. -- Consider adding humidity measurements and ranges as an additional variable. - -Analytics -- Using weather models, predict which locations in the world this mechanism would be most effective. -- Predict global energy savings and carbon footprint reduction could be vs traditional air conditioner. - -Prototype -- Prototype your fan system using readily available components and an Arduino (or equivalent) and the [Arduino Support package for Simulink](https://www.mathworks.com/hardware-support/arduino-simulink.html). - - -## Background Material - -- [1] IEA (2018), [The Future of Cooling](https://www.iea.org/reports/the-future-of-cooling.), IEA, Paris -- [2] Department of Energy, [Air Conditioning Overview](https://www.energy.gov/energysaver/home-cooling-systems/air-conditioning#:~:text=Three%2Dquarters%20of%20all%20homes,into%20the%20air%20each%20year.conditioning) -- [3] [Simulink thermal model of a house](https://www.mathworks.com/help/simulink/slref/thermal-model-of-a-house.html), Simulink Reference Manual -- [4] [Simscape House Heating System](https://www.mathworks.com/help/physmod/simscape/ug/house-heating-system.html), Simscape Reference Manual -- [5] [Building and HVAC Simulation in MATLAB/Simulink](https://it.mathworks.com/content/dam/mathworks/mathworks-dot-com/solutions/aerospace-defense/files/2017/expo-de/gebaude-und-anlagensimulation-mit-matlab-und-simulink-am-beispiel-des-ffg-projekts-saluh.pdf) – FFG Project SaLüH!, MATLAB Expo, Munich, 2017 - - -## Impact - -Contribute to energy and carbon footprint reduction. - -## Expertise Gained - -Sustainability and Renewable Energy, Control, Modeling and Simulation, Optimization - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/17) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[mw-agrace](https://github.com/mw-agrace) - -## Project Number - -161 diff --git a/projects/Landslide Susceptibility Mapping using Machine Learning/README.md b/projects/Landslide Susceptibility Mapping using Machine Learning/README.md deleted file mode 100644 index ab855241..00000000 --- a/projects/Landslide Susceptibility Mapping using Machine Learning/README.md +++ /dev/null @@ -1,55 +0,0 @@ -**Project 228:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Landslide Susceptibility Mapping using Machine Learning

-

Develop a tool to identify and visualize geographical areas susceptible to landslides.

-
- -## Motivation - -The [United States Geological Survey (USGS)](https://www.usgs.gov/) reports that the worldwide death toll due to landslides is in the thousands with most fatalities due to rock falls, debris flows, or volcanic debris flows. A Landslide Susceptibility Map (LSM) is an effective visualization for identifying the relative likelihood of future landslides in low and high-risk landslide regions. These maps can aid with understanding the potential for future landslides that can have devastating impacts on cites, power grids, transportation, and most importantly people. - -## Project Description - -Develop a complete MATLAB-based example that generates a Landslide Susceptibility Map as seen in [1]. The example would obtain the historical landslide data e.g., distance from roads, distance from geological faults, distance from water streams, land use type, etc. and apply the three-stage methodology (outlined in Figure 2 of [1]) to generate and visualize a geographic map, -color-coded with high and low landslide risk regions (as see in Figures 5 and 6 of [1]). - -Suggested steps: -1. Identify the regions for which required landslide data exists e.g., Muş, Turkey, Bandar Torkaman, Iran, etc. -2. Fetch the required training data. You can adopt the tool [here](https://github.com/jeffreyevans/GradientMetrics) as per Reference [1]. -3. Implement the three-stage methodology as seen in Figure 2 of [1]. - 1. Pre-process spatial data using MATLAB® - 2. Develop a cascade neural network model (cascadeforwardnet using trainlm training) using the [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) - 3. Use the [Mapping Toolbox™](https://www.mathworks.com/products/mapping.html) to generate the risk map using a total of 24 variables - -## Background Material - -References: -- [1] [Abujayyab, Sohaib K. M. and Azlan Saleh. “Landslides Risk Prediction Using Cascade Neural Networks Model at Muş In Turkey.” (2020).](https://iopscience.iop.org/article/10.1088/1755-1315/540/1/012081/pdf). -- [2] [Vakhshoori, Vali, Hamid R. Pourghasemi, Mohammad Zare, and Thomas Blaschke. 2019. "Landslide Susceptibility Mapping Using GIS-Based Data Mining Algorithms" Water 11, no. 11: 2292. https://doi.org/10.3390/w11112292](https://www.mdpi.com/2073-4441/11/11/2292/pdf) - -Examples and tutorials: -- ArcGIS Geomorphometry and Gradient Metrics ([Documentation](https://evansmurphy.wixsite.com/evansspatial/arcgis-gradient-metrics-toolbox), [GitHub repository](https://github.com/jeffreyevans/GradientMetrics)) -- [What is a Convolutional Neural Network?](https://www.mathworks.com/discovery/convolutional-neural-network-matlab.html) - -## Impact - -Identify areas that are at risk for landslides to help mitigate devastating impacts on people and infrastructure. - -## Expertise Gained - -Sustainability and Renewable Energy, Machine Learning - - -## Project Difficulty - -Bachelor - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/62) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -228 diff --git a/projects/MIMO Engine Airpath Control/README.md b/projects/MIMO Engine Airpath Control/README.md deleted file mode 100644 index 632e6193..00000000 --- a/projects/MIMO Engine Airpath Control/README.md +++ /dev/null @@ -1,64 +0,0 @@ -**Project 45:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

MIMO Engine Airpath Control

-

Internal combustion engines will continue to be used in the automotive marketplace well into the future. Build a MIMO airflow control to improve, engine performances, fuel economy, and emissions, and start your career in the automotive industry!

-
- -## Motivation - -Internal combustion engines will continue to be used in the automotive marketplace well into the future. -However, engine performance, fuel economy, and emissions must continually be improved to keep them viable relative to more expensive full electric power sources. -For example, modern gasoline and diesel engines use direct fuel injection which can result in particulate matter emissions at low engine loads with particles small -enough to enter the human bloodstream directly through the lungs, so control of particulates is a major area of focus in diesel and gasoline engine control design today. -Transient air-fuel ratio control is critical to the particulate matter problem. - -## Project Description - -The ability to both estimate and control the airflow into the engine is at least half of the Air-Fuel Ratio (AFR) control problem that is so important to good engine performance, fuel economy, and emissions. -Since multiple actuators , e.g. throttle, wastegate, cam-phasers, Gas Re-circulation (EGR), affect the airflow into the engine, multi-input/multi-output control (MIMO) of engine airflow is an attractive way to improve engine airflow control and estimation performance that has yet to be implemented fully in the marketplace. General Motors is the first company known to implement a MIMO control approach in low-cost production ECU hardware using Model-Predictive Control (https://www.sae.org/publications/technical-papers/content/2018-01-0875). - -Suggested steps: - -1. Replace the existing schedule-based non-MIMO engine air system controls within the Powertrain Blockset Spark-Ignition (SI) and Compression-Ignition (CI) Engine Controller subsystems in the MathWorks Powertrain Blockset Conventional Vehicle Reference Application with a MIMO control approach similar to but more generic than that implemented by GM. -2. Calculate and present ability to arrive at pre-determined EGR flow and boost setpoints better than with the existing open-loop table-based controls during significant engine transients on USFTP75 and all shipping drive-cycles, resulting in better open-loop AFR control on the drive-cycle. -3. Provide and present solution for reproduction of results and understanding of designed approach. - -After designing and testing the proposed MIMO controller you can provide: -- A drop-in modification of the existing SI and CI engine control blocks in the Powertrain Blockset Conventional Vehicle Reference Example. Model-Predictive Control is required using Model Predictive Control Toolbox. -- A written review of simulation results showing significantly improved transient boost/EGR target control during transient operation in Simulation Data Inspector tool. - - -## Background Material - -- [Powertrain Blockset](https://www.mathworks.com/products/powertrain.html) - -- [Model Predictive Control of Diesel Engine Airpath](https://www.mathworks.com/videos/model-predictive-control-of-diesel-engine-airpath-81995.html) - -- [Model Predictive Control of Turbocharged Gasoline Engines for Mass Production (GM article)](https://www.sae.org/publications/technical-papers/content/2018-01-0875) - -- [Robust Model Predictive Control of a Diesel Engine Airpath](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.457.5766&rep=rep1&type=pdf) - -## Impact - -Improve environmental friendliness of engine control by tier 1 automotive supplier. - -## Expertise Gained - -Autonomous Driving, Control, Modeling and Simulation - -## Project Difficulty - -Master’s, Doctoral level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/10) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[pjmalo](https://github.com/pjmalo) - -## Project Number - -45 diff --git a/projects/MIMO Engine Airpath Control/block.png b/projects/MIMO Engine Airpath Control/block.png deleted file mode 100644 index 63af53a9..00000000 Binary files a/projects/MIMO Engine Airpath Control/block.png and /dev/null differ diff --git a/projects/MIMO Engine Airpath Control/students submissions/T513---SIEngineDynamometer b/projects/MIMO Engine Airpath Control/students submissions/T513---SIEngineDynamometer deleted file mode 160000 index 798b4646..00000000 --- a/projects/MIMO Engine Airpath Control/students submissions/T513---SIEngineDynamometer +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 798b4646fc686e98a5fdbadcd98559023bd7d2fa diff --git a/projects/Machine Learning for Motor Control/README.md b/projects/Machine Learning for Motor Control/README.md deleted file mode 100644 index 6539410a..00000000 --- a/projects/Machine Learning for Motor Control/README.md +++ /dev/null @@ -1,71 +0,0 @@ -**Project 218:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Machine Learning for Motor Control

-

Enhance the performance and product quality required to develop a motor control application.

-
- -## Motivation - -Motor control is one of the core skillsets in robotics and electrification areas which are becoming more and more widely used in the industry. Currently, many industrial motor applications are driven by classical and robust control-based methods. These methods are also cost-effective to run on embedded systems. Conventional control approaches are effective when the system can be modelled predictably. It can be difficult to predict system nonlinearities due to motor parameter changes caused by aging and temperature variation. Therefore, implementing Machine Learning-based motor control methods may provide an alternate pathway to overcome the real-world challenges. - -## Project Description - -First, build a motor drive system based on conventional control approach using Simulink® and/or Simscape™. Then, introduce machine learning based models to certain blocks or parts of the system. Use [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/products/statistics.html) and/or [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) and/or [Reinforcement Learning Toolbox™](https://www.mathworks.com/products/reinforcement-learning.html) for the machine learning based models. For example, work with the [Motor Control Blockset™](https://www.mathworks.com/products/motor-control.html) product to automate the motor control system design using MATLAB® and Simulink® with the key characteristics being Electrical Systems and Controller. The machine learning based-models/parameters can be set up using [Deep Network Designer](https://www.mathworks.com/help/deeplearning/gs/get-started-with-deep-network-designer.html) and [Reinforcement Learning Designer](https://www.mathworks.com/help/reinforcement-learning/ug/design-dqn-using-rl-designer.html). Finally, develop and evaluate a workflow that demonstrates controller design and optimization using classical control theory and machine learning-based approaches. An example introducing reinforcement learning is laid out in the suggested steps. - -Suggested steps: -1. Use Surface Mount Permanent Magnet Synchronous Motor (PMSM) technologies. -2. Perform literature research prior to starting the work. -3. Use the [PMSM Field-Oriented Control (FOC) block](https://www.mathworks.com/help/physmod/sps/ref/pmsmfieldorientedcontrol.html) to implement a PI speed and current control cascade structure. -4. Run and check the reference speed and censored speed using [FOC Autotuner](https://www.mathworks.com/help/mcb/gs/tune-pi-controllers-using-foc-autotuner.html). -5. Using Reinforcement Learning design apps, change the current control block as an agent model. Based on workflow, the PMSM and drive inverter are set as environment. Also, D-axis current, Q-axis current, and its errors can be determined as an observation. Determine the actions with D-axis and Q-axis voltages, then the reward signal to minimize the control effort from previous time step. -6. To build policy, create network architecture such as action-critic networks using Deep Network Designer apps. For example, the agent is a twin-delayed deep deterministic policy gradient (TD3) agent. A TD3 agent approximates the long-term reward given the observations and actions using two critics. Run each training for at most 1000 episodes and stop training when the agent receives an average cumulative reward greater than -190 over 100 consecutive episodes. At this point, the agent can track the reference speeds. To validate the performance of the trained agent, simulate the model and view the closed-loop performance through the Speed Tracking Scope block. -https://www.mathworks.com/help/reinforcement-learning/ug/train-td3-agent-for-pmsm-control.html - -Project variations: -1. Try the other types of reference signals such as ramp or sine wave -2. Insert the parameter errors to compare the classical control and machine learning-based approaches. For example, Rs_hat in Controller gain = 0.5 * Rs in motor model. - -Advanced project work: -1. Accurate Torque Estimation of electrical machines based on hybrid machine learning approaches -2. Estimation of Permanent Magnet temperature in rotor utilizing data-driven modeling and machine learning -3. Design optimization of motor parameters using Deep Neural Network -4. Predictive Maintenance of Electric Drive systems - -## Background Material - -Examples: -- [PMSM Control Examples](https://www.mathworks.com/help/mcb/pmsm.html) -- [Train TD3 Agent for PMSM Control](https://www.mathworks.com/help/reinforcement-learning/ug/train-td3-agent-for-pmsm-control.html) -- [PMSM Field Oriented Control Reinforcement Learning Example](https://www.mathworks.com/help/reinforcement-learning/ug/train-td3-agent-for-pmsm-control.html) - -Suggested readings: -- [1] Shuai Zhao, Frede Blaabjerg, Huai Wang. "An Overview of Artificial Intelligence Applications for Power Electronics." IEEE TRANSCTIONS ON POWER ELECTRONICS, VOL. 36, NO. 4, APRIL 2021. -- [2] Kano Matsura, Kan Akatsu, “A motor control method by using Machine learning,” 23rd international Conference on Electrical Machines and Systems (ICEMS), Nov. 2020, DOI: 10.23919/ICEMS50442.2020.9290989 -- [3] Wei-Lun Peng, Yung-Wen Lan, Shih-Gang Chen, Faa-Jeng Lin, Ray-I Chang, Jan-Ming Ho. "Reinforcement Learning Control for Six-Phase Permanent Magnet Synchronous Motor Position Servo Drive." 2020 3rd IEEE International Conference on Knowledge Innovation and Invention (ICKII), AUGUST 2020. -- [4] Soumava Bhattacharjee, Sukanta Halder, Aiswarya Balamurali, Muhammad Towhidi, Lakshmi Varaha Iyer, Narayan C. Kar. "An Advanced Policy Gradient Based Vector Control of PMSM for EV Application." 2020 10th International Electric Drives Production Conference (EDPC), DECEMBER 2020. -- [5] Wilhelm Kirchgassner, Oliver Wallcheid, “Estimating Electric Motor Temperatures with Deep Residual Machine Learning,” IEEE Transactions Power Electronics, vol. 36, pp. 7480-7488, Jul. 2021. -- [6] Mikko Tahkola, Janne Keranen, Denis Sedov, Mehrnaz Farzam Far, Juha Kortelaine, “Surrogate Mdeling of Electrical Machine Torque Using Artificial Neural Networks,” IEEE Access, pp. 2200027-220045, Dec. 2021 -- [7] Marius Stender, Oliver Wallcheid, Joachim Boker, “Accurate Torque Estimation for Induction Motors by Utilizing a Hybrid Machine Learning approach,” IEEE 19th International Power Electronics and Motion Control Conference (PEMC), Apr. 2021, DOI: 10.1109/PEMC48073.2021.9432615 - -## Impact - -Contribute to the global transition to smart manufacturing and electrification. - -## Expertise Gained - -Artificial Intelligence, Power Electronics, Control, Machine Learning, Reinforcement Learning, Automotive - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/49) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -218 diff --git a/projects/Machine Learning for Motor Control/student submissions/Machine-Learning-for-Motor-Control- b/projects/Machine Learning for Motor Control/student submissions/Machine-Learning-for-Motor-Control- deleted file mode 160000 index e2035fb3..00000000 --- a/projects/Machine Learning for Motor Control/student submissions/Machine-Learning-for-Motor-Control- +++ /dev/null @@ -1 +0,0 @@ -Subproject commit e2035fb30ffb6c9a7d00870da27d034a5726e8eb diff --git a/projects/Monitoring and Control of Bioreactor for Pharmaceutical Production/README.md b/projects/Monitoring and Control of Bioreactor for Pharmaceutical Production/README.md deleted file mode 100644 index b955e121..00000000 --- a/projects/Monitoring and Control of Bioreactor for Pharmaceutical Production/README.md +++ /dev/null @@ -1,74 +0,0 @@ -**Project 188:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Monitoring and Control of Bioreactor for Pharmaceutical Production

-

Monitor and control an industrial scale bioreactor process for pharmaceutical production.

-
- -## Motivation - -In the era of industry 4.0, which envisions a highly intelligent data-driven manufacturing environment, the need for online monitoring and control is paramount. This is especially applicable to the COVID crisis where the pharma industry is being forced to scale up manufacturing of the vaccine very quickly. -The biopharmaceutical sector is still significantly lagging other sectors in their adoption of advanced process control, particularly in their use of innovative process analytical technology (PAT) solutions. A major push from industrial regulators to rectify this has been the implementation of the Quality by Design (QbD) and PAT initiatives set out by the FDA in 2004 and 2009, respectively. While regulatory initiatives have had an impact, the uptake has been slow. A major challenge is the expertise and confidence required to adopt and implement these novel control solutions throughout industrial biopharmaceutical processes. - -## Project Description - -The objective of this project is to ensure a predefined product quality target is consistently achieved for all batches regardless of inherent process disturbances and batch-to-batch fluctuations. This is demonstrated by applying the QbD methodology utilising the PAT framework to an industrial bioprocess with MATLAB® and Simulink®. - -Suggested Steps: - -1. Familiarize yourself with the industrial-scale penicillin fermentation simulator built in MATLAB, referred to as [IndPenSim](https://www.mathworks.com/matlabcentral/fileexchange/49041-industrial-scale-penicillin-simulationv2?s_tid=srchtitle). -2. Identify the critical process parameters (CPPs) and subsequent critical quality attributes (CQAs) influencing production using multivariate analysis. Use functions from the [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/help/stats/) -3. Select an appropriate CPP and utilise the spectra recorded by the Raman spectroscopy device to develop a soft sensor enabling an on-line prediction of biomass, substrate or product concentration in real-time. -4. Develop a control strategy that manipulates one or more of the following flowrates: substrate, nitrogen or phenylacetic acid, to maintain control variables within their acceptable ranges. Consider using the [Control System Toolbox™](https://www.mathworks.com/help/control/). -5. Demonstrate benefits of this control strategy over baseline in terms of product yield increase. - -Project variations: - -1. Create a digital twin of the process by calibrating the process model with the plant data. Estimate the parameters of the model using measured data in the parameter estimator using [Simulink® Design Optimization™](https://www.mathworks.com/help/sldo/). -2. Improve the estimation of a CPP using a state-based filter (e.g. Kalman Filter, Extended Kalman Filter, Moving Horizon Estimator, etc.) this is helpful for process monitoring and control. -3. Adopt a learning-based approach (machine/deep learning) to develop soft sensors. - -Advanced project work: - -1. Develop an algorithm to detect the endpoint of reactions – the point at which the production of a batch should be terminated. -2. Implement a fault detection strategy to detect the root cause of abnormal process behavior. -3. Reduce fluctuation of variables such as pH and temperature by implementing different control strategies. - - -## Background Material - -- [Industrial-scale penicillin fermentation simulator (code and dataset)](https://www.mathworks.com/matlabcentral/fileexchange/49041-industrial-scale-penicillin-simulationv2?s_tid=srchtitle). -- [PID Controller in Simulink](https://www.mathworks.com/help/simulink/slref/pidcontroller.html). - -Suggested readings: - -[1] [Stephen Goldrick, Andrei Ştefan, David Lovett, Gary Montague, Barry Lennox, The development of an industrial-scale fed-batch fermentation simulation, Journal of Biotechnology, 2015](https://www.researchgate.net/publication/267816104_The_development_of_an_industrial-scale_fed-batch_fermentation_simulation). - -[2] [Stephen Goldrick, Carlos A. Duran-Villalobos, Karolis Jankauskas, David Lovett, Suzanne S. Farid, Barry Lennox,Modern day monitoring and control challenges outlined on an industrial-scale benchmark fermentation process, Computers & Chemical Engineering, 2019](https://www.sciencedirect.com/science/article/pii/S0098135418305106?via%3Dihub). - - -## Impact - -Improve quality and consistency of pharmaceutical products and contribute to transitioning the pharmaceutical sector to Industry 4.0. - - -## Expertise Gained - -Big Data, Industry 4.0, Control, IoT, Modeling and Simulation, Optimization, Machine Learning - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/22) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[samvrao]() - -## Project Number - -188 diff --git a/projects/Music Composition with Deep Learning/README.md b/projects/Music Composition with Deep Learning/README.md deleted file mode 100644 index 5e5dc456..00000000 --- a/projects/Music Composition with Deep Learning/README.md +++ /dev/null @@ -1,69 +0,0 @@ -**Project 243:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Music Composition with Deep Learning

-

Design and train a deep learning model to compose music.

-
- -## Motivation - -Generative deep learning models have shown promise in several areas, particularly text and image generation. A musical piece can be described by a language, is often highly structured, and pleasing music lies in a low-dimensional subspace embedded within all possible combinations of notes and sounds. Thus, it is plausible that similar models could be applied to music composition without human supervision. This is an enticing prospect for, e.g., media production companies, as bespoke music could be generated for any purpose without relying on ‘stock music’ marketplaces. - -## Project Description - -Use the [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) and the [Audio Toolbox™](https://www.mathworks.com/help/audio/musical-instrument-digital-interface-midi.html?s_tid=CRUX_lftnav) to create a generative model that can compose and play back a melody. A melody can be represented using the [MIDI](https://en.wikipedia.org/wiki/MIDI) format to describe a collection of musical notes, which come together to form a piece. One option is to train a generative model on collections of MIDI sequences to predict the next note(s) in a sequence, and this can then be used recurrently to compose entirely new melodies. - -Suggested steps are: - --Perform a literature review and decide on a network architecture to use (e.g., an LSTM). - -- Decide on the style of music and download an appropriate training data set, e.g., a [collection of classical music](https://paperswithcode.com/search?q_meta=&q_type=&q=midi+dataset) - -- Load and pre-process the training data as necessary. - -- Design and train a generative network on the training data set. - -- Play the resultant sequence, either by writing a MIDI file and playing it outside of MATLAB, or synthesizing audio with the [Audio Toolbox™](https://www.mathworks.com/help/audio/musical-instrument-digital-interface-midi.html?s_tid=CRUX_lftnav). - -- Assess the quality of the outputs from the network and make adjustments as necessary. - -Advanced extensions: - -- Instead of a melody, try to generate chord sequences. This could then be combined with the melody to form a more sophisticated piece. - -- Instead of a single melody, try to generate a polyphonic piece, or a piece with multiple instruments. This could include percussion. - -- Make use of note properties other than pitch, such as velocity (volume). This can produce more expressive pieces. - -## Background Material - -[Long Short-Term Memory Networks in MATLAB](https://www.mathworks.com/help/deeplearning/ug/long-short-term-memory-networks.html) - -[Example Project](https://www.mathworks.com/matlabcentral/fileexchange/50791-diffusion-music?s_tid=srchtitle) Includes tools useful for reading and writing MIDI files. - -[Design and play a MIDI synthesizer](https://www.mathworks.com/help/audio/ug/midi-synthesizer.html) - -[1] Allen Huang and Raymond Wu, [Deep Learning for Music](https://cs224d.stanford.edu/reports/allenh.pdf) - -[2] Carlos Hernandez-Olivan and Jose R. Beltran, [Music Composition with Deep Learning: A Review](https://arxiv.org/abs/2108.12290), arXiv:2108.12290 - -## Impact - -Generative music models can be used to create new assets on demand. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Machine Learning, Neural Networks, Audio - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/78) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -243 diff --git a/projects/Optimal Data Center Cooling/README.md b/projects/Optimal Data Center Cooling/README.md deleted file mode 100644 index 7fde7325..00000000 --- a/projects/Optimal Data Center Cooling/README.md +++ /dev/null @@ -1,65 +0,0 @@ -**Project 196:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - - -

Optimal Data Center Cooling

-

Improve performance, stability, and cost effectiveness of data centers by designing a cooling algorithm that keeps the system running as efficiently as possible.

-
- -## Motivation - -Data centers are growing at an amazing rate. They enable cloud computing and accessing incredible resources from anywhere across the globe. At their core, data centers house many power-hungry computers. Maintaining these computers at ideal temperatures allows them to perform at peak capacity and minimizes downtime. Cooling data centers requires large amounts of energy. Cooling too aggressively can increase costs and the carbon footprint of data centers, while under cooling can force systems offline or damage expensive equipment. -Intelligently cooling data centers given an understanding of the heat loads and cooling dynamics can help provide the highest uptime with the lowest carbon footprint. - -## Project Description - -Work with [Simscape™ Fluids™](https://www.mathworks.com/products/simscape-fluids.html) to create a plant and controller for a data center cooling system with dynamic loads using MATLAB® and Simulink®. The model should be detailed enough to capture important dynamics. -Dynamic loads include outside environmental conditions and server loads. The heat generated should vary with load and temperature. -The cooling system should have a level of fidelity that includes performance under various operating conditions. -A baseline thermostatic controller should be developed, and this should be improved with predictive or more intelligent control. -Demonstrate that the advanced control system can keep temperature in the desired range. -Demonstrate whether the control can allow for a greater performance envelope for the data center. Compute the difference in carbon footprint of using a baseline vs. advanced controller. - -Suggested steps: - -1. Perform a literature search to understand data center loads and cooling systems. -2. Create a dynamic data center cooling model. -3. Model different loads that the data center can experience such as times of increased load or high external temperatures. -4. Create a baseline by implementing a simple controller. -5. Create a predictive controller using information such as expected load and external temperature. -6. Demonstrate the value of your controller in keeping the data center temperature controlled and compare the carbon footprint of the cooling system with different controllers. - -Advanced project work: - -Extend the work to predict component failures with different controllers and optimal placement of critical loads within the data center. - -## Background Material - -- [Simscape Fluids documentation](https://www.mathworks.com/help/physmod/hydro/index.html) -- Ebrahimi, Khosrow, Gerard F. Jones, and Amy S. Fleischer. "A review of data center cooling technology, operating conditions and the corresponding low-grade waste heat recovery opportunities." Renewable and Sustainable Energy Reviews 31 (2014): 622-638. -- Moazamigoodarzi, Hosein, et al. "Influence of cooling architecture on data center power consumption." Energy 183 (2019): 525-535. -- A. Mousavi, V. Vyatkin, Y. Berezovskaya and X. Zhang, "Towards energy smart data centers: Simulation of server room cooling system," 2015 IEEE 20th Conference on Emerging Technologies & Factory Automation (ETFA), Luxembourg, Luxembourg, 2015, pp. 1-6, doi: 10.1109/ETFA.2015.7301573. - -## Impact - -Contribute to the performance, reliability, and efficiency of data centers worldwide. - -## Expertise Gained - -Big Data, Sustainability and Renewable Energy, Cloud Computing, Control, Deep Learning, Modeling and Simulation, Parallel Computing, Predictive Maintenance - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/27) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -196 - - diff --git a/projects/Optimization of Large Antenna Arrays for Astronomical Applications/README.md b/projects/Optimization of Large Antenna Arrays for Astronomical Applications/README.md deleted file mode 100644 index 2e3c43a7..00000000 --- a/projects/Optimization of Large Antenna Arrays for Astronomical Applications/README.md +++ /dev/null @@ -1,66 +0,0 @@ -**Project 205:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Optimization of Large Antenna Arrays for Astronomical Applications

-

Design a large antenna array and optimize its multiple design variables to achieve desired transmission/reception characteristics.

-
- -## Motivation - -Drone-based photography in Mars or landing on the moon has become a realized dream today due to tremendous advancement in radio-communication technology. Intelligently designed antenna arrays are potential contributors for transmission and reception of such long distant signals. Large arrays of antennas find applications in recent technologies like in square kilometer array or in massive MIMO communication. When array size is electrically large like in reflector arrays, it urges for large amount of computational time and infrastructure, followed by a complex theoretical characterization. In such practical design applications of large arrays, MATLAB® based toolboxes like Antenna Toolbox™ and Optimization Toolbox™ can be potential solution for fast computation with limited resource. - -## Project Description - -Use [Antenna Toolbox™](https://www.mathworks.com/products/antenna.html) to characterize and optimize the desired transmission and reception characteristics of large antenna arrays. Unlike the popular stochastic approximation of array analysis, Antenna Toolbox provides a full-wave electromagnetic simulation to achieve more accurate results. - -Suggested steps: -1. Use the [*design*](https://www.mathworks.com/help/antenna/ref/design.html) function on an existing [antenna catalog]( https://www.mathworks.com/help/antenna/antenna-catalog.html) in Antenna Toolbox to design an isolated antenna element at any arbitrary frequency. -2. Use [array catalog]( https://www.mathworks.com/help/antenna/array-catalog.html) of Antenna Toolbox to define a large array. -3. Optimize multiple design parameters of the large array like antenna physical parameters, inter-element spacing or tilting, etc. using Antenna Toolbox in association with other MATLAB optimization Toolboxes. -4. Compare large finite arrays with infinite array analysis of Antenna Toolbox and investigate the difference in the far-field radiation patterns. See this [example](https://www.mathworks.com/help/antenna/ug/modeling-mutual-coupling-in-large-arrays-using-infinite-array-analysis.html?searchHighlight=infinite%20array&s_tid=srchtitle) - -Advanced project work: - -Currently design function provides the initial design of isolated metal only antenna. -1. Use the simulation data obtained using Antenna Toolbox to build a data-driven prediction modelling of initial design parameters of the large array. -2. Derive standard design models for different class of arrays for different antenna elements. - -## Background Material - -- [Square Kilometre Array](https://www.skatelescope.org/the-ska-project/) - -- [Antenna Toolbox](https://www.mathworks.com/help/antenna/) - -- [Optimization Toolbox](https://www.mathworks.com/products/optimization.html) - -Suggested readings: - -[1] E.W. Reid, L. Ortiz-Balbuena, A. Ghadiri, and K. Moez, "[A 324-Element Vivaldi Antenna Array for Radio Astronomy Instrumentation](https://doi.org/10.1109/TIM.2011.2159414)", IEEE Transactions on Instrumentation and Measurement, 61(1):2012. - -[2] M. Ivashina, and J.D.B.A. van Ardenne. "[A way to improve the field of view of the radio telescope with a dense focal plane array](https://doi.org/10.1109/CRMICO.2002.1137238)", 12th International Conference Microwave and Telecommunication Technology, 2002. - - -## Impact - -Advance long distance communication capabilities for astronomical applications - -## Expertise Gained - -5G, Smart Antennas, Wireless Communication, Optimization - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/36) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[sghosalcem](https://github.com/sghosalcem) - -## Project Number - -205 diff --git a/projects/Optimizing Antenna Performance in an Indoor Propagation Environment/README.md b/projects/Optimizing Antenna Performance in an Indoor Propagation Environment/README.md deleted file mode 100644 index dab49f34..00000000 --- a/projects/Optimizing Antenna Performance in an Indoor Propagation Environment/README.md +++ /dev/null @@ -1,70 +0,0 @@ -**Project 206:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Optimizing Antenna Performance in an Indoor Propagation Environment

-

Design an antenna to optimize transmission and reception in indoor environment

-
- -## Motivation - -Design and optimization of antennas considering indoor propagation environment can find realistic applications like the placement and orientation of router antennas inside a large room/hall or to construct a multi-point communication network in a large building. It is easy to characterize antennas in ideal free-space environment. However, in an indoor environment, there will be multiple reflection from the floors, walls, and surrounding objects to perturb the ideal radiation behavior of the antenna. This will influence both the signal strength as well as the direction of the beam pattern. As next generation communication technology is moving towards higher frequency region, the impacts of the neighboring objects are more prominent. Currently, these issues are addressed by measurement-based observations only. In this regard, with a prior information of the building environment, one can utilize the Antenna Toolbox™ and other toolboxes like Optimization toolbox™ to optimize an indoor transmission/reception performance of the antenna or antenna arrays. - -## Project Description - -Use [Antenna Toolbox ™](https://www.mathworks.com/products/antenna.html) to characterize and optimize the desired transmission and reception characteristics of antenna or antenna arrays inside an indoor propagation environment. Unlike the commonly used measurement-based approach, Antenna Toolbox provides a suitable alternative to optimize the antenna design and placement with prior information of indoor topology. Utilize the material loss properties of Antenna Toolbox to have a more accurate insight of realistic transmission/reception characteristics. A high-level sequence of steps is given below: - -1. Use the [*design*](https://www.mathworks.com/help/antenna/ref/design.html) function on an existing [antenna catalog]( https://www.mathworks.com/help/antenna/antenna-catalog.html) in Antenna Toolbox ™ to design an isolated antenna element at any arbitrary frequency with the assumption of ideal free-space propagation. - -2. Create indoor propagation scenario as per requirement while specifying the desired location of the receiving user(/s). Use the [installedAntenna](https://www.mathworks.com/help/antenna/ref/installedantenna.html) object of Antenna Toolbox for this purpose. - -3. Optimize the design and orientation parameters of the antenna to achieve expected radiation characteristics like to maximize the signal strength at a particular receiving location or to maximize the average signal strength in multiple receiving ends. - -4. Visualize a full-wave signal coverage map using the capabilities of the Antenna Toolbox ™. - -Advanced project work: - -1. Using the simulation data obtained using Antenna Toolbox, develop a data-based prediction modelling of arbitrary shaped indoor propagation scenario. - -2. Investigate such a model with different radiating elements. - -3. Derive suitable design suggestions for different class of antennas and arrays for future references. - -## Background Material - -- [Antenna Toolbox](https://www.mathworks.com/help/antenna/) -- [Optimization Toolbox](https://www.mathworks.com/products/optimization.html) -- [RF Propagation and Visualization](https://www.mathworks.com/help/antenna/gs/rf-propagation-and-visualization.html) - -Suggested readings: - -[1] A. Cidronali, et. al., "Analysis and Performance of a Smart Antenna for 2.45-GHz Single-Anchor Indoor Positioning", IEEE Transactions on Microwave Theory and Techniques, 58 (1):2010. - -[2] X. Wu, et.al., "60-GHz Millimeter-Wave Channel Measurements and Modeling for Indoor Office Environments", IEEE Transactions on Antennas and Propagation, 65(4): 2017. - -[3] Z. Genc, et.al.,"Robust 60 GHz Indoor Connectivity: Is It Possible with Reflections?”, 2010 IEEE 71st Vehicular Technology Conference, 2010. - -## Impact - -Maximize indoor radio signal coverage and reduce energy consumption of signal booster devices. - -## Expertise Gained - -5G, Optimization, Smart Antennas, Wireless Communication - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/37) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[sghosalcem](https://github.com/sghosalcem) - -## Project Number - -206 diff --git a/projects/Path Planning for Autonomous Race Cars/README.md b/projects/Path Planning for Autonomous Race Cars/README.md deleted file mode 100644 index b02e9d3d..00000000 --- a/projects/Path Planning for Autonomous Race Cars/README.md +++ /dev/null @@ -1,65 +0,0 @@ -**Project 208:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Path Planning for Autonomous Race Cars

-

Develop an algorithm to compute an optimal path for racing tracks.

-
- -## Motivation - -With advancements in automotive technology, various industries and academia are investing in path planning algorithms for driverless vehicles. As an example, Formula Student competitions have introduced the driverless category, where the goal of the teams is to design and build an autonomous vehicle that can compete in different disciplines. In such competitions or global racing championships such as Formula 1, the optimal racing line is one of the winning factors. In basic terms, an optimal racing line is the shortest path through a race circuit to achieve the best lap-time. -Developing an optimal racing line algorithm will help the teams to compute the path for new tracks, improving the vehicle’s overall performance. This will also serve as a reference to train the drivers participating in conventional racing championships. - - -## Project Description - -Finding an optimal racing line can be considered an optimization problem. Work with MATLAB® and the [Optimization Toolbox™](https://www.mathworks.com/products/optimization.html) to develop the algorithm. - -Suggested steps: -1. Perform a Google search on optimization-based path planning. -2. Define a basic [bicycle vehicle model](https://www.mathworks.com/help/robotics/ref/bicyclekinematics.html) of the car. -3. Create or import a racetrack using [RoadRunner](https://www.mathworks.com/products/roadrunner.html) or [Driving Scenario Designer] (https://www.mathworks.com/help/driving/ref/drivingscenariodesigner-app.html) -4. Given a known racetrack, generate a minimum-time velocity profile [1]. -5. Formulate an optimization problem. The two most common approaches for defining the optimization problem are minimizing the time [3] and minimizing the curvature [4]. You can follow either of these two approaches and use the [Optimization Toolbox™](https://www.mathworks.com/products/optimization.html) to solve the problem. -6. Include a 2D visualization plot to demonstrate the vehicle trajectory in real-time. -7. Test the robustness of the algorithm on different racetracks. - -Advanced project work: -1. Add static obstacles on racetracks and compute the optimal racing line. Use the generated optimal racing line and velocity profile as reference for a high-fidelity vehicle dynamics models provided by the [Vehicle Dynamics Blockset™](https://www.mathworks.com/products/vehicle-dynamics.html). Extend the model to track the reference path using autonomous driving techniques. -2. Incorporate electrical and 3D mechanical behavior into the model using a Simscape model for the vehicle. Extend optimization examples in the [Simscape Vehicle Templates](https://www.mathworks.com/matlabcentral/fileexchange/79484-simscape-vehicle-templates) to tune the suspension design and include battery usage into the optimization problem. - - -## Background Material - -- [Real-time Trajectory Optimization for Autonomous Vehicle Racing](https://github.com/janismac/RacingTrajectoryOptimization) -- [Solving Optimization Problems with MATLAB](https://www.youtube.com/watch?v=4wgI3-RQqTY). -- [Optimization Toolbox](https://www.mathworks.com/products/global-optimization.html) -- [Global Optimization Toolbox](https://www.mathworks.com/products/global-optimization.html) -- [Design, simulate, and deploy path planning algorithms](https://www.mathworks.com/discovery/path-planning.html) - - -## Impact - -Push racing car competitions into fully autonomous mode. - -## Expertise Gained - -Autonomous Vehicles, Automotive, Optimization, Modeling and Simulation - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/39) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[mw-veeralakshendra](https://github.com/veeralakshendra) - -## Project Number - -208 diff --git a/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_AUTON_RACECARS b/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_AUTON_RACECARS deleted file mode 160000 index f5b26b82..00000000 --- a/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_AUTON_RACECARS +++ /dev/null @@ -1 +0,0 @@ -Subproject commit f5b26b8289c44989aab2fba4683b0a9c0830a4d5 diff --git a/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_Raceline_Optimization b/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_Raceline_Optimization deleted file mode 160000 index 94e9a6e9..00000000 --- a/projects/Path Planning for Autonomous Race Cars/students submissions/MW208_Raceline_Optimization +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 94e9a6e9fb23d14f9718b5472682fb379cc6c130 diff --git a/projects/Path Planning for Autonomous Race Cars/students submissions/MW_EiI_208_Trajectory_Planning_and_Tracking b/projects/Path Planning for Autonomous Race Cars/students submissions/MW_EiI_208_Trajectory_Planning_and_Tracking deleted file mode 160000 index fc023269..00000000 --- a/projects/Path Planning for Autonomous Race Cars/students submissions/MW_EiI_208_Trajectory_Planning_and_Tracking +++ /dev/null @@ -1 +0,0 @@ -Subproject commit fc023269cc667f4eee9c02b63cbfdc8a60f0d318 diff --git a/projects/Portable Charging System for Electric Vehicles/README.md b/projects/Portable Charging System for Electric Vehicles/README.md deleted file mode 100644 index aa410b61..00000000 --- a/projects/Portable Charging System for Electric Vehicles/README.md +++ /dev/null @@ -1,70 +0,0 @@ -**Project 216:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Portable Charging System for Electric Vehicles

-

Design a portable charger for Electric Vehicles

-
- -## Motivation - -Electric Vehicles are rising in popularity as the global trend towards electrification continues. However, this new and exciting technology also poses different challenges compared to conventional internal combustion engines, which will need new engineering solutions to resolve. -One such challenge is the possibility of an electric vehicle running out of charge before it can reach a recharging station. In such a scenario, it might be easier to simply recharge the EV on-the-spot using a charged battery pack rather than putting in the much more substantial effort of towing it to another location. -Solving this problem requires a detailed understanding of how battery packs and power electronics operate, and also of the control algorithms necessary to manage the charging process in accordance with established standards. - - -## Project Description - -Work with [Simscape™](https://in.mathworks.com/help/physmod/simscape/index.html) and [Simscape™ Electrical™](https://in.mathworks.com/help/physmod/sps/index.html) blocks to create a simulation model of a reference EV’s battery pack with electrical and thermal characteristics modelled. Next, devise a solution for transferring charge from an external battery through an intermediate power electronic circuit to the reference EV. Select an appropriate value of the charging rate and total power transferred, considering the trade-off against time taken and the cost of hardware needed for faster charging. - -Suggested steps: -1. Perform literature research prior to starting the work to understand the basics of EV battery packs, and charging/discharging circuits. Consider using the resources [here](https://mathworks.com/solutions/power-electronics-control/battery-models.html). -2. Create a model of the Battery Pack within an EV, ensuring the sizing and battery chemistry is appropriate for contemporary applications. Consider using a [generic battery model]( https://in.mathworks.com/help/physmod/sps/powersys/ref/battery.html) as a starting point. -3. Create a model of a second, external battery pack, which will be used to recharge the EV. The size and configuration of this should be parametrized, so that the cost and charging rate can be optimized. -4. Model the intermediate power electronic charging circuit, ensuring the electrical and thermal characteristics of the conversion are within an acceptable range. Consider using the examples provided in the Simscape Electrical library [here]( https://in.mathworks.com/help/physmod/sps/power-electronics.html). -5. Verify the suitability of the charging control algorithms and power electronics design through simulations. - -Advanced project work: -1. Customize the design to match various popular EV manufacturers. -2. Use the model to compare the speed and cost of different charging technologies, battery chemistries and cell configurations. -3. Build and test the actual hardware - - -## Background Material - -- [Battery Electric Vehicle Model in Simscape](https://in.mathworks.com/matlabcentral/fileexchange/82250-battery-electric-vehicle-model-in-simscape) -- [Lithium Battery Charger Block](https://in.mathworks.com/matlabcentral/fileexchange/72570-lithium-battery-charger-block) -- Video Series: [Developing DC-DC Converter Control with Simulink](https://in.mathworks.com/videos/series/developing-dc-dc-converter-control-with-simulink.html) -- Example: [Buck Converter With Thermal Dynamics](https://in.mathworks.com/help/physmod/sps/ug/buck-converter-with-thermal-dynamics.html) - -Suggested readings: -- [Modern Electric, Hybrid Electric, and Fuel Cell Vehicles, 3e](https://in.mathworks.com/academia/books/modern-electric-hybrid-electric-and-fuel-cell-vehicles-ehsani.html) -- A. Arancibia and K. Strunz, "Modeling of an electric vehicle charging station for fast DC charging," 2012 IEEE International Electric Vehicle Conference, Greenville, SC, USA, 2012, pp. 1-6, doi: 10.1109/IEVC.2012.6183232. -- Falvo, Maria Carmen, et al. "EV charging stations and modes: International standards." 2014 International Symposium on Power Electronics, Electrical Drives, Automation and Motion. IEEE, 2014. -- Das, H. S., et al. "Electric vehicles standards, charging infrastructure, and impact on grid integration: A technological review." Renewable and Sustainable Energy Reviews 120 (2020): 109618. -- Atmaja, Tinton Dwi. "Energy storage system using battery and ultracapacitor on mobile charging station for electric vehicle." Energy Procedia 68 (2015): 429-437. -- Huang, Shisheng, et al. "Design of a mobile charging service for electric vehicles in an urban environment." IEEE Transactions on Intelligent Transportation Systems 16.2 (2014): 787-798. -- Khan, Abdul Basit, et al. "Multistage constant-current charging method for Li-Ion batteries." 2016 IEEE Transportation Electrification Conference and Expo, Asia-Pacific (ITEC Asia-Pacific). IEEE, 2016. - - -## Impact - -Help make Electric Vehicles more reliable for general use. - - -## Expertise Gained - -Sustainability and Renewable Energy, Control, Electrification, Modeling and Simulation - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/47) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -216 diff --git a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Buck-Converter-EV-charger b/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Buck-Converter-EV-charger deleted file mode 160000 index 4ed96491..00000000 --- a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Buck-Converter-EV-charger +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 4ed9649166da28cf4617944b175908706a6a9b15 diff --git a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System b/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System deleted file mode 160000 index 994668fb..00000000 --- a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 994668fb316805f870f85da4512c8bc56e093c6f diff --git a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System-for-EVs b/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System-for-EVs deleted file mode 160000 index 98e15601..00000000 --- a/projects/Portable Charging System for Electric Vehicles/student submissions/Portable-Charging-System-for-EVs +++ /dev/null @@ -1 +0,0 @@ -Subproject commit 98e156017431efc9a4b35a295dd267a3d5619806 diff --git a/projects/Predictive Electric Vehicle Cooling/README.md b/projects/Predictive Electric Vehicle Cooling/README.md deleted file mode 100644 index cab28505..00000000 --- a/projects/Predictive Electric Vehicle Cooling/README.md +++ /dev/null @@ -1,63 +0,0 @@ -**Project 194:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Predictive Electric Vehicle Cooling

-

Improve range, performance, and battery life by designing a cooling algorithm that keep EV battery packs cool when they need it most.

-
- -## Motivation - -Electric vehicle (EV) adoption is growing at an amazing rate. The battery packs that these vehicles carry are their lifeblood, and excessive heat can limit their performance. -Charging and discharging performance can be limited by thermal constraints. Heat exposure can limit the range and life of battery packs. The thermal time constant of larger battery packs is long and, thermostatic cooling models can be too slow, as they may first cool the pack when it already has a large amount of thermal energy. -Predictively cooling the pack based on expected thermal demands can help keep it near its optimal temperature range. This can allow faster charging and discharging, longer range, and extended battery life. - -## Project Description - -Work with [Simscape™ Fluids™](https://www.mathworks.com/products/simscape-fluids.html) to create a plant and predictive controller for EV cooling system with dynamic loads using MATLAB® and Simulink®. -The model should be detailed enough to capture important dynamics. Dynamic loads include outside environmental conditions, fast charging, and rapid acceleration/deceleration. Demonstrate that the predictive control system can keep battery temperature in the desired range. Demonstrate whether the control can allow for a greater performance envelope for motor loads and fast charging. Compute the change in energy requirement from operating the cooling system predictively vs. reactively. - -Suggested steps: -1. Perform a literature search to understand EV cooling systems, battery management, and drive cycles. -2. Study a [dynamic EV cooling model](https://www.mathworks.com/help/physmod/hydro/ug/sscfluids_ev_battery_cooling.html) -3. Model different loads that the battery can experience such as fast charging and rapid acceleration. -4. Create a baseline by implementing a simple controller. -5. Create a predictive controller using information such as incoming charge, throttle position, and location. -6. Demonstrate the value of your controller in keeping the battery temperature controlled and compare the energy requirements of the cooling system with different controllers. - -Advanced project work: - -Extend the work to predict battery range and life expectancy improvement with the predictive controller. - - -## Background Material - -- [Simscape Fluids Documentation](https://www.mathworks.com/help/physmod/hydro/index.html). -- T. Huria, M. Ceraolo, J. Gazzarri,R. Jackey. "High Fidelity Electrical Model with Thermal Dependence for Characterization and Simulation of High Power Lithium Battery Cells," IEEE International Electric Vehicle Conference, March 2012 - - - -## Impact - -Contribute to the electrification of transport worldwide. Increase the range, performance, and battery life of EVs. - -## Expertise Gained - -Sustainability and Renewable Energy, Autonomous Vehicles, Automotive, Control, Electrification, Modeling and Simulation, Optimization - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/25) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[MUdengaard](https://github.com/MUdengaard) - -## Project Number - -194 diff --git a/projects/Quadruped Robot with a Manipulator/README.md b/projects/Quadruped Robot with a Manipulator/README.md deleted file mode 100644 index b782eaf8..00000000 --- a/projects/Quadruped Robot with a Manipulator/README.md +++ /dev/null @@ -1,65 +0,0 @@ -**Project 29:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - -

Quadruped Robot with a Manipulator

-

Legged robots with manipulators will be the ideal platforms to traverse rough terrains and interact with the environment. Are you ready to tackle the challenge of operating robots outdoor?

-
- -## Motivation - -Boston Dynamics and MIT Biomimetics Robotics Lab have led the revolution in legged robots. -Quadruped robots can walk over terrains difficult for their wheeled counterparts performing a diverse set of tasks. -Moreover, legged robots have the capability to change the body position and posture without changing the foot points. -Thus, they have captured interest from a wide range of industries--military scout, street patrol, factory inspection, and even home butler. -Robot manipulators with the locomotion ability provided by legged robots enable robot arms to work in a much larger variety of fields including construction, agriculture, and forestry. -Get prepared for the industry behind this blooming application! - - -## Project Description - -Create a Simulink® model of a quadruped robot with a manipulator. Design control algorithms for it to do the following tasks: -- Scan its 360-degree surrounding -- Going upstairs -- Holding a cup of water while going upstairs -- One more fun task you define - -Suggested high-level steps: - -1. Use CAD software to model the robot and import it to Simscape™ -2. Add sensors and motors to the Simscape model, so you can command the movements of the robot and measure the motion of every link of the robot -3. Define a controller in Simulink to perform individual tasks -4. Use Stateflow® to connect a sequence of tasks together. - - -## Background Material - -- [Quadruped Robot Locomotion Using DDPG Agent](https://www.mathworks.com/help/reinforcement-learning/ug/quadruped-robot-locomotion-using-ddpg-gent.html) -- [Running Robot Model in Simscape](https://www.mathworks.com/matlabcentral/fileexchange/64237-running-robot-model-in-simscape) - - -## Impact - -Contribute to state-of-the-art technologies for exploration, and search and rescue transformation. - - -## Expertise Gained - -Robotics, Control, Image Processing, Manipulators, Mobile Robots, Modeling and Simulation - - -## Project Difficulty - -Master’s level - - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/6) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[dryouwu](https://github.com/dryouwu) - -## Project Number - -29 diff --git a/projects/Reinforcement Learning Based Fault Tolerant Control of a Quadrotor/README.md b/projects/Reinforcement Learning Based Fault Tolerant Control of a Quadrotor/README.md deleted file mode 100644 index bdaf4f1e..00000000 --- a/projects/Reinforcement Learning Based Fault Tolerant Control of a Quadrotor/README.md +++ /dev/null @@ -1,65 +0,0 @@ -**Project 235:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Reinforcement Learning Based Fault Tolerant Control of a Quadrotor

-

Develop a fault-tolerant controller for a quadcopter using model-based reinforcement learning.

-
- -## Motivation - -Unmanned aerial vehicles (UAVs), such as quadcopters, are nowadays popular vehicles adopted for various applications. The actual operation of quadcopters can be affected by disturbances such as wind, rain, dust, etc., which may cause component faults. Fault is defined as a change in a system's property or parameters that cause the system to behave differently from its design. A fault-tolerant controller (FTC) is a control strategy that aims to improve the performance of a system that is operating in degraded performance due to fault [1]. - -Deployment of multi-rotor drones for applications like Urban Air Mobility (UAM), product delivery, etc., requires a proper behavior of the vechile at all times to ensure safety of the environemnt and people nearby. That is why a FTC is a necessary block for such systems and both academic researchers and industry professionals are looking to improve methods to observe faults and provide a control strategy able to overcome them and ensuring safe behavior. - -FTCs are characterized as model-based or data-driven based on the method used to develop the controllers. Model-based techniques necessitate knowledge of the system's model and parameters in order to design a fault-tolerant controller. Data-driven approaches, on the other hand, learn the FTC directly from system data. The fundamental problem of model-based FTC approaches is that their effectiveness is dependent on the correctness of the system model, which is difficult to establish when system parameters vary due to faults. Furthermore, complex systems necessitate complicated controllers, which has an impact on the controllers' robustness. Data-driven techniques, on the other hand, utilize data to design FTC without knowing the full dynamics of the system. As a result, data-driven methods, particularly reinforcement learning (RL)-based techniques, have recently gained the attention of a number of researchers. - -## Project Description - -Train an RL agent to develop a fault-tolerant controller for a quadcopter using model-based reinforcement learning. The framework uses the system dynamics and a Kalman filter-based estimator to estimate the fault-related parameters online, which will be used to identify the occurrence of a fault in the system. Once you identify the event of a fault, you will use the fault-related parameters to train an RL agent that tunes the position and attitude controller gains of the quadcopter to compensate for the happening fault. - -Suggested steps: -1. Review [Tune PI Controller using Reinforcement Learning](https://www.mathworks.com/help/reinforcement-learning/ug/tune-pi-controller-using-td3.html) example to learn how to use the [Reinforcement Learning Toolbox](https://www.mathworks.com/help/reinforcement-learning/index.html?s_tid=CRUX_lftnav) to tune a PI controller for a system. -2. Review [Quadcopter Drone Model in Simscape](https://www.mathworks.com/matlabcentral/fileexchange/63580-quadcopter-drone-model-in-simscape?s_tid=srchtitle) example, which contains a detailed model of the quadcopter including the airframe, battery, and propulsion systems, and learn how a PID control can be applied for a quadcopter's position and attitude control. -3. Design a reward function that will be used for training the RL agent (consider a reward function that takes into account the error between the reference and actual trajectory). To represent fault behavior of the system, you may use the equivalent resistance of the motors as in [2]. -4. Use the simulation environment to simulate faulty behaviors and train an RL agent to tune the quadcopter position and attitude PID controller gains. -5. Apply the trained model for tuning the PID controllers in the presence of fault/s. - - -Advanced Project work: -- Implement a state estimator for monitoring the fault related parameters that will be used for training the RL agent (you may refer to [Fault Detection Using an Extended Kalman Filter](https://www.mathworks.com/help/predmaint/ug/Fault-Detection-Using-an-Extended-Kalman-Filter.html), [2], and [3]). -- Consider complete sub-component failure instead of fault (degradation). - -## Background Material - - Examples: -- [PID Autotuning for UAV Quadcopter](https://www.mathworks.com/help/slcontrol/ug/pid-controller-tuning-for-a-uav-quadcopter.html) -- [UAV Inflight Failure Recovery](https://www.mathworks.com/help/slcontrol/ug/uav-quadcopter-controller-tuning-and-inflight-failure-recovery.html) -- [Simulink Drone Reference Application](https://www.mathworks.com/matlabcentral/fileexchange/67625-simulink-drone-reference-application) -- [UAV Package Delivery](https://www.mathworks.com/help/uav/ug/uav-package-delivery.html) - -Suggested readings: -- [1] Blanke, M., Kinnaert, M., Lunze, J., Staroswiecki, M., and Schröder, J., Diagnosis and fault-tolerant control, Vol. 2, Springer, 2006. -- [2] Bhan, L., Quinones-Grueiro, M., and Biswas, G., “Fault Tolerant Control combining Reinforcement Learning and Model-based Control,” 2021 5th International Conference on Control and Fault-Tolerant Systems (SysTol), pp. 31–36, 2021. -- [3] Matthew Daigle, Bhaskar Saha, and Kai Goebel. A comparison of filter-based approaches for model based prognostics. In 2012 IEEE Aerospace Conference, pages 1–10, 2012. - -## Impact - -Improve safety of multi-rotor drones - -## Expertise Gained - -Drones, Artificial Intelligence, Robotics, Control, Reinforcement Learning, UAV - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/71) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -235 diff --git a/projects/Robust Visual SLAM Using MATLAB Mobile Sensor Streaming/README.md b/projects/Robust Visual SLAM Using MATLAB Mobile Sensor Streaming/README.md deleted file mode 100644 index 18b13297..00000000 --- a/projects/Robust Visual SLAM Using MATLAB Mobile Sensor Streaming/README.md +++ /dev/null @@ -1,63 +0,0 @@ -**Project 213:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Robust Visual SLAM Using Mobile Sensor Streaming

-

Perform robust visual SLAM using MATLAB Mobile sensor streaming.

-
- -## Motivation - -Mobiles phones are within everyone’s reach and have sensors such as IMU, magnetic compass, GPS, camera, which can be used for visual SLAM. The challenge has been in streaming multi-sensor data such as GPS and camera data in sync to enable robust visual SLAM. This could be potentially used for developing and testing SLAM and navigation algorithms (map building and path planning) by many researchers without the need for additional complex hardware and setup. Furthermore, the resulting point cloud of an object or environment could be leveraged for augmented reality applications. - -## Project Description - -Work with [Computer Vision Toolbox™](https://www.mathworks.com/products/communications.html) (CVT), [Navigation Toolbox™](https://www.mathworks.com/products/navigation.html) (NAV), [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html) (SFTT), [webcam hardware support]( https://www.mathworks.com/videos/webcam-support-89504.html )), and [MATLAB® Mobile™](https://www.mathworks.com/products/matlab-mobile.html) to enable robust visual SLAM from mobile sensors. Using the examples provided in Steps 1 and 2 below, acquire streaming sensor data and calibrate the data. Other examples in steps 3 and 4 help build a map and export to a point cloud. This project will extend the suggested basic example and focus on integrating all of these into a complete workflow that is general enough for robust visual SLAM and works independent of the source of the streaming data. - -Suggested steps: -1. Obtain streaming sensor data from [MATLAB® Mobile™](https://www.mathworks.com/products/matlab-mobile.html) and [webcam]( https://www.mathworks.com/videos/webcam-support-89504.html ) -2. Calibrate the sensors [example](https://www.mathworks.com/help/vision/ug/camera-calibration.html). -3. Use CVT, NAV, and SFTT to build robust visual SLAM -4. Export resulting map to a point cloud -5. Finally, a workflow that demonstrates robust visual SLAM can be developed. - -Project variations: -- Use MATLAB Mobile and webcam/Mobile together to obtain the sensor data -- Upgrade MATLAB Mobile to stream camera data with integrated IMU and GPS readings -- Crowdsourcing raw sensor data (images, GPS, IMU readings) to build large scale maps along with dynamic updates. - - -## Background Material - -Examples: -- [Monocular Visual SLAM](https://www.mathworks.com/help/vision/ug/monocular-visual-simultaneous-localization-and-mapping.html) -- [Structure from Motion](https://www.mathworks.com/help/vision/ug/structure-from-motion-from-multiple-views.html) -- [Optimize Pose Graph](https://www.mathworks.com/help/nav/ref/optimizeposegraph.html) - -Suggested readings: - -[1] P. Tanskanen, K. Kolev, L. Meier, F. Camposeco, O. Saurer and M. Pollefeys, "Live Metric 3D Reconstruction on Mobile Phones," 2013 IEEE International Conference on Computer Vision, -2013, pp. 65-72, doi: 10.1109/ICCV.2013.15. - - -## Impact - -Enable visual SLAM from streaming sensors and extend the state-of-art in real-time visual SLAM algorithms. - -## Expertise Gained - -Autonomous Vehicles, Computer Vision, Drones, Robotics, Automotive, AUV, Mobile Robots, Manipulators, Humanoid, UAV, UGV - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/44) to ask/answer questions, comment, or share your ideas for solutions for this project. - - -## Project Number - -213 diff --git a/projects/Rotor-Flying Manipulator Simulation/README.md b/projects/Rotor-Flying Manipulator Simulation/README.md deleted file mode 100644 index e1bc57fa..00000000 --- a/projects/Rotor-Flying Manipulator Simulation/README.md +++ /dev/null @@ -1,74 +0,0 @@ -**Project 47:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Rotor-Flying Manipulator Simulation

-

Rotor-flying manipulation will change the future of aerial transportation and manipulation in construction and hazardous environments. Take robotics manipulation to the next level with an autonomous UAV

-
- -## Motivation - -Rotor-flying manipulators are Unmanned Aerial Vehicles (UAVs) equipped -with a lightweight manipulator and have the potential to transform major -industries thanks to their unconstrained 3D motion making them ideal for deployent -in cluttered environments. Rotor-flying manipulation is a natural evolution -of mobile manipulation and a popular research area in robotics that -attracts the interest of many companies and public institutions. Its -applications range from aerial transportation in construction, -manipulations in hazard places, inspections and installations on sites -with a difficult access, to search and rescue. - -Autonomous aerial manipulation is challenging problem because of the -coupled dynamics between the two systems. - -## Project Description - -Develop an autonomous aerial manipulation simulation including a UAV equipped with a multi-DoF manipulator to pick an object and place it into a goal location. Pose estimation and perception of the environment will be developed using a visual system. Global motion planning with obstacle avoidance will allow the system to reach the target location to approach and pick an object and eventually place it into a goal location. -Model the UAV and the manipulator and couple them together. The coupled system will be the plant for the controller you will need to design and implement in order to stabilize and fly the system. -For this project we can neglect aerodynamic effects, assuming low-speed flight, and external disturbances. - -Suggested steps: - -1. Create a Simulink or MATLAB model of the coupled system of UAV + manipulator. Simscape multibody is a physical modeling tool that you can use for modeling the system and import the model in the environment of your choice (Simulink or MATLAB) -2. Design and implement a controller to control the orientation and position of the coupled system -3. Build your environment as an open space to start with (without obstacles), and then if you have enough time, add obstacles and use an obstacle avoidance algorithm to fly around them. -4. Use the position of the objects and the target locations as waypoints to the flying manipulator (using a waypoint follower algorithm from the UAV Toolbox) to get close to the objects to pick. -5. Once the UAV is above the object, plan the manipulator motion using a path planner algorithm. - -Advanced project work: - -1. Implement a visual-based pose estimation algorithm - - Use Lidar scans to estimate the system’s pose (ego-motion) - - Use Computer Vision and Deep Learning (for example the YOLOv2 neural network) to detect and locate the object to pick and place. - - -## Background Material - -- [UAV Toolbox](https://www.mathworks.com/products/uav.html) -- [Robotics System Toolbox](https://www.mathworks.com/products/robotics.html) -- [Navigation Toolbox](https://www.mathworks.com/help/nav/getting-started-with-navigation-toolbox.html) -- [Simscape Multibody](https://www.mathworks.com/products/simmechanics.html) -- [Quadcopter programming in Simulink](https://www.mathworks.com/videos/programming-drones-with-simulink-1513024653640.html) -- [Quadcopter modelling and Control with Simscape and Simulink](https://www.mathworks.com/matlabcentral/fileexchange/44902-quadrotor-modelling-and-control-with-simmechanics?s_tid=srchtitle) -- [Pick-and-place example](https://www.mathworks.com/help/robotics/examples/pick-and-place-workflow-using-stateflow.html) - -## Impact - -Transform the field of robot manipulation. - -## Expertise Gained - -Drones, Robotics, Manipulators, Modeling and Simulation, UAV - -## Project Difficulty - -Master’s level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/12) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -47 - diff --git a/projects/Satellite Collision Avoidance/README.md b/projects/Satellite Collision Avoidance/README.md deleted file mode 100644 index c94320f1..00000000 --- a/projects/Satellite Collision Avoidance/README.md +++ /dev/null @@ -1,97 +0,0 @@ -**Project 225:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Satellite Collision Avoidance

-

Model satellites in Low Earth Orbit (LEO) to identify conjunctions and prevent collisions with space debris, while maintaining orbital requirements.

-
- -## Motivation - -The satellite industry is experiencing unprecedented growth, with new companies launching record numbers of satellites transforming global communication, providing near real-time images of our planet, and even enabling commercial space tourism. - -As more and more satellites are added to low Earth orbit, the probability of collisions between satellites and orbital debris continues to rise. This is compounded by the fact that any collisions that do occur (either between operational satellites and debris or two pieces of debris) create more space debris on impact. Satellites and orbital debris travel at orbital speeds greater than 17,500 mph (7825 m/s) in low earth orbit. At these speeds, even relatively small orbital debris can cause substantial to catastrophic damage to a satellite or a spacecraft if a collision occurs. - -In [March 2021, a Chinese military satellite launched in September 2019 collided](https://www.space.com/space-junk-collision-chinese-satellite-yunhai-1-02) with a 10-50 cm piece of debris from the Russian Zenit-2 rocket that launched Russia’s Tselina-2 satellite in September 1996. This collision resulted in 37 additional, trackable debris objects, and likely many more smaller pieces that remain untracked. - -According to the [NASA Orbital Debris Program Office](https://www.orbitaldebris.jsc.nasa.gov/faq/#), there are approximately 23,000 pieces of debris larger than 10 cm orbiting the Earth. There are half a million pieces of debris between 1 and 10 cm, and approximately 100 million pieces of debris 1 mm or larger. The United States Space Command (USSPACECOM) Space Surveillance Network (SSN) sensors catalog and track approximately 27,000 pieces of orbital debris in near-Earth orbit (larger than 10 cm). This information can be obtained from [Space-track.org](https://www.space-track.org/). - -Historically, satellite operators predict and manually maneuver spacecraft to avoid conjunctions with other spacecraft and cataloged space debris. The SSN, for example, analyses known orbital debris trajectories over time to identify upcoming potential close encounters with the International Space Station. If large debris is projected to pass within a few kilometers of the space station and probability of collision is greater than 1 in 10,000, a maneuver is scheduled to avoid conjunction. - -The SpaceX Starlink constellation is the [first commercial application](https://spectrum.ieee.org/spacex-preps-selfdriving-satellites-for-launch) of an autonomous conjunction avoidance system. - -## Project Description - -Work with the [Aerospace Toolbox](https://www.mathworks.com/products/aerospace-toolbox.html) and [Aerospace Blockset™](https://www.mathworks.com/products/aerospace-blockset.html) products to identify potential satellite-debris conjunctions, and develop and analyze an automated algorithm for satellites to avoid space debris over time while maintaining orbital requirements. Test and validate your algorithm. - -Suggested steps: -1. Review [Comparison of Orbit Propagators](https://www.mathworks.com/help/aerotbx/ug/comparison-of-orbit-propagators.html) and Constellation Modeling with the [Orbit Propagator Block](https://www.mathworks.com/help/aeroblks/constellation-modeling-with-the-orbit-propagator-block.html) examples to learn how to propagate orbital trajectories using Two-Body-Keplerian, SGP4 and SDP4 propagators in Aerospace Toolbox and Aerospace Blockset. -2. Run additional examples listed in the Background Material section below. -3. Download space debris states from [Space-track.org](https://www.space-track.org/). -4. Propagate the debris trajectories using SGP4 propagation algorithm. -5. Define a satellite mission with specific orbit definition and requirements. e.g.: earth imaging or broadband communications. -6. Propagate the satellite’s states over time, along with the space debris. -7. Identify potential conjunctions during the analysis window [5,6]. - - Determine close contacts by distance criteria. - -Extended project work: -1. Plan avoidance maneuver for upcoming conjunction with debris. - - Maintain orbit requirements when possible. -2. Simulate avoidance maneuver. -3. Design and execute a test and validation method for your algorithm. Consider using [Verification and Validation products](https://www.mathworks.com/solutions/verification-validation.html). - -Advanced project work: -1. Develop probability of collision analysis to enhance conjunction detection in step 7 by determining close contacts by both distance criteria and probability of collision [5]. -2. Expand the analysis to cover a constellation of satellites. Size of the constellation can be defined from your chosen satellite mission. -3. Improve the avoidance maneuver algorithm by optimizing fuel consumption during the maneuver including optimizations for the orbit experiencing changes due to predictable Earth perturbations and the impact of flying in a variable drag environment ([1]). Fuel consumption is a main constraint during the satellite mission determining the satellite operational lifetime. Determine a fuel consumption estimate to establish a maneuver propulsion budget and lifetime analysis ([1], [2]). Fuel consumption may be assumed to be in terms of ΔV. -4. Automate the avoidance maneuver to continue indefinitely. This will be an iterative process to continue avoiding debris as orbit changes. -5. Design conceptual architecture of satellite including propulsion system and Attitude and Orbit Control System (AOCS). Consider using [System Composer](https://www.mathworks.com/products/system-composer.html) and [CubeSat Vehicle Model](https://www.mathworks.com/help/aeroblks/model-and-simulate-cubesats.html). - - Number and type of actuators. - - Number and type of sensors. -6. Design and implement AOCS and propulsion system control system in MATLAB and Simulink. - -## Background Material - -- [Get Started with Aerospace Toolbox](https://www.mathworks.com/help/aerotbx/getting-started.html) for general information about the toolbox. -- [Get Started with Aerospace Blockset](https://www.mathworks.com/help/aeroblks/getting-started-1.html) for general information about the blockset. -- [Space Applications](https://www.mathworks.com/help/aerotbx/satellite-scenario.html) has an overview of objects and functions for modeling, analyzing, and visualizing satellites with Aerospace Toolbox and Aerospace Blockset. -- [Space Applications — Examples](https://www.mathworks.com/help/aerotbx/examples.html?category=satellite-scenario) provides a set of examples satellite scenario modeling within the MATLAB environment. -- [Spacecraft](https://www.mathworks.com/help/aeroblks/spacecraft.html) has an overview of available blocks and examples for modeling, simulating and visualizing spacecraft with Aerospace Blockset. -- [Comparison of Orbit Propagators](https://www.mathworks.com/help/aerotbx/ug/comparison-of-orbit-propagators.html). -- [Constellation Modeling with the Orbit Propagator Block](https://www.mathworks.com/help/aeroblks/constellation-modeling-with-the-orbit-propagator-block.html) provides an example of how to propagate the orbits of a constellation of satellites in a Simulink model. -- [Mission Analysis with the Orbit Propagator Block](https://www.mathworks.com/help/aeroblks/mission-analysis-with-the-orbit-propagator-block.html) is an example showing how to perform line-of-sight access analysis from Simulink simulations. -- [Aerospace Blockset CubeSat Simulation Library](https://www.mathworks.com/matlabcentral/fileexchange/70030-aerospace-blockset-cubesat-simulation-library?s_tid=srchtitle_cubesat%20library_1). - -Suggested readings: - -[1] Patano, S., Myers, R., Aviles, J., and Bock, G., “GPM Orbital Maintenance Planning and Operations in Low Solar Activity Environment”, 2018 SpaceOps Conference, Marseille, France, May 2018. - -[2] Matko, D., Rodič, T., Oštir, K., Marsetič, A., and Peljhan, M., “Optimization of Fuel Consumption with Respect to Orbital Requirements for High Resolution Remote Sensing Satellite Constellations”, V: 25th AIAA/USU Conference on Small Satellites, Aug 8-11, 2011, Logan, UT, USA. - -[3] Office of the Chief Engineer, NASA, “Collision Avoidance for Space Environment Protection”, November 19, 2020, NID 7120.132. - -[4] Office of Safety and Mission Assurance, NASA, “NASA Procedural Requirements for Limiting Orbital Debris and Evaluating the Meteoroid and Orbital Debris Environments”, February 16, 2017, NPR 8715.6B. - -[5] Aida, Saika, “Conjunction Risk Assessment and Avoidance Maneuver Planning Tools”, DLR German Space Operations Center (GSOC), Oberpfaffenhofen, Weßling, Germany. - -[6] NASA, “NASA Spacecraft Conjunction Assessment and Collision Avoidance Best Practices Handbook”, NASA/SP-20205011318, Dec 2020. - -## Impact - -Contribute to the success of satellite mega-constellations and improve the safety of the Low Earth Orbit (LEO) environment. - -## Expertise Gained - -Autonomous Vehicles, Aerospace, Satellite, Control, Modeling and Simulation - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/57) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -225 diff --git a/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README-ah-rvalenti.md b/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README-ah-rvalenti.md deleted file mode 100644 index ef0e55d6..00000000 --- a/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README-ah-rvalenti.md +++ /dev/null @@ -1,73 +0,0 @@ - - -### Be the first to sign up for this project and receive a MathWorks T-shirt! -
- -**Project 148:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Selection of Mechanical Actuators Using Simulation-Based Analysis

-

Help accelerate the design and development of autonomous systems by providing a framework for mechanical actuators analysis and selection.

-
- -## Motivation - -Changing to novel actuation systems and/or system electrification often requires major redesign of a product. System-level simulation analysis using tools like Simscape™ and Simulink® is being used in many industries to manage deployment of these disruptive technologies. In this project you will develop analytical and simulation-based methods to help product designers make good and informed actuation choices. Through this you will develop skills that are of interest to multiple industries. - -## Project Description - -In this project you will explore the behaviors of different types of actuation. To provide a rigorous framework within which to compare actuation technologies, it is suggested that you make use of the property charts proposed by Huber, Fleck and Ashby [1]. You may want to focus on a subset of the available actuation technologies – for example you could compare different types of rotary electric motors, or compare all types of linear actuators (electric, hydraulic, pneumatic and variants thereof). - - -When comparing technologies, you will find that some properties require simulation-based analysis. For the latter, The MathWorks Simscape™ product family already includes example models of many common actuation technologies. You will probably also need to build some of your own models, possibly also creating custom Simscape blocks to represent novel actuation solutions. Something to consider is combination of two different actuator types to make up for inherent deficiencies in each. For example, take a look at the Hybrid Linear Actuator example in Simscape Electrical™. - - -The outcome of this project could be an analysis of two or more actuation technologies with results presented using the property charts proposed in [1]. The analysis should be presented as a short write-up with supporting Simulink/Simscape and associated MATLAB scripts used to produce the results. A good solution will be written in a way that enables others to build on your work by adding information for other actuation technologies. - - -This project could take many different directions, but by way of example, here are some suggested steps: - -1. Read and familiarize yourself with the Huber, Fleck and Ashby paper. -2. Think about what other property charts a product designer might find useful – is the list in the Huber paper complete? -3. Add the Hybrid Linear Actuator example actuation to the property charts. In what applications might this hybrid actuator be useful? -4. Come up with two or more other hybrid actuation ideas, developing supporting models in Simscape. Use these to add these hybrid actuation solutions to the property charts. -5. Write more generalized MATLAB scripts that can be used to automatically extract property charts from any Simscape actuator model. -6. Using MATLAB develop a solution that maps from an actuation requirement to a set of suitable technologies using data from the Huber paper and from your analysis of hybrid solutions. - - -Depending on how much time you have for your project, you might want to pick a subset of these steps. - - -[1] Huber, J.E., Fleck, N.A. & Ashby, M.F. “The selection of mechanical actuators based on performance indices”, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, Vol. 453, Issue 1965, pp. 2185-2205, 1997. - -## Background Material - -- [1] Huber, J.E., Fleck, N.A. & Ashby, M.F. “The selection of mechanical actuators based on performance indices”, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, Vol. 453, Issue 1965, pp. 2185-2205, 1997. [[link](https://royalsocietypublishing.org/doi/10.1098/rspa.1997.0117)] -- [Simscape](https://www.mathworks.com/products/simscape.html#ssfam) -- [Simscape Electrical](https://www.mathworks.com/help/physmod/sps/index.html?s_tid=CRUX_lftnav) -- [Simscape Fluids](https://www.mathworks.com/help/physmod/sps/index.html?s_tid=CRUX_lftnav) -- [Hybrid Linear Actuator](https://www.mathworks.com/help/physmod/sps/ug/hybrid-linear-actuator.html) - -## Impact - -Help evaluate and select actuation systems across multiple industries (robotic, automotive, manufacturing, aerospace) and help designers come up with novel actuation solutions. - -## Expertise Gained - -Drones, Robotics, Control, Cyber-physical Systems, Electrification, Humanoid, Manipulators, Modeling and Simulation - -## Project Difficulty - -Master's, Doctoral level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/14) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[rhyde005](https://github.com/rhyde005) - -## Project Number - -148 diff --git a/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README.md b/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README.md deleted file mode 100644 index 51461c2a..00000000 --- a/projects/Selection of Mechanical Actuators Using Simulation-Based Analysis/README.md +++ /dev/null @@ -1,73 +0,0 @@ - - -### Be the first to sign up for this project and receive a MathWorks T-shirt! -
- -**Project 148:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Selection of Mechanical Actuators Using Simulation-Based Analysis

-

Help accelerate the design and development of autonomous systems by providing a framework for mechanical actuators analysis and selection.

-
- -## Motivation - -Changing to novel actuation systems and/or system electrification often requires major redesign of a product. System-level simulation analysis using tools like Simscape™ and Simulink® is being used in many industries to manage deployment of these disruptive technologies. In this project you will develop analytical and simulation-based methods to help product designers make good and informed actuation choices. Through this you will develop skills that are of interest to multiple industries. - -## Project Description - -In this project you will explore the behaviors of different types of actuation. To provide a rigorous framework within which to compare actuation technologies, it is suggested that you make use of the property charts proposed by Huber, Fleck and Ashby [1]. You may want to focus on a subset of the available actuation technologies – for example you could compare different types of rotary electric motors, or compare all types of linear actuators (electric, hydraulic, pneumatic and variants thereof). - - -When comparing technologies, you will find that some properties require simulation-based analysis. For the latter, The MathWorks Simscape™ product family already includes example models of many common actuation technologies. You will probably also need to build some of your own models, possibly also creating custom Simscape blocks to represent novel actuation solutions. Something to consider is combination of two different actuator types to make up for inherent deficiencies in each. For example, take a look at the Hybrid Linear Actuator example in Simscape Electrical™. - - -The outcome of this project could be an analysis of two or more actuation technologies with results presented using the property charts proposed in [1]. The analysis should be presented as a short write-up with supporting Simulink/Simscape and associated MATLAB scripts used to produce the results. A good solution will be written in a way that enables others to build on your work by adding information for other actuation technologies. - - -This project could take many different directions, but by way of example, here are some suggested steps: - -1. Read and familiarize yourself with the Huber, Fleck and Ashby paper. -2. Think about what other property charts a product designer might find useful – is the list in the Huber paper complete? -3. Add the Hybrid Linear Actuator example actuation to the property charts. In what applications might this hybrid actuator be useful? -4. Come up with two or more other hybrid actuation ideas, developing supporting models in Simscape. Use these to add these hybrid actuation solutions to the property charts. -5. Write more generalized MATLAB scripts that can be used to automatically extract property charts from any Simscape actuator model. -6. Using MATLAB develop a solution that maps from an actuation requirement to a set of suitable technologies using data from the Huber paper and from your analysis of hybrid solutions. - - -Depending on how much time you have for your project, you might want to pick a subset of these steps. - - -[1] Huber, J.E., Fleck, N.A. & Ashby, M.F. “The selection of mechanical actuators based on performance indices”, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, Vol. 453, Issue 1965, pp. 2185-2205, 1997. - -## Background Material - -- [1] Huber, J.E., Fleck, N.A. & Ashby, M.F. “The selection of mechanical actuators based on performance indices”, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, Vol. 453, Issue 1965, pp. 2185-2205, 1997. [[link](https://royalsocietypublishing.org/doi/10.1098/rspa.1997.0117)] -- [Simscape](https://www.mathworks.com/products/simscape.html#ssfam) -- [Simscape Electrical](https://www.mathworks.com/help/physmod/sps/index.html?s_tid=CRUX_lftnav) -- [Simscape Fluids](https://www.mathworks.com/help/physmod/sps/index.html?s_tid=CRUX_lftnav) -- [Hybrid Linear Actuator](https://www.mathworks.com/help/physmod/sps/ug/hybrid-linear-actuator.html) - -## Impact - -Help evaluate and select actuation systems across multiple industries (robotic, automotive, manufacturing, aerospace) and help designers come up with novel actuation solutions. - -## Expertise Gained - -Drones, Robotics, Control, Cyber-physical Systems, Electrification, Humanoid, Manipulators, Modeling and Simulation - -## Project Difficulty - -Master's, Doctoral level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/14) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[rhyde005](https://github.com/rhyde005) - -## Project Number - -148 diff --git a/projects/Sensor Fusion for Autonomous Systems/README.md b/projects/Sensor Fusion for Autonomous Systems/README.md deleted file mode 100644 index e3a2f48d..00000000 --- a/projects/Sensor Fusion for Autonomous Systems/README.md +++ /dev/null @@ -1,71 +0,0 @@ -**Project 233:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Sensor Fusion for Autonomous Systems

-

Develop a sensor fusion algorithm for vehicle pose estimation using classical filtering or AI-based techniques.

-
- -## Motivation - -A key piece of any autonomous system is answering the question “Where am I?” This can be achieved with high accuracy today because of the abundance of sensors available. But each autonomous vehicle works in a specific environment and requires specially chosen sensors. - -Each sensor measures its environment with varying precision and has unique benefits and drawbacks. Classical techniques like the Extended Kalman Filter are often used to combine varying sensors to achieve a higher precision estimate of a vehicle’s position and orientation (pose) than any one sensor could achieve alone. Recently, machine learning and deep learning approaches have tried to improve localization performance and, in some cases, have been able to achieve outcomes which are not possible with classical filtering. - - -## Project Description - -This project focuses on fusing sensors on a ground robot or quadcopter to determine position and orientation. The [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html), [Automated Driving Toolbox]( https://www.mathworks.com/products/automated-driving.html), [LIDAR toolbox]( https://www.mathworks.com/products/lidar.html), [Navigation Toolbox™](https://www.mathworks.com/products/navigation.html), [UAV Toolbox](https://www.mathworks.com/products/uav.html), and [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html) enable you to simulate a vehicle trajectory and many commonly used sensors. Popular sensors for autonomous systems include IMUs, GPS, LIDAR, visual odometry, wheel encoder odometry, altimeters, and pitot tubes among many others. In this project you will design a fusion algorithm for a group of these sensors to localize your ground robot or quadcopter - -Suggested steps: -1. Use one of the many publicly available inertial navigation datasets, including but not limited to [1], [2], [3], [4]. Each of these datasets uses some subset of the aforementioned sensors. -2. Use an [insEKF](https://www.mathworks.com/help/nav/ref/insekf.html) in Navigation Toolbox to create an extended Kalman filter to fuse the simulated sensor data and compare it to the recorded ground truth in the dataset. You may need to develop sensor plugins for the [insEKF]( https://www.mathworks.com/help/nav/ref/insekf.html), like the [insGyroscope](https://www.mathworks.com/help/nav/ref/insgyroscope.html?searchHighlight=insGyroscope&s_tid=srchtitle_insGyroscope_1) or [insAccelerometer](https://www.mathworks.com/help/nav/ref/insaccelerometer.html?searchHighlight=insAccelerometer&s_tid=srchtitle_insAccelerometer_1). See [5] for a description of how to build a custom sensor. -3. Make your filter robust to sensor dropout by detecting bad sensor data and/or adding sensors. You can manually modify/corrupt/remove some of the recorded sensor data to see how your filter handles the situation. - -Project variations: -- Build trajectory simulation using one of the toolboxes listed above. Consider using the [uavScenario](https://www.mathworks.com/help/uav/ug/uav-scenario-tutorial.html), [robotScenario](https://www.mathworks.com/help/robotics/ref/robotscenario.html), [drivingScenario](https://www.mathworks.com/help/driving/ref/drivingscenario.html) or [waypointTrajectory](https://www.mathworks.com/help/fusion/ref/waypointtrajectory-system-object.html) depending on which autonomous system you are modeling. Save the ground truth pose of the vehicle created by this trajectory. Use the sensor simulators in these toolboxes ([imuSensor](https://www.mathworks.com/help/nav/ref/imusensor-system-object.html?searchHighlight=imusensor&s_tid=srchtitle_imusensor_2), [gpsSensor](https://www.mathworks.com/help/nav/ref/gpssensor-system-object.html?searchHighlight=gpsSensor&s_tid=srchtitle_gpsSensor_1), etc) to simulate input to an insEKF. Tune the insEKF and compare its performance to ground truth. -- Use a machine learning or deep learning based approach to fusing data rather than an extended Kalman filter, such as in [6] and [7]. - - -## Background Material - -Datasets: - -[1] [The EuRoC MAV Dataset](https://projects.asl.ethz.ch/datasets/doku.php?id=kmavvisualinertialdatasets) M. Burri, J. Nikolic, P. Gohl, T. Schneider, J. Rehder, S. Omari, M. Achtelik and R. Siegwart, The EuRoC micro aerial vehicle datasets, International Journal of Robotic Research, DOI: 10.1177/0278364915620033, early 2016. - -[2] [The TUM VI Dataset](https://vision.in.tum.de/data/datasets/visual-inertial-dataset) Klenk, Simon, et al. "TUM-VIE: The TUM Stereo Visual-Inertial Event Dataset." 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021. - -[3] [The KTTI Dataset](http://www.cvlibs.net/datasets/kitti/) Geiger, Andreas, et al. "Vision meets robotics: The kitti dataset." The International Journal of Robotics Research 32.11 (2013): 1231-1237. - -[4] [ANSFL Dataset](https://github.com/ansfl/Navigation-Data-Project/) A. Shurin et al., "The Autonomous Platforms Inertial Dataset," in IEEE Access, vol. 10, pp. 10191-10201, 2022, doi: 10.1109/ACCESS.2022.3144076. - -Examples and papers: - -[5] [Design Fusion Filter for Custom Sensors](https://www.mathworks.com/help/nav/ug/design-fusion-filter-for-custom-sensors.html) - -[6] Brossard, Martin, Silvere Bonnabel, and Axel Barrau. "Denoising imu gyroscopes with deep learning for open-loop attitude estimation." IEEE Robotics and Automation Letters 5.3 (2020): 4796-4803. -Suggested readings: - -[7] Esfahani, Mahdi Abolfazli, et al. "OriNet: Robust 3-D orientation estimation with a single particular IMU." IEEE Robotics and Automation Letters 5.2 (2019): 399-406. - - -## Impact - -Enhance navigation accuracy of autonomous vehicles. - -## Expertise Gained - -Autonomous Vehicles, Sensor Fusion and Tracking, State Estimation - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/67) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -233 diff --git a/projects/Sentiment Analysis in Cryptocurrency Trading/README.md b/projects/Sentiment Analysis in Cryptocurrency Trading/README.md deleted file mode 100644 index fbe5dc73..00000000 --- a/projects/Sentiment Analysis in Cryptocurrency Trading/README.md +++ /dev/null @@ -1,75 +0,0 @@ -**Project 239:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Sentiment Analysis in Cryptocurrency Trading

-

Build your own cryptocurrency trading strategies based on sentiment analysis.

-
- -## Motivation - -Recently, more than 2,300 US businesses accept bitcoin, according to an estimate from late 2020. The cryptocurrency trading presents a host of opportunities and challenges. External factors, such as financial news and social media, have wide impacts on the crypto price movements. For this reason, it is important to research on the effectiveness of twitter posts, as a main social media platform, to the cryptocurrency trading strategies. Sentiment analysis is a sub-research area of Natural Language Processing (NLP) studies. Twitter sentiment analysis can provide insights that indicate positive or negative attitudes on the cryptocurrency. Moreover, due to the high volatility of cryptocurrency price, the analysis on social media sentiments and cryptocurrencies’ time series will improve the trading strategies a lot. Based on your research, build the optimal portfolio of cash and cryptocurrencies to maximize the revenues given a certain level of risk. - -## Project Description - -Analyze data from Twitter posts about cryptocurrency to discover the current overall feelings of people towards cryptocurrency and use your findings to build the optimal trading strategy. -Work with the [Datafeed Toolbox™](https://www.mathworks.com/products/datafeed.html), [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/products/statistics.html) and [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html ) products to retrieve the social media data, build the time series models with the sentiment scores, and trading strategies. - -Suggested steps: -1. Become familiar with the MATLAB based deep learning, text analytics examples listed in Background Material section below. -2. Retrieve the tweets from Twitter on cryptocurrencies using the [Twitter connection object]( https://www.mathworks.com/help/datafeed/twitter.html) from the Datafeed Toolbox -3. Apply a [classification algorithm]( https://www.mathworks.com/help/stats/classification.html) from the Statistics Machine learning and Statistics Toolbox to determine the sentiment scores and compare the results against the existing [VADER](https://www.mathworks.com/help/textanalytics/ref/vadersentimentscores.html) and [ratio rule](https://www.mathworks.com/help/textanalytics/ref/ratiosentimentscores.html) methods from the [Text Analytics toolbox](https://www.mathworks.com/help/stats/classification.html) -4. Build the time series model of cryptocurrency, considering the sentiment scores as a factor in the model using the [Econometrics Toolbox](https://www.mathworks.com/products/econometrics.html). -5. Design algorithm trading strategies on cryptocurrency using [Financial Toolbox](https://www.mathworks.com/products/finance.html). Back test the portfolio performance. - -Project variations: -1. Apply [python integration in MATLAB]( https://www.mathworks.com/help/matlab/call-python-libraries.html) to overcome the current limited data range retrieval functionality. -2. Model selection: Select one or more of these models to test the performance of the trading strategy: cryptocurrency time series model, factor model, deep learning or reinforcement learning model. - -Advanced project work: -1. Build an interactive app on trading strategies on cryptocurrency data. -2. Utilize more advanced techniques to improve the performance of the crypto currency trading strategies. For example see some methods introduced in the papers in the background material section. -3. Immigrate the same methodology onto the equity market or fixed-income market - - -## Background Material - -- [Sentiment Analysis in MATLAB](https://www.mathworks.com/discovery/sentiment-analysis.html?s_tid=srchtitle) -- [Text Analytics Examples](https://www.mathworks.com/help/textanalytics/examples.html) -- [Twitter API](https://developer.twitter.com/en/docs/twitter-api) -- [Back test examples](https://www.mathworks.com/help/finance/backtest-investment-strategies.html) -- [Algorithmic trading in MATLAB]( https://www.mathworks.com/discovery/algorithmic-trading.html?s_tid=srchtitle_algorithm%2520trading_1) -- [Time series model example]( https://www.mathworks.com/help/econ/introduction-to-vector-autoregressive-var-models.html?s_tid=srchtitle_vector%20autoregression_1) -- [Factor model example](https://www.mathworks.com/help/finance/portfolio-optimization-using-factor-models.html) -- [Reinforcement Learning Example](https://www.mathworks.com/matlabcentral/fileexchange/74176-reinforcement-learning-for-financial-trading) -- [Deep learning with time series sequences](https://www.mathworks.com/help/deeplearning/examples.html?category=deep-learning-with-time-series-sequences-and-text&s_tid=CRUX_topnav) - -Suggested readings: - -[1] Mittal, Anshul. “Stock Prediction Using Twitter Sentiment Analysis.” (2011). - -[2] E. Şaşmaz and F. B. Tek, "Tweet Sentiment Analysis for Cryptocurrencies," 2021 6th International Conference on Computer Science and Engineering (UBMK), 2021, pp. 613-618, doi: 10.1109/UBMK52708.2021.9558914. - -[3] J. Bollen and H. Mao, "Twitter Mood as a Stock Market Predictor" in Computer, vol. 44, no. 10, pp. 91-94, 2011. doi: 10.1109/MC.2011.323. - - -## Impact - -Have a foundation on the potential opportunities on Environmental, Social, and Governance (ESG) portfolio analysis. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Machine Learning, Text Analytics - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/75) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -239 diff --git a/projects/Signal Coverage Maps Using Measurements and Machine Learning/README.md b/projects/Signal Coverage Maps Using Measurements and Machine Learning/README.md deleted file mode 100644 index 2783587b..00000000 --- a/projects/Signal Coverage Maps Using Measurements and Machine Learning/README.md +++ /dev/null @@ -1,53 +0,0 @@ -**Project 151:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Signal Coverage Maps Using Measurements and Machine Learning

-

Reduce the cost of 5G and IoT network deployment by generating coverage maps from limited measurements.

-
- -## Motivation - -New 5G and IoT systems are driving the deployment of dense small cell networks and leading the way to the smart cities of tomorrow. -Companies need to use signal coverage maps to optimize and validate their networks but collecting signal measurements to generate those maps is an expensive process. -To reduce cost there is a need to apply prediction models to generate coverage maps from limited, sparse measurement data. - -## Projects Description - -Use MATLAB® and toolboxes to implement, explore, and compare different techniques for predicting signal coverage from real signal measurements. Develop propagation models from the measurements data and use them to generate coverage maps for new scenarios. - -Suggested steps: - -1. Locate signal strength measurement data sets. One portal of wireless data which may help is [CRAWDAD](https://crawdad.org/keyword-signal-strength.html). An example data set is available from [mySignals](http://www.mysignals.gr/research.php). -2. Research data-driven techniques used to predict coverage maps from signal strength measurements. In particular, study [City-Wide Signal Strength Maps: Prediction with Random Forests](https://dl.acm.org/doi/fullHtml/10.1145/3308558.3313726). -3. Design and implement data-driven propagation models in MATLAB using the measurements data. Using the paper mentioned above as a guide, implement one model as a baseline using geospatial interpolation techniques, and implement another model using machine learning techniques such as random forests. -4. Generate coverage maps using your propagation models and compare your results against theoretical propagation models which are available in Communications Toolbox. - -## Background Material - -- [Communication Toolbox™](https://www.mathworks.com/products/communications.html) -- [Propagation and Channel Models](https://www.mathworks.com/help/comm/propagation-and-channel-models.html) - -## Impact - -Contribute to the evolution and deployment of new wireless communications systems. - -## Expertise Gained - -Artificial Intelligence, 5G, Machine Learning, Wireless Communication - -## Project Difficulty - -Master's, Doctoral level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/16) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[jhalbr](https://github.com/jhalbr) - -## Project Number - -151 - diff --git a/projects/Signal Coverage Maps Using Measurements and Machine Learning/student submissions/coverageMap b/projects/Signal Coverage Maps Using Measurements and Machine Learning/student submissions/coverageMap deleted file mode 160000 index e6f8cdc0..00000000 --- a/projects/Signal Coverage Maps Using Measurements and Machine Learning/student submissions/coverageMap +++ /dev/null @@ -1 +0,0 @@ -Subproject commit e6f8cdc0662ca3059d74177ca96b5669223526c2 diff --git a/projects/Signal Integrity Channel Feature Extraction for Deep Learning/README.md b/projects/Signal Integrity Channel Feature Extraction for Deep Learning/README.md deleted file mode 100644 index 5bfd4d3d..00000000 --- a/projects/Signal Integrity Channel Feature Extraction for Deep Learning/README.md +++ /dev/null @@ -1,111 +0,0 @@ -**Project 198:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Signal Integrity Channel Feature Extraction for Deep Learning

-

Develop a deep learning approach for signal integrity applications.

-
- -## Motivation - -Modern society is dependent on a fast, reliable, and inexpensive internet. Signal integrity engineers design the boards, packages and computer systems that carry internet traffic and provide internet access. New approaches would enable more robust and faster communications systems that are delivered to market sooner. - -So far, deep learning has not significantly impacted computer interconnect design. Many approaches have been tried but the channel S-parameter models are much too complex and difficult to apply [[1](#schierholz), [2](#swaminathan)]. Among them is the lack of data to train DL models with. Further each signal integrity system design has its own unique characteristics, and a general DL model would likely not apply to more than a few designs. - -Explore using a segmented parameterized channel model approach to extract channel features that may lead to better neural network models of the interconnect. - -## Project Description - -This project proposes a feature extraction approach where the total end-to-end channel S-parameter is represented by a series of parameterized transmission lines. If this decomposition of the channel into sub-blocks has sufficient fidelity, then it can be used as proxy for the total end-to-end channel. Further, the parameters of certain sub-blocks represent design variables, like transmission line length, and other parameters represent manufacturing variation variables, like characteristic impedance and the span of these variables forms the design parameter space. By judiciously choosing values from this space, a set of synthetic system end-to-end features, S-parameters and system performance metrics can be used to train a neural network (NN). - -Determining the interconnect features is analogous to the labeling process of an image for an automated driving application. The project is roughly divided into two main questions, 1) Can the features be used to predict system performance? And 2) can the features be automatically extracted from end-to-end system S-parameters? - -Suggested steps: - -1. Perform literature research prior to starting the work. -2. Validate general approach, Can the features be used to predict system performance with a NN?: - 1. Forward model creation with [SerDes Toolbox™](https://www.mathworks.com/products/serdes.html): - 1. Transmission line creation [[3](#serdesfun), [4](#ctlm)] - 2. Cascading of S-parameters [[5](#cascade)] - 3. Convert to time-domain [[3](#serdesfun)] - 4. Quantify system performance [[6](#optpulse)] - 2. Prove that transmission line parameters can be used as features [[12](#fe)] for NN modeling. - 1. For a network of N<10 transmission line blocks, vary length and impedance parameters to create M≈1000 end-to-end S-parameters. Quantify system performance for each. - 2. Use the Deep Learning Toolbox to train a NN model on 90% of the data and validate on remaining 10%. - -Advanced project work: -1. Inverse process [13](#ip) part I: Extract the features from synthetic channel S-parameter models. - 1. For synthetic end-to-end S-parameters like those created in step 2.b.i, extract the transmission line parameters. - 2. Ideas to explore: - 1. Impedance peeling algorithms [[7](#liu), [8](#schuster)] - 2. De-embedding [[9](#deembed)] - 3. Cascading S-parameter Linearization Technique (CSLT) [[10](#allred21), [11](#allred18)] -2. Inverse process part II: Extract the features from general S-parameter models. - 1. Select an end-to-end channel model S-parameter [[1](#schierholz), [14](#toolschannels)] - 2. Apply Inverse process to extract a representative set of parameterized transmission lines that best fits the data. -3. Create synthetic DL data set - 1. Using the extracted features, vary the trace length, impedance or other properties to create a training set of end-to-end S-parameters. - 2. Train NN on data set - 3. Identify ways to validate approach - -## Background Material - -[1] M. Schierholz et al., "SI/PI-Database of PCB-Based Interconnects for Machine Learning Applications," in IEEE Access, vol. 9, pp. 34423-34432, 2021 - -[2] M. Swaminathan, H. M. Torun, H. Yu, J. A. Hejase and W. D. Becker, "Demystifying Machine Learning for Signal and Power Integrity Problems in Packaging," in IEEE Transactions on Components, Packaging and Manufacturing Technology, vol. 10, no. 8, pp. 1276-1295, Aug. 2020. --- see section “IV.C Problem 8 - High speed channel signaling” - -[3] See SerDes Toolbox function: serdes.ChannelLoss.causalTransmissionLine - -[4] “[Proposal for a causal transmission line model](http://www.ieee802.org/3/bj/public/mar14/healey_3bj_01_0314.pdf)”, IEEE P802.3bj Task Force, March 2014. - -[5] [Cascading of S-parameters](https://www.mathworks.com/help/rf/ref/cascadesparams.html) - -[6] [Pulse response metric for optimization routines](https://www.mathworks.com/help/serdes/ref/optpulsemetric.html) - -[7] Liu, P., J. Zhang, and J. Fang, “Accurate characterization of lossy interconnects from TDR waveforms," 2013 IEEE 22nd Conference on Electrical Performance of Electronic Packaging and Systems, 187-190, 2013. - -[8] Schuster, C. and W. Fichtner, “Signal integrity analysis of interconnects using the FDTD method and a layer peeling technique," IEEE Transactions on Electromagnetic Compatibility, Vol. 42, No. 2, 229-233, 2000. - -[9] [De-Embedding S-Parameters](https://www.mathworks.com/help/rf/ug/de-embedding-s-parameters.html) - -[10] R. J. Allred and C. Furse, “Reflection budgeting methodology for high-speed serial link signal integrity design,” Progress In Electromagnetics Research B, vol. 91, pp. 59–77, 2021. - -[11] R. J. Allred and C. M. Furse, “Linearization of s-parameter cascading for analysis of multiple reflections,” Applied Computational Electromagnetics Society Journal, vol. 33, no. 12, 2018. - -[12] [Feature_extraction](https://en.wikipedia.org/wiki/Feature_extraction) - -[13] [Inverse problem](https://en.wikipedia.org/wiki/Inverse_problem) - -[14] [IEEE P802.3ck Task Force - Tools and Channels](https://www.ieee802.org/3/ck/public/tools/index.html) - -## Impact - -Accelerate signal integrity design and analysis to enable society with more robust and connected internet communications. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Machine Learning, Modeling and Simulation, Neural Networks, RF and Mixed Signal - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/29) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[richardallredmathworks](https://github.com/richardallredmathworks) - -## Project Number - -198 - -## - - -
- -### Be the first to sign up for this project and receive a MathWorks T-shirt! diff --git a/projects/Simulation-Based Design of Humanoid Robots/README.md b/projects/Simulation-Based Design of Humanoid Robots/README.md deleted file mode 100644 index d2770933..00000000 --- a/projects/Simulation-Based Design of Humanoid Robots/README.md +++ /dev/null @@ -1,75 +0,0 @@ -**Project 170:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Simulation-Based Design of Humanoid Robots

-

Develop and use models of humanoid robots to increase understanding of how best to control them and direct them to do useful tasks.

-
- -## Motivation - -Robots that can be repurposed are predicted to revolutionize many things, ranging from the construction industry to healthcare. Although great progress has been made both in actuation and sensing hardware, understanding how to control these types of robot is still in its infancy. Fast-simulating models provide a way to explore solutions to this, whether it be related to lower-level motion control or to higher-level task programming using AI. - -## Project Description - -This project could take on many different forms depending on your interests. A good starting point is the humanoid robot example already modeled in Simscape Multibody™ that is linked from the background material section below. Review this model and also the associated MATLAB® live scripts that design algorithms that make the robot walk. Having done this, it is then suggested that you pick at least one modeling task and one research task as the basis of your project. Some ideas for modeling and research tasks are listed below. - -**Modeling tasks:** -- Add more degrees of freedom e.g. in the hip and in the foot. Currently the robot relies on having a wide flat foot and flat ground surface. Adding another degree of freedom at the foot will help with balance when coupled with suitable control. -- Add actuation and sensor models. The model currently assumes ideal actuators that deliver demanded forces and torques. Replacing these with models of real actuators will bring some more real-world realism into the model. You will find examples of electrical, hydraulic, and pneumatic actuation in Simscape™ product examples to help you get started. -- Automate definition of the robot from MATLAB to support easy sizing of the robot. -  - - **Research tasks:** -- Develop algorithms to manage balance as measured by, for example, the minimum force required to make it fall over. You might want to tackle this task before trying any of the motion tasks to gain understanding and/or to provide an inner-loop balance feedback system. -- Develop algorithms to perform specific tasks such as walking, stopping, crouching, jumping, reaching/leaning. Consider using conventional approaches and/or AI methods such as reinforcement learning. -- Develop motion planning and control algorithms to perform more complex tasks. For example, can you get the robot to hit an incoming cricket ball or baseball? In the context of healthcare, can you develop solutions to one robot helping stabilize a second one that has more limited actuator capability? For construction, can you get two robots to coordinate lifting an I-beam from its two ends? Again, consider using conventional approaches and/or AI methods such as reinforcement learning and deep learning. -- Research controller architectures. Humans and animals have a structure of multiple nested feedback loops; would this help with a robot’s balance too? What is the role for pattern generators and feedforward in robotics versus other approaches? -- Does including natural mechanical compliance into the actuation system help or hinder control? Under what circumstances? This may be particularly pertinent to maintaining balance. How does compliance help with efficiency of locomotion? - - -## Background Material - -It is suggested that you start with this [humanoid robot example](https://www.mathworks.com/help/physmod/sm/ug/humanoid_walker.html). -The example includes two approaches to getting the robot to walk. - -For electrical actuation modeling, take a look at: -- [Motor & Drive modeling](https://www.mathworks.com/help/physmod/sps/ref/motordrivesystemlevel.html) -- [RC Servo modeling](https://www.mathworks.com/help/physmod/sps/ref/rcservo.html) -- [DC Motor](https://www.mathworks.com/help/physmod/sps/ref/dcmotor.html) - -For hydraulic and pneumatic actuation, take a look at: -- [Air muscles (McKibben actuation)](https://www.mathworks.com/help/physmod/hydro/ug/antagonistic-mcKibben-muscle-actuator.html) -- [Hydraulic actuation](https://www.mathworks.com/help/physmod/hydro/ug/creating-a-simple-model.html) - -For control and AI (including reinforcement learning) see: -- [Control Systems Toolbox](https://www.mathworks.com/products/control.html) -- [Reinforcement Learning Toolbox](https://www.mathworks.com/products/reinforcement-learning.html) -- [Deep Learning Toolbox](https://www.mathworks.com/products/deep-learning.html) -- [Optimization Toolbox](https://www.mathworks.com/products/optimization.html) -- [Global Optimization Toolbox](https://www.mathworks.com/products/global-optimization.html) - - -## Impact - -Accelerate the deployment of humanoid robots to real-world tasks including in healthcare, construction, and manufacturing - -## Expertise Gained - -Artificial Intelligence, Robotics, Control, Cyber-Physical Systems, Deep Learning, Humanoid, Human-Robot Interaction, Machine Learning, Mobile Robots, Modeling and Simulation, Optimization, Reinforcement Learning - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/19) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By -[rhyde005](https://github.com/rhyde005) - -## Project Number - -170 diff --git a/projects/Simulink Hearing Aid/README.md b/projects/Simulink Hearing Aid/README.md deleted file mode 100644 index decb7963..00000000 --- a/projects/Simulink Hearing Aid/README.md +++ /dev/null @@ -1,91 +0,0 @@ -**Project 241:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Simulink Hearing Aid

-

Develop a hearing aid simulation in Simulink.

-
- -## Motivation - -Over 1.8 billion people worldwide suffer from some level of hearing loss, however, only around [1 in 5](https://www.hearingloss.org/wp-content/uploads/HLAA_HearingLoss_Facts_Statistics.pdf) of those people use a hearing aid. Many medical device companies are working towards creating hearing aids that are cheaper, last longer, and work better, to improve their adoption amongst those who need them. The development of open-source hearing aid models could provide a platform to help improve both the efficacy, and accessibility, of hearing aids. - -## Project Description - -Use the [Audio Toolbox™](https://www.mathworks.com/products/audio.html) and/or [Simulink™](https://www.mathworks.com/products/simulink.html) to implement a block-based hearing aid simulation. Test in real time using a headset with microphone as the input, or with pre-recorded audio. - -Suggested steps: - -- Perform a literature review to discover the signal processing blocks commonly included in hearing aids and understand their purpose. - -- Create a simple input->output model that performs no signal processing but includes a filtered and delayed feedback loop from output->input to simulate crosstalk between microphone and speakers. - -- Develop signal processing blocks and link them together to create a full hearing aid model. - -Suggested signal processing blocks: - -- Single or multi-band dynamic range compression (‘DRC’) (and/or automatic gain control (‘AGC’)). - -- Noise suppression / reduction (‘NR’). - -- Filtering / equalization. - -- Feedback suppression / cancellation (‘DFS’/’DFC’) or adaptive feedback cancellation (‘AFC’). - -Advanced project extension(s): - -- Deploy the completed model to a cell phone, Raspberry Pi™ or other similar device, with an attached headset/microphone. - -- Adaption: Automatically adapt algorithm parameters to better suit the current listening conditions. Can include classification of the current environment, e.g., ‘music’, ‘speech’, ‘noise’. - -## Background Material - -[MATLAB and Simulink for Hearing Aids](https://www.mathworks.com/solutions/medical-devices/hearing-aids.html) - -Real-Time Audio in MATLAB: - -- [Real-Time Audio in Simulink](https://www.mathworks.com/help/audio/gs/real-time-audio-in-simulink.html) - -- [Real-Time Audio in MATLAB](https://www.mathworks.com/help/audio/gs/real-time-audio-in-matlab.html) - -Audio signal processing examples: - -- [Audio Processing Algorithm Design](https://www.mathworks.com/help/audio/audio-processing-algorithm-design.html?s_tid=CRUX_lftnav) - -- [Acoustic Noise Cancellation using LMS](https://www.mathworks.com/help/audio/ug/acoustic-noise-cancellation-using-lms.html) - -- [Cochlear Implant Speech Processor](https://www.mathworks.com/help/audio/ug/cochlear-implant-speech-processor.html) - -- [Active Noise Control with Simulink Real-Time](https://www.mathworks.com/help/audio/ug/active-noise-control-with-simulink.html) and [video](https://www.mathworks.com/videos/active-noise-control-from-modeling-to-real-time-prototyping-1561451814853.html) - -- [Code generation and deployment](https://www.mathworks.com/help/audio/examples.html?category=code-generation-and-deployment&s_tid=CRUX_topnav) - -Reading materials: - -[1] [Launer, S., Zakis, J.A., Moore, B.C.J. (2016). Hearing Aid Signal Processing. ](https://link.springer.com/chapter/10.1007/978-3-319-33036-5_4) - -[2] [H. Puder, Hearing aids: an overview of the state-of-the-art, challenges, and future trends of an interesting audio signal processing application, ISPA 2009.](https://ieeexplore.ieee.org/abstract/document/5297793) - -[3] [Kates JM, Principles of Digital Dynamic-Range Compression, Trends in Amplification.](https://journals.sagepub.com/doi/full/10.1177/108471380500900202) - -[4] [F. Strasser and H. Puder, Adaptive Feedback Cancellation for Realistic Hearing Aid Applications, TASLP 2015.](https://ieeexplore.ieee.org/abstract/document/7268853) - -## Impact - -Improve hearing aid simulation and create a testbed for new audio processing algorithm prototyping. - -## Expertise Gained - -Signal Processing, Audio, Modeling and Simulation - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/79) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -241 diff --git a/projects/Smart Watering System with Internet of Things/README.md b/projects/Smart Watering System with Internet of Things/README.md deleted file mode 100644 index dd38abce..00000000 --- a/projects/Smart Watering System with Internet of Things/README.md +++ /dev/null @@ -1,83 +0,0 @@ -**Project 219:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Smart Watering System with Internet of Things

-

Develop a smart plant water system using Internet of Things (IoT) and low-cost hardware

-
- -## Motivation - -Drought and hunger affect hundreds of millions of people around the world. Pressures from increasing population, inequity, climate change, and water shortages contribute to this challenge. Agriculture accounts for approximately 80 percent of the consumptive water use in the United States. Cutting-edge technologies like AI and IoT will be instrumental in fostering sustainable agricultural practices. -Efficient irrigation systems can help conserve resources and maintain farm profitability in an era of increasing food demand and rising costs. IoT-driven systems can automatically monitor, analyze, and precisely regulate water demand and supply, to minimize water overuse and eliminate the need for human intervention. -You can be a pioneer in the field of smart agriculture. Use cutting-edge technology to save water and optimize agricultural practices. - - -## Project Description - -The Internet of Things gives increased access to and control of smart devices. Sensors on these smart devices measure certain types of data, and the power of IoT can transfer that data to the cloud, where it can be used to make real-time decisions. - -Design an IoT-enabled system that will manage plant irrigation to optimize water use. Use low-cost hardware to interface with sensors, collect the data on the cloud and determine when and how the plants are watered. - -Suggested steps: -1. Collect agriculture data from sensors (Temp., Humidity, Soil Moisture, etc) -2. Upload this data to [ThingSpeak™](https://thingspeak.com/) (IoT analytics platform) -3. Access and Analyze this data using ThingSpeak, see some [Examples](https://www.mathworks.com/help/thingspeak/examples.html). E.g., Estimate [Evapotranspiration](http://www.fao.org/3/X0490E/x0490e0a.htm) of the field based on the input data -4. Gather other relevant information like weather forecast data or crop parameters -5. Determine a watering strategy based on the crop type and the data you have gathered. For example, identify how your measurements and data could help develop a predictive model in MATLAB® to determine when the crop needs to be watered -6. Send an action based on the analysis. This action could be as simple as a notification (tweet, e-mail) to water the plants or actuate a watering system at the plant. - -Here are the various phases of the project: -- Develop the requirements -- Design the architecture and specifications of the system -- Select the hardware (sensors, embedded systems) -- Build the system -- Test it on a simplified use case -- Develop analytics for the system (E.g., Forecast the amount of water used over the season for a location) - -Advanced project work: -- Add more sensors to gather different types of data such as light, airflow, and pH -- Develop models for soil management (e.g., fertilization) -- Imaging to monitor crop health (e.g., webcam, hyperspectral, or drone) -- Develop AI models for Weather Prediction/Soil Characterization - - -## Background Material - -- [Collect Agricultural Data over The Things Network](http://www.mathworks.com/help/thingspeak/things_network_ag_data.html) -- [Arduino Based Smart Watering of Plants](https://www.mathworks.com/help/supportpkg/arduino/examples/arduino-based-smart-watering-of-plants.html) -- [Forecast Tidal Depths Using ThingSpeak Data](https:/www.mathworks.com/help/thingspeak/forecast-tidal-wave-depths.html) -- [How to gather data from weather forecast?](https://www.mathworks.com/matlabcentral/answers/417426-how-to-gather-data-from-weather-forecast#answer_335736) -- [Analyzing weather data from an Arduino-based weather station](https://www.mathworks.com/matlabcentral/fileexchange/47049-analyzing-weather-data-from-an-arduino-based-weather-station) - -Videos: -- [Using ThingSpeak for IoT in Agriculture](https://www.mathworks.com/videos/using-thingspeak-for-iot-in-agriculture-1594044754903.html) -- [Using MATLAB to Empower Modern Numerical Weather Forecasts](https://www.mathworks.com/videos/using-matlab-to-empower-modern-numerical-weather-forecasts-1562096395625.html) -- [Machine Learning for Agriculture](https://www.mathworks.com/videos/machine-learning-for-agriculture-1600457289413.html) -- [Build a Solar Tracking System using Simulink and ThingSpeak](https://www.youtube.com/watch?v=57GxzjSaKhA) - -Suggested readings: - -[Real Time Weather Analysis Using ThingSpeak](https://acadpubl.eu/hub/2018-120-6/1/46.pdf) - - -## Impact - -Minimize the negative effects of the overuse of water in farming and preserve water resources. - -## Expertise Gained - -Sustainability and Renewable Energy, Artificial Intelligence, IoT, Low-Cost Hardware, Deep Learning, Cloud Computing - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/51) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -219 diff --git a/projects/Snake-like Robot Modeling and Navigation/README.md b/projects/Snake-like Robot Modeling and Navigation/README.md deleted file mode 100644 index 64191a11..00000000 --- a/projects/Snake-like Robot Modeling and Navigation/README.md +++ /dev/null @@ -1,71 +0,0 @@ -**Project 224:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Snake-like Robot Modeling and Navigation

-

Model and control an autonomous snake-like robot to navigate an unknown environment.

-
- -## Motivation - -Snake-like robots offer impressive dexterity owing to their high degree-of-freedom (DOF) serial linkage. These bio-inspired designs are becoming increasingly popular for inspection tasks, for which the manipulator’s slender body can access an internal space via a tight aperture and navigate through a narrow environment. Applications include the inspection of vessels and engines in the nuclear and aeronautical industries, as well as endoscopic imaging for medical diagnosis. In both cases, the accurate control of the entire manipulator body is of critical importance to the safe operation of the device. - -## Project Description - -Construct a [Simscape™ Multibody™](https://www.mathworks.com/help/physmod/sm/ref/simscape.multibody.kinematicssolver.html?searchHighlight=multibody%20inverse%20kinematics&s_tid=srchtitle) assembly of a snake-like robot and develop an autonomous controller to navigate the robot in a constricted environment by entering it from a small opening. - -Suggested steps: -1. Model the robot as with a serial chain of solid bodies and revolute joints. -2. Solve the inverse kinematics of the end-effector and robot body, cf. [Solve kinematic problems for a Multibody model](https://www.mathworks.com/help/physmod/sm/ref/simscape.multibody.kinematicssolver.html?searchHighlight=multibody%20inverse%20kinematics&s_tid=srchtitle). -3. Adopt an appropriate actuation mechanism for the revolute joints. These robots are commonly designed to use pneumatic actuators, magnetic elements or a combination of cables and springs, cf. Multibody [Assemblies](https://www.mathworks.com/help/physmod/sm/multibody-systems.html). -4. Implement a control algorithm to follow a trajectory using the inverse kinematics models developed in step 2. Learn about trajectory generation algorithms [here](https://www.mathworks.com/help/robotics/coordinate-system-transformations.html). - -Project variations: - -Model your snake-like robot, using CAD software of your choice -- Export it as a URDF -- Import the URDF into Simscape Multibody using the [smimport](https://www.mathworks.com/help/physmod/sm/ref/smimport.html) function -- Alternatively, import the URDF into Gazebo and control it using Simulink with Gazebo co-simulation or [ROS Toolbox®](https://www.mathworks.com/products/ros.html) - -Advanced project work: -- Pick and build the scenario in which the robot will navigate, e.g. pipeline, aircraft engine, reaction vessel. -- Create a model of the environment using occupancy grids ([2D](https://www.mathworks.com/help/robotics/ug/occupancy-grids.html),[3D](https://www.mathworks.com/help/nav/ref/occupancymap3d.html)). -- Add sensors to model distance and inertial sensors, e.g. LiDAR, cameras, and IMUs. -- Integrate approaches for [planning](https://www.mathworks.com/discovery/path-planning.html) and obstacle avoidance. -- Develop searching and mapping algorithms. -- Incorporate optimization-based or reinforcement learning-based control techniques in the motion planning hierarchy using …toolboxes -- Test a perception-based workflow by modelling the inspected [VR in Simulink 3d Animation](https://www.mathworks.com/products/3d-animation.html)or using [UE4 co-simulation](https://www.mathworks.com/help/driving/unreal-engine-scenario-simulation.html). -- Model aerodynamic forces (drag and lift) experienced by the robot body during motion. -- Develop an advanced robot using multiple snake-like components. For example, you could consider each component as finger in a robotic hand or gripper, or as legs in a robotic walker or swimmer. - -## Background Material - -Examples: -- [Create a simple part in Simscape Multibody](https://www.mathworks.com/help/physmod/sm/ug/creating-a-simple-part.html) -- [Solve kinematic problems for a Multibody model](https://www.mathworks.com/help/physmod/sm/ref/simscape.multibody.kinematicssolver.html?searchHighlight=multibody%20inverse%20kinematics&s_tid=srchtitle) -- [Modelling flexible bodies in Simscape Multibody](https://www.mathworks.com/campaigns/offers/model-flexible-bodies.html) - - Suggested readings: -- Hughes, J., Culha, U., Giardina, F., Guenther, F., Rosendo, A., & Iida, F. (2016). Soft manipulators and grippers: A review. Frontiers in Robotics and AI, 3, 69. -- SMH Sadati, SE Naghibi, A Shiva, B Michael, L Renson. (2019) TMTDyn: A matlab package for modeling and control of hybrid rigid–continuum robots based on discretized lumped systems and reduced-order models The International Journal of Robotics Research, 2019 -- S Kim, C Laschi, B Trimmer  Soft robotics: a bioinspired evolution in robotics - Trends in biotechnology, 2013 - -## Impact - -Advance robotics design for hazardous environments inspection and operation in constricted spaces. - -## Expertise Gained - -Robotics, Manipulators, Modeling and Simulation - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/56) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -224 diff --git a/projects/Speech Background Noise Suppression with Deep Learning/README.md b/projects/Speech Background Noise Suppression with Deep Learning/README.md deleted file mode 100644 index 3fec2e06..00000000 --- a/projects/Speech Background Noise Suppression with Deep Learning/README.md +++ /dev/null @@ -1,103 +0,0 @@ -**Project 193:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Speech Background Noise Suppression with Deep Learning

-

Develop a deep learning neural network for audio background noise suppression.

-
- -## Motivation - -Globally, 1.5 billion people live with some degree of hearing loss, according to the [World Health Organization](https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearing-loss). Of those who could benefit with the use of a hearing aid, only 17% use one. Background noise can be very detrimental to such devices because it reduces the intelligibility and quality of speech. Noise suppression is an active research area with multiple applications, beside hearing aids, including smart devices, and online meetings. Moreover, the recent rapid rise of “remote work” has widened the need for efficient and robust noise suppression. - -Noise suppression is a challenging problem. Noise suppression systems must handle a wide variety of types and sources of noise, possibly at low signal-to-noise ratios. Additionally, they must strike a balance between removing unwanted noise and minimizing speech distortion in order maintain speech intelligibility. - -Noise suppression approaches based on artificial intelligence (AI) and, more specifically, deep learning have recently shown promising results. The availability of large, public datasets for clean speech and noise audio signals has enabled researchers and engineers to design and train AI noise suppression models that potentially outperform more traditional signal processing techniques. - -## Project Description - -Work with the [Audio Toolbox™](https://www.mathworks.com/products/audio.html) and [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) products to develop and train a noise suppression deep learning network using MATLAB®. Test and validate your trained network using both subjective and objective methods. - - Suggested steps: - -1. Perform a literature search on speech noise suppression using deep learning. - -2. Download the speech and noise dataset by following the instructions on the [Microsoft DNS challenge repository](https://github.com/microsoft/DNS-Challenge). - -3. Design the deep learning network using Deep Learning Toolbox. There are many network architectures to choose from. The two most common ones are convolutional neural networks, or CNN, ([[5]](#choi), [[6]](#isik)), and recurrent neural networks, or RNN ([[2]](#valin), [[3]](#xiang), [[4]](#nils)). -Your solution may require applying signal processing at the input or output of your network (for example, for signal pre-processing, feature extraction, or -time-frequency transformation). In that case, you can use Audio Toolbox and [Signal Processing Toolbox™](https://www.mathworks.com/products/signal.html) functionalities. - -4. Train the network using the downloaded speech/noise dataset. - -5. Evaluate your noise suppression system on a test set, using both subjective evaluation (score based on hearing) and objective metrics (for example, signal-to-interference ratio). - -Project variations: - -1. Use an end-to-end noise suppression system. Many noise suppression techniques in the literature transform the audio time-series into a time-frequency representation (such as a magnitude spectrogram, or a short-time Fourier transform) before feeding it to the network ([[3]](#xiang), [[4]](#nils), [[5]](#choi), [[6]](#isik)). However, end-to-end noise suppression systems have also been proposed recently, where the audio signal is directly fed to the system ([[7]](#alamdari), [[8]](#rethage)). - -2. Use a self-supervised technique where you train only on noisy data (see [[7]](#alamdari) for an example), instead of training your network using clean/noisy speech pairs. - -Advanced project work: - -1. Run your noise suppressor model in real time: it must take less than the frame stride time to process an input frame of speech otherwise, your system will drop audio samples, degrading the speech quality or rendering it non-intelligible. - -2. Take your noise suppressor beyond the desktop and deploy it on an embedded target, for example a Raspberry Pi™ board. Refer to the “Background material” section for an example. - -3. Extend the scope of your network to perform speech dereverberation as well. The available dataset contains measured and synthetic room impulse responses that you can use to generate reverberant speech training data. - -4. Consider personalized noise suppression, where your network is trained to improve the speech quality of a particular speaker. In this scenario, you have access to a few minutes of speech from this speaker, from which you can extract features in the training stage. - - -## Background Material - -1. [Get started with Deep Learning Toolbox](https://www.mathworks.com/help/deeplearning/examples.html?category=getting-started-with-deep-learning-toolbox&exampleproduct=all&s_tid=CRUX_lftnav) for simple examples of designing and training deep learning networks. - -2. [Deep Learning with Time Series, Sequences, and Text](https://www.mathworks.com/help/deeplearning/examples.html?category=deep-learning-with-time-series-sequences-and-text&s_tid=CRUX_topnav) for examples of deep learning applied to time-series data, including audio and speech. - -3. [Denoise Speech Using Deep Learning Networks](https://www.mathworks.com/help/audio/ug/denoise-speech-using-deep-learning-networks.html) for illustration purposes. - -4. [Speech Command Recognition Code Generation on Raspberry Pi](https://www.mathworks.com/help/deeplearning/ug/speech-command-recognition-code-generation-on-raspberry-pi.html). - - -Suggested readings: - -[1] Chandan K.A.Reddy et al, “The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Results”, INTERSPEECH 2020, October 2020, Shanghai, China. - -[2] J.-M. Valin, "A Hybrid DSP/Deep Learning Approach to Real-Time Full-Band Speech Enhancement", International Workshop on Multimedia Signal Processing, 2018. (arXiv:1709.08243) - -[3] Hao Xiang, Su Xiangdong, Horaud Radu, Li Xiaofei, "FullSubNet: A Full-Band and Sub-Band Fusion Model for Real-Time Single-Channel Speech Enhancement", ICASSP 2021. (arXiv:2010.15508) - -[4] Nils L. Westhausen, Bernd T. Meyer, "Dual-Signal Transformation LSTM Network for Real-Time Noise Suppression", INTERSPEECH 2020. (arXiv:2005.07551) - -[5] Hyeong-Seok Choi, Hoon Heo, Jie Hwan Lee, Kyogu Lee, "Phase-aware Single-stage Speech Denoising and Dereverberation with U-Net", INTERSPEECH 2020. (arXiv:2006.00687) - -[6] Umut Isik et Al, "PoCoNet: Better Speech Enhancement with Frequency-Positional Embeddings, Semi-Supervised Conversational Data, and Biased Loss", INTERSPEECH 2020. (arXiv:2008.04470) - -[7] N. Alamdari, A. Azarang, N. Kehtarnavaz, “Improving deep speech denoising by Noisy2Noisy signal mapping”, Applied Acoustics, Volume 172, 2021. (arXiv:1904.12069) - -[8] D. Rethage, J. Pons, X, Serra, “A Wavenet for Speech Denoising", ICASSP 2018. (arXiv:1706.07162) - -## Impact - -Advance hearing aid technology through research in speech enhancement and noise suppression and improve the quality of life of persons with a hearing impairment. - -## Expertise Gained - -Artificial Intelligence, Deep Learning, Neural Networks, Signal Processing - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/24) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[jibrahim](https://github.com/jibrahim80) - -## Project Number - -193 diff --git a/projects/Speech Background Noise Suppression with Deep Learning/student submissions/MATLAB-denoise b/projects/Speech Background Noise Suppression with Deep Learning/student submissions/MATLAB-denoise deleted file mode 160000 index cd98bbfd..00000000 --- a/projects/Speech Background Noise Suppression with Deep Learning/student submissions/MATLAB-denoise +++ /dev/null @@ -1 +0,0 @@ -Subproject commit cd98bbfd014d84b742b596c2857edc16b789cf66 diff --git a/projects/Synthetic Aperture Radar (SAR) Simulator/README.md b/projects/Synthetic Aperture Radar (SAR) Simulator/README.md deleted file mode 100644 index 8c1dc2e7..00000000 --- a/projects/Synthetic Aperture Radar (SAR) Simulator/README.md +++ /dev/null @@ -1,78 +0,0 @@ -**Project 211:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Synthetic Aperture Radar (SAR) Simulator

-

Develop a lightweight Synthetic Aperture Radar (SAR) raw data simulator.

-
- -## Motivation - -Achieving all-weather imaging capabilities is a critical need for remote-sensing, airborne imaging, and monitoring applications. SAR provides a suitable mechanism to attain this weather agnostic capability. SAR imaging allows us to create high spatial resolution imagery from reflectivity signals of objects and environments obtained through radar sensing. -The cost of validating such an image formation algorithm on-board a moving platform is considerably high and to reduce this, an offline/lab-based algorithm validation is needed. Hence, developing simulators that can generate Radar in-phase quadrature (IQ) data for a scenario will significantly help validate the algorithms. There can be different aspects of object resolution refinement that can be evaluated in the simulator. The focus here is on refining the radar cross-section (RCS) and/or impact of various waveforms and their characteristics on the generated image. - -## Project Description - -The following steps (shown in order) provide guidelines to help you develop a high-quality SAR data simulator and imaging tool. The end goal would be to support a workflow where you define SAR system parameters, setup the environment, simulate IQ data based on the system definition, and validate it using an image formation algorithm of your choice. - -Suggested Steps: -1. Conduct a literature survey of SAR systems, SAR data simulation, and image formation algorithms [1-3]. -2. Use [Phased Array System Toolbox™](https://www.mathworks.com/help/phased/index.html) and [Radar Toolbox](https://www.mathworks.com/help/radar/index.html) to build the simulator. -3. Use MATLAB® for development of a simulation mechanism to generate IQ data for a single point scatterer under [Strip-map SAR mode](https://www.mathworks.com/help/radar/ug/stripmap-synthetic-aperture-radar-sar-image-formation.html;jsessionid=91731983519938346125a98e50b0 ). -4. Incorporate the effects of atmospheric loss, Earth’s curvature, and antenna grazing angle in the simulator. (https://www.mathworks.com/help/radar/radar-system-analysis.html) -5. Process the raw data from the simulator and reconstruct the point target using a SAR image formation algorithm of your choice [5-6]. -6. The results can be verified using the desired versus achieved image resolution for point targets using metrics like PSLR (Peak Sidelobe Ratio) and ISLR (Integrated Sidelobe Ratio). - -Project variations: -1. RCS refinement: - 1. Incremental refinement of the simulator by varying [RCS models](https://www.mathworks.com/help/radar/ug/modeling-target-radar-cross-section.html) and characteristics, adding targets of different variety (distributed, extended), effect of ground clutter and analysis of the generated image. - 2. Analyze the impact using metrics in step 6 above or other metrics of your choice. -2. Waveform refinement: - 1. Perform Literature survey of waveforms for SAR systems, SAR data simulation and image formation algorithms. [1-4] - 2. Vary the type of [waveforms and the waveform characteristics] -(https://www.mathworks.com/help/phased/waveform-design-and-analysis.html) including bandwidth, operating frequency, pulse width, PRF etc. and analyze the impact of this change on the target characteristics and efficiency in mitigating clutter. - - -## Background Material - -Examples: -- [Stepped FM based SAR system design in Simulink](https://www.mathworks.com/help/radar/ug/synthetic-aperture-radar-system-simulation-and-image-formation.html) -- [Stripmap SAR Image Formation](https://www.mathworks.com/help/phased/ug/stripmap-synthetic-aperture-radar-image-formation.html) -- [Squinted Spotlight Image Formation](https://www.mathworks.com/help/phased/ug/squinted-spotlight-synthetic-aperture-radar-sar-image-formation.html) - -Suggested Readings - -[1] A. Moreira, P. Prats-Iraola, M. Younis, G. Krieger, I. Hajnsek and K. P. Papathanassiou, "A tutorial on synthetic aperture radar," in IEEE Geoscience and Remote Sensing Magazine, vol. 1, no. 1, pp. 6-43, March 2013, doi: 10.1109/MGRS.2013.2248301. - -[2] Balmer, R. Principles of Synthetic Aperture Radar. Surveys in Geophysics 21, 147–157 (2000). https://doi.org/10.1023/A:1006790026612. - -[3] Yee Kit Chan and Voon Koo, "An Introduction to Synthetic Aperture Radar (SAR)," Progress In Electromagnetics Research B, Vol. 2, 27-60, 2008. - -[4] J. Yang, X. Huang, T. Jin, J. Thompson and Z. Zhou, "Synthetic Aperture Radar Imaging Using Stepped Frequency Waveform," in IEEE Transactions on Geoscience and Remote Sensing, vol. 50, no. 5, pp. 2026-2036, May 2012, doi: 10.1109/TGRS.2011.2170176. - -[5] D. C. Munson and R. L. Visentin, "A signal processing view of strip-mapping synthetic aperture radar," in IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 37, no. 12, pp. 2131-2147, Dec. 1989, doi: 10.1109/29.45556. - -[6] John R. Hupton, John A. Saghri, "Three-dimensional target modeling with synthetic aperture radar," Proc. SPIE 7798, Applications of Digital Image Processing XXXIII, 77980P (7 September 2010) - - -## Impact - -Accelerate design of SAR imaging systems and reduce time and cost for their development for aerial and terrestrial applications - -## Expertise Gained - -Autonomous Vehicles, Automotive, AUV, Image Processing, Signal Processing, Radar Processing - - -## Project Difficulty - -Master's, Doctoral, Bachelor - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/43) to ask/answer questions, comment, or share your ideas for solutions for this project. - - -## Project Number - -211 diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/README.md b/projects/Techno-Economic Assessment of Green Hydrogen Production/README.md deleted file mode 100644 index 6e2bbe2b..00000000 --- a/projects/Techno-Economic Assessment of Green Hydrogen Production/README.md +++ /dev/null @@ -1,60 +0,0 @@ -**Project 236:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Techno-Economic Assessment of Green Hydrogen Production

-

Perform early-stage economic feasibility of an energy project to determine project viability.

-
- -## Motivation - -The main sources of global greenhouse gas emissions come from transportation, electricity generation and industrial processes. Green hydrogen is produced from the electrolysis of water, powered by renewable energy sources, and is being considered as a pathway to decarbonize energy dense sectors such as aviation, and industrial manufacturing. Most hydrogen produced today is from reforming of natural gas, but this method releases greenhouse gas into the atmosphere. A key challenge is how to produce green hydrogen economically and in a repeatable way, given the variable nature of renewable energy. - -## Project Description - -Develop a framework to perform techno-economic assessment of a green hydrogen production system. - -Suggested steps: -1. Become familiar with the included green hydrogen production model and use this Simscape model as the basis of your project. -2. Add economic signals to the Simscape model, including capital cost, operational cost, and cost of grid electricity. -3. Download and import electricity price data from appropriate independent system operators (for example [New England ISO Express](https://www.iso-ne.com/markets-operations/iso-express) ). -4. Change the location of the system, by importing solar irradiance data (for example [North America National Solar Radiation Database](https://nsrdb.nrel.gov/data-sets/how-to-access-data)) from different locations included in the StationData data structure. -5. Create a MATLAB script that will automate the loading of data, run the simulation at different locations, and calculate economic cost. Use parallel computing if possible. -6. Document your approach and findings, and make recommendations on how to effectively assess techno-economic feasibility of energy projects. Document any limitations in your approach. - -Project variations: - -1. Modify the operation of the energy storage system, to reduce cost of grid electricity at a given location. -2. Size the energy storage system at each location so no grid electricity is needed. -3. Extend economic analysis by adding balance-of-plant considerations, such as cost of water supply. - - -## Background Material - -- [Model included in this repository](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/tree/main/projects/Techno-Economic%20Assessment%20of%20Green%20Hydrogen%20Production/techno_economic_green_hydrogen_model) -- Video: -[Techno-Economic Analysis of a Solar-Powered Green Hydrogen Production System](https://www.youtube.com/watch?v=cpttz8Q7jww) -- Example -[Techno-Economic Analysis of Hybrid Renewable Energy System with PSO - File Exchange - MATLAB Central (mathworks.com)](https://www.mathworks.com/matlabcentral/fileexchange/54205-techno-economic-analysis-of-hybrid-renewable-energy-system-with-pso) - - -## Impact - -Connect economic aspect to technical design. - -## Expertise Gained - -Sustainability and Renewable Energy, Modeling and Simulation, Electrification - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/72) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -236 diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen.png b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen.png deleted file mode 100644 index 67c2c4a1..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen.png and /dev/null differ diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen_production_ROM_8760.slx b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen_production_ROM_8760.slx deleted file mode 100644 index e40526ab..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/green_hydrogen_production_ROM_8760.slx and /dev/null differ diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/grid.png b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/grid.png deleted file mode 100644 index 7908e8f1..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/grid.png and /dev/null differ diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/initData.mat b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/initData.mat deleted file mode 100644 index 4a0345f8..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/initData.mat and /dev/null differ diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/solar.png b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/solar.png deleted file mode 100644 index 4f45bb7a..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/solar.png and /dev/null differ diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/startup.m b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/startup.m deleted file mode 100644 index 73d8c7ed..00000000 --- a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/startup.m +++ /dev/null @@ -1,8 +0,0 @@ - -% load data - -load initData - -% set irradiance - -irradiance = StationData(1).Irradiance; diff --git a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/storage.png b/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/storage.png deleted file mode 100644 index e8b3bab8..00000000 Binary files a/projects/Techno-Economic Assessment of Green Hydrogen Production/techno_economic_green_hydrogen_model/storage.png and /dev/null differ diff --git a/projects/Testing Realtime Robustness of ROS in Autonomous Driving/README.md b/projects/Testing Realtime Robustness of ROS in Autonomous Driving/README.md deleted file mode 100644 index feee6cc0..00000000 --- a/projects/Testing Realtime Robustness of ROS in Autonomous Driving/README.md +++ /dev/null @@ -1,79 +0,0 @@ -**Project 220:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Testing Realtime Robustness of ROS in Autonomous Driving

-

Develop a realtime collision avoidance system using ROS2 that will execute a safe vehicle response.

-
- -## Motivation - -As automated driving systems become more complex, there is a greater need to leverage middleware such as the [Robotic Operating System (ROS)](https://www.ros.org/). However, the use of such middleware requires trade-offs with respect to aspects such as architecture, execution speed, and control over execution. When designing complicated automated driving systems for production, it is important to be aware of such trade-offs and related issues to produce a vehicle that meets performance and safety standards. - -## Project Description - -Using the principles of Model-Based Design, leveraging the toolchain from the MathWorks including [Simulink®](https://www.mathworks.com/products/simulink.html), [System Composer™](https://www.mathworks.com/products/system-composer.html), and [ROS Toolbox](https://www.mathworks.com/products/ros.html), develop a realtime collision avoidance robot using ROS2 that will execute a safe vehicle response for a vehicle travelling at, for example 50 mph, to prevent a frontal collision. There are three major parts of this project: -1. System design and analysis -2. System prototyping using low-cost hardware -3. System testing and analysis of performance - -Suggested high-level steps: -1. Develop system- and component-level requirements. The system only needs to consist of two applications where ROS sits between them. -2. Develop architecture using System Composer™. -3. Develop component algorithms such as image acquisition/computer vision, navigation, and control of vehicle, using Simulink as the common algorithm platform. -4. Integrate components using ROS Toolbox. -5. Use simulation to predict the performance of the system and perform tradeoff studies of the design or architecture. Performance metrics may include lag, responses to data drop-outs, and irregular data transfer rates. [Automated Driving Toolbox™](https://www.mathworks.com/products/automated-driving.html) coupled with [Unreal Engine](https://www.unrealengine.com/en-US/) can also be used to perform simulations in a virtual environment. -6. Deploy algorithms to hardware using either [hardware support packages](https://www.mathworks.com/hardware-support/home.html) or embedded controller. Example hardware platforms could be: Jetbot, Raspberry Pi -7. Test system, closely monitoring performance metrics such as data throughput, response time, and tracking accuracy. Also note any issues with integrating the system within the ROS architecture. Compare results with ideal performance of a system with no middleware latencies. - -Project variations: -- Study effects of architecture, such as where sensor fusion is performed (in vehicle management computer or nearer to sensors) -- Implement cloud processing for vision algorithms (mainly to offload processing from low-cost hardware) -- Use more advanced sensors such as LIDAR or a combination of sensors + sensor fusion - -Advanced project work (optional): -- Add multiple computation nodes (ex. One for image processing, another for vehicle controls) -- Deploy on a full-scale vehicle -- Perform optimization to determine best interface specifications (speed, message size, etc.) - -## Background Material - -- [ROS Toolbox](https://www.mathworks.com/products/ros.html) -- [Get Started with ROS Toolbox](https://www.mathworks.com/help/ros/getting-started-with-ros-toolbox.html) -- [System Composer](https://www.mathworks.com/products/system-composer.html) -- [Simulink](https://www.mathworks.com/products/simulink.html) -- [Automated Driving Toolbox](https://www.mathworks.com/products/automated-driving.html) -- [Automated Parking Valet](https://www.mathworks.com/help/driving/ug/automated-parking-valet.html) -- [Deep Learning with Raspberry Pi and MATLAB](https://www.mathworks.com/company/events/webinars/upcoming/deep-learning-with-raspberry-pi-and-matlab-3251374.html) -- [Deploying ROS Node on Raspberry Pi](https://youtu.be/6IWImhKpihA) -- [Deep Learning with NVIDIA Jetson and ROS](https://youtu.be/0FPPBGAKw8k) -- [Simulink Onramp](https://www.mathworks.com/learn/tutorials/simulink-onramp.html) -- [MATLAB Onramp](https://www.mathworks.com/learn/tutorials/matlab-onramp.html) - -Suggested readings: -- [ROS Robotics By Example, 2e](https://www.mathworks.com/academia/books/ros-robotics-by-example-fairchild.html?s_tid=srchtitle) -- [Intelligent Control of Robotic Systems](https://www.mathworks.com/academia/books/intelligent-control-of-robotic-systems-behera.html?s_tid=srchtitle) - - - -## Impact - -Contribute to improving access and safety of transportation through robust automated driving systems. - - -## Expertise Gained - -Autonomous Vehicles, Robotics, Automotive, Image Processing, Modeling and Simulation, Sensor Fusion and Tracking, Low-Cost Hardware - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/52) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -220 diff --git a/projects/Top Quark Detection with Deep Learning and Big Data/README.md b/projects/Top Quark Detection with Deep Learning and Big Data/README.md deleted file mode 100644 index 0e61d0b3..00000000 --- a/projects/Top Quark Detection with Deep Learning and Big Data/README.md +++ /dev/null @@ -1,80 +0,0 @@ -**Project 238:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Top Quark Detection with Deep Learning and Big Data

-

Develop a predictive classifier model able to discriminate jets produced by top quark decays from the background jets

-
- -## Motivation - -Ability to detect and retain only interesting particle jets (that usually are rare events) and throw away all the boring jets (that are produced within well understood standard model) is crucial in searching for new physics (rare events), especially when high luminosity upgrade on Large Hadron Collider (LHC) will generate petabytes of data daily. - - Modern deep learning algorithms trained on jet images out-perform standard physically-motivated feature driven approaches to jet tagging developed by physicists previously. - -After the high-luminosity upgrade of LHC collider, most efficient Deep Learning models will be deployed on CERN triggers that will automatically filter (in real-time) background jets and retain only desired jets. This will intelligently reduce the amount of data to be stored for discovering new fundamental physics by more than 80%, avoiding deluge of data. - - -## Project Description - -Aim of the project is to develop a predictive classifier model, based on deep convolutional neural network trained on publicly available big data, that can discriminate efficiently jets produced by top quark decays (signal) from the background jets (noise). - -Work with the [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) products to develop and train, validate, and test a deep learning classification network using MATLAB® and publicly available data on [CERN’s Zenodo database](https://zenodo.org/record/2603256#.Y20xysvMLmE) - -Suggested steps: - -1. Become familiar with the MATLAB based deep learning examples listed in the Background Material section below. -2. Download Live Script on: [Deep Learning for Real-Time Top Quark Jet Tagging](https://www.mathworks.com/matlabcentral/fileexchange/105635-deep-learning-for-real-time-top-quark-jet-tagging?s_tid=srchtitle) -3. Download particle jets open datasets from: Kasieczka, Gregor, Plehn, Tilman, Thompson, Jennifer, & Russel, Michael. (2019). Top Quark Tagging Reference Dataset (v0 (2018_03_27)) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.2603256 -4. Import datasets as pandas dataframes and save as parquet files. -Run Python directly from MATLAB with the following commands: -``` -pyrun("import pandas as pd") -pyrun("df=pd.read_hdf('train.h5','table')") -pyrun("df.to_parquet('jets.parquet.gzip', compression='gzip')") -``` -5. Use [parquet datastore]( https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.parquetdatastore.html) and [Tall Array](https://www.mathworks.com/help/matlab/tall-arrays.html), to manage and preprocess the particle jets data and arrange it into 2D grayscale images. -6. Build a deep convolutional neural network, analogous to the one in: [Deep Learning for real-Time Top Quark Jet Tagging](https://www.mathworks.com/matlabcentral/fileexchange/105635-deep-learning-for-real-time-top-quark-jet-tagging?s_tid=srchtitle) using [Deep Network designer app](https://www.mathworks.com/help/deeplearning/ref/deepnetworkdesigner-app.html) and train network using training datasets. -7. Check accuracy of the network on test datasets. Use [imageDatastores](https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.imagedatastore.html), where label sources come from folder names. - -Project variations: - -1. Instead of a simple 3-layer CNN, Adopt more complex deep learning networks such as resnet18. Call blank architecture of resnet18 and train its weights. GPU can be handy when training more complicated deep networks, especially when dealing with big data. -2. Instead of gray-scale images use additional variables of the jets and encode them into the color of the image. In practice researchers use up to 7 or 8 colors. - -Advanced project work: - -Run your model in real time. Using [Deep Learning HDL Toolbox™](https://www.mathworks.com/products/deep-learning-hdl.html) generate HDL code for deploying model on FPGA, following steps described in: [Deep Learning for Real-Time Top Quark Jet Tagging](https://www.mathworks.com/matlabcentral/fileexchange/105635-deep-learning-for-real-time-top-quark-jet-tagging?s_tid=srchtitle) - - -## Background Material - -For Deep Learning: -- [Get started with Deep Learning Toolbox](https://www.mathworks.com/help/deeplearning/examples.html?category=getting-started-with-deep-learning-toolbox&exampleproduct=all&s_tid=CRUX_lftnav) for simple examples of designing and training deep learning networks. -- [Deep Learning for Top Quark Jet Tagging, without using Big Data](https://www.mathworks.com/matlabcentral/fileexchange/105635-deep-learning-for-real-time-top-quark-jet-tagging?s_tid=srchtitle) - -For Big Data: -- [Datastore for Parquet files](https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.parquetdatastore.html) -- [Tall Array for working with Big Data](https://www.mathworks.com/help/matlab/ref/tall.tall.html) - - -## Impact - -Reduce the interference of background jets and help the discovery of new fundamental physics - -## Expertise Gained - -Artificial Intelligence, Big Data, Deep Learning, Physics - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/74) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -238 diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/README.md b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/README.md deleted file mode 100644 index e48acbcf..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/README.md +++ /dev/null @@ -1,88 +0,0 @@ -**Project 222:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Traffic Data Analysis for Modeling and Prediction of Traffic Scenarios

-

Analyze real-world traffic data to understand, model, and predict human driving trajectories.

-
- -## Motivation - -The past few years, the world has been overwhelmed with large quantities of data from ubiquitous sources with the hope that AI can one day leverage them to evolve. With the autonomous driving industry booming, an increasing number of vehicle trajectories datasets are being recorded to help engineers and researchers improve autonomy algorithms and bring traffic simulation tools to the next level of realism. Therefore, the need arises to establish data processing and analysis methodologies using machine learning techniques that would efficiently leverage the growing traffic data. - -## Project Description - -Work with [Deep Learning Toolbox™](https://www.mathworks.com/products/deep-learning.html) and [Statistics and Machine Learning Toolbox™](https://www.mathworks.com/products/statistics.html) to process and analyze real-world vehicle trajectories data, such as the [Next Generation Simulation (NGSIM) Vehicle Trajectories](https://data.transportation.gov/Automobiles/Next-Generation-Simulation-NGSIM-Vehicle-Trajector/8ect-6jqj) data provided by the U.S. Department of Transportation. - -Objective: - -Enable decision making in autonomous driving by predicting future vehicle trajectories using deep learning and machine learning algorithms, for example, LSTM neural networks in [2] and [3], and Bayes and decision-tree classifiers in [4]. - -Advanced objective: - -Classify the trajectories based on intrinsic attributes (e.g., human driving styles – tired, in a hurry, distracted, aggressive, etc., driving modes – entering the driveway, exiting the driveway, passing a car, trailing a car, changing lanes, etc.), using unsupervised learning. - -Suggested steps: -1. Perform literature research prior to starting the work. Suggested readings are provided in the Background Material below. -2. Select the dataset(s) you will be using for your analysis. Depending on your analysis objective, consider parameters such as road topology, number of provided vehicle trajectories, measured quantities (position, velocity, acceleration, etc.), any available labels (e.g., type of vehicles, type of maneuver, driving profile, etc.) and more. -3. Download the vehicle trajectories datasets. Some options are: - - [Next Generation Simulation (NGSIM) Vehicle Trajectories](https://data.transportation.gov/Automobiles/Next-Generation-Simulation-NGSIM-Vehicle-Trajector/8ect-6jqj) provided by the U.S. Department of Transportation - - [Dataset of Annotated Car Trajectories (DACT)](https://figshare.com/articles/dataset/DACT_Dataset_of_Annotated_Car_Trajectories/5005289) -4. Preprocess and visualize the data using MATLAB. The live script [preprocess_visualize_NGSIM_US101data.mlx](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/blob/main/projects/Traffic%20Data%20Analysis%20for%20Modelling%20and%20Prediction%20of%20Traffic%20Scenarios/preprocess_visualize_NGSIM_US101data.mlx) provided in the supporting material preprocesses the NGSIM data captured on US101 highway and visualizes them using Mapping Toolbox. You are encouraged to use this script as a starting point for other datasets. -5. [Optional] Visualize and replay the recorded trajectories in the Unreal Engine using [RoadRunner](https://www.mathworks.com/products/roadrunner.html) and Simulink. -6. Select the algorithmic solution that best suits your analysis objective, e.g., Bayesian Networks, Long Short-term Memory (LSTM) networks, Clustering (K-means, Hierarchical, etc.), Gaussian mixture models, Support Vector Machines (SVM), etc. -7. Separate the data in testing and training datasets. -8. Setup the problem formulation: determine the inputs, outputs, required intermediate data that are not directly available in the original datasets, such as information about the preceding cars, etc. -9. Use MATLAB and Simulink to design and/or implement an algorithmic solution that uses the data to achieve your analysis objective. You are encouraged to get inspiration from the provided readings. -10. Test the approach on the testing datasets and report efficiency. -11. Document the results, challenges, and potential future work. - - -## Background Material - -Deep Learning examples: -- [Time Series Forecasting Using Deep Learning](https://www.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html) -- [Long Short-Term Memory (LSTM) discovery page](https://www.mathworks.com/discovery/lstm.html) -- [Moving Towards Automating Model Selection Using Bayesian Optimization](https://www.mathworks.com/help/stats/towards-automating-model-selection.html) -- [Guidance for choosing the appropriate clustering method](https://www.mathworks.com/help/stats/choose-cluster-analysis-method.html) -- [Discover Gene Expression profiles using k-Means Clustering](https://www.mathworks.com/help/bioinfo/ug/gene-expression-profile-analysis.html) - -Suggested readings: - -[1] He, Zhengbing. "Research based on high-fidelity NGSIM vehicle trajectory datasets: A review." Research Gate (2017): 1-33. - -[2] N. Deo and M. M. Trivedi, "Convolutional Social Pooling for Vehicle Trajectory Prediction," 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Salt Lake City, UT, USA, 2018, pp. 1549-15498, doi: 10.1109/CVPRW.2018.00196. - -[3] F. Altché and A. de La Fortelle, "An LSTM network for highway trajectory prediction," 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC), Yokohama, Japan, 2017, pp. 353-359, doi: 10.1109/ITSC.2017.8317913. - -[4] Y. Hou, P. Edara and C. Sun, "Modeling Mandatory Lane Changing Using Bayes Classifier and Decision Trees," in IEEE Transactions on Intelligent Transportation Systems, vol. 15, no. 2, pp. 647-655, April 2014, doi: 10.1109/TITS.2013.2285337. - -[5] Moosavi, Sobhan, Behrooz Omidvar-Tehrani, and Rajiv Ramnath. “Trajectory Annotation by Discovering Driving Patterns.” Proceedings of the 3rd ACM SIGSPATIAL Workshop on Smart Cities and Urban Analytics. ACM, 2017. - -[6] Martinelli, F., Mercaldo, F., Nardone, V., Orlando, A., & Santone, A. (2018). Cluster analysis for driver aggressiveness identification. In ICISSP (pp. 562-569). - -[7] Warren, J., Lipkowitz, J., & Sokolov, V. (2019). Clusters of driving behavior from observational smartphone data. IEEE Intelligent Transportation Systems Magazine, 11(3), 171-180. - -[8] Yan, F., Liu, M., Ding, C., Wang, Y., & Yan, L. (2019). Driving style recognition based on electroencephalography data from a simulated driving experiment. Frontiers in psychology, 10, 1254. - - -## Impact - -Contribute to autonomous driving technologies and intelligent transportation research. - -## Expertise Gained - -Big Data, Autonomous Vehicles, Support Vector Machines, Machine Learning, Deep Learning, Automotive - - -## Project Difficulty - -Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/54) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -222 diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.m b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.m deleted file mode 100644 index 2d2053d9..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.m +++ /dev/null @@ -1,16 +0,0 @@ -function data = add_time_step_column(data, delta_time) -arguments - data table - delta_time double = 100 %millisec -end - -% Generate "Time_Step" column -ts = unique(data.Global_Time); - -%dts = ts(2:end) - ts(1:end-1); -%assert(length(unique(dts))==1 && dts(1) == 100); % assert delta glbal time values is constant (100ms) -% ^ This assertion holds only within a single time-block dataset. - -min_t = min(ts); -data.Time_Step = (data.Global_Time - min_t)/delta_time; -end \ No newline at end of file diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.mlx b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.mlx deleted file mode 100644 index 71d47ea9..00000000 Binary files a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/add_time_step_column.mlx and /dev/null differ diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/cell2mat_seqs.mlx b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/cell2mat_seqs.mlx deleted file mode 100644 index a831ae73..00000000 Binary files a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/cell2mat_seqs.mlx and /dev/null differ diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/clip_gradient.m b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/clip_gradient.m deleted file mode 100644 index 5cb14f43..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/clip_gradient.m +++ /dev/null @@ -1,8 +0,0 @@ -function g = clip_gradient(g, thresh) - -g_norm = norm(extractdata(g)); -if g_norm > thresh - g = (thresh/g_norm).*g; -end - -end \ No newline at end of file diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/create_one_step_regression_XY.m b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/create_one_step_regression_XY.m deleted file mode 100644 index ad3030ff..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/create_one_step_regression_XY.m +++ /dev/null @@ -1,30 +0,0 @@ -function [X,Y] = create_one_step_regression_XY(data, in_vars, out_vars) - % ngsim_data (Table): a table of ego states (eg. speed, acceleration, space_hdwy) sorted by vehicle_id and time - % in_vars (cell of strings or string array): name of input variables - % out_vars (same as in_vars(: output/target variable names - ego_ids = unique(data.Vehicle_ID); - n_cars = length(ego_ids); - X = cell(1, n_cars); - Y = cell(1, n_cars); - for k = 1:length(ego_ids) - - ego_id = ego_ids(k); - is_ego = (data.Vehicle_ID == ego_id); - data_ego = data(is_ego, :); %table - data_ego_mtx = data_ego{1:end-1, in_vars}'; - target_mtx = data_ego{2:end, out_vars}'; %Q: end symbol / exit symbol - - X{k} = data_ego_mtx; - Y{k} = target_mtx; - % size(data_ego_mtx); %debug - % size(target_mtx); - - % if ego_id > 10 - % break - % end - % if mod(ego_id, 300) == 0 - % disp(ego_id) - % end - - end -end \ No newline at end of file diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/get_train_ind.m b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/get_train_ind.m deleted file mode 100644 index 260e92d6..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/get_train_ind.m +++ /dev/null @@ -1,18 +0,0 @@ -function [isTrain] = get_train_ind(n, p_train, seed) - % n (int): total number of data points - % p_train (float): in range [0,1]; percentage of training points; the - % rest will become validation points - % seed (int or NaN by default): random number genertor's seed - arguments - n (1,1) {mustBeNumeric} - p_train (1,1) {mustBeNumeric} - seed = NaN - end - - if ~isnan(seed) - rng(seed); - end - isTrain = false(1,n); % create logical index vector - isTrain(1:round(p_train*n)) = true; - isTrain = isTrain(randperm(n)); -end diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/time_integrate.m b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/time_integrate.m deleted file mode 100644 index 684bc63e..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/Utils/time_integrate.m +++ /dev/null @@ -1,10 +0,0 @@ -function xs = time_integrate(init_x, delta_time, dxs) - % time integration of 1D position given speeds - xs = [init_x, zeros(1, numel(dxs))]; - for k = 1:numel(dxs) - dx = dxs(k); - xs(k+1) = xs(k) + delta_time * dx; - end -end - - diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/preprocess_visualize_NGSIM_US101data.mlx b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/preprocess_visualize_NGSIM_US101data.mlx deleted file mode 100644 index aae5df5e..00000000 Binary files a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/preprocess_visualize_NGSIM_US101data.mlx and /dev/null differ diff --git a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/student submissions/Project222 b/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/student submissions/Project222 deleted file mode 160000 index cfa28b80..00000000 --- a/projects/Traffic Data Analysis for Modelling and Prediction of Traffic Scenarios/student submissions/Project222 +++ /dev/null @@ -1 +0,0 @@ -Subproject commit cfa28b80b4f943199f9ba6acf13280313d8a90b6 diff --git a/projects/Traffic Light Negotiation and Perception-Based Detection/README.md b/projects/Traffic Light Negotiation and Perception-Based Detection/README.md deleted file mode 100644 index 7c92eb0e..00000000 --- a/projects/Traffic Light Negotiation and Perception-Based Detection/README.md +++ /dev/null @@ -1,72 +0,0 @@ -**Project 223:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Traffic Light Negotiation and Perception-Based Detection

-

Detect traffic lights and perform traffic light negotiation at an intersection in Unreal environment.

-
- -## Motivation - -Navigating an intersection with traffic lights is a challenging but safety critical task for a self-driving car. Using perception to identify traffic light positions and states despite lighting conditions and occlusions is an interesting problem that a 3D simulation environment can safely and effectively provide a solution for. Automated Driving Toolbox™ provides a 3D simulation environment powered by Unreal Engine® from Epic Games® that can be used to visualize traffic lights and the motion of a vehicle in a 3D scene. - -## Project Description - -Use MATLAB® and Simulink® to load a pre-built Unreal scene, detect and identify the state of the traffic light nearest to the ego-vehicle, design a Stateflow® model to control traffic lights in the scene, and control the reaction of the ego-vehicle in accordance to the traffic lights and surrounding vehicles. - -The position and color state of the traffic light at an intersection can be obtained by using a combination of sensors and perception algorithms. Perception can be used to identify surrounding vehicles which can also be used to inform decisions. Identify distance between the traffic light nearest to the ego-vehicle and the ego-vehicle in a pre-built Unreal scene intersection. Identify the color of the traffic light using camera output and perception. Control the change of state of the traffic light using Stateflow. The ego-vehicle should react to the traffic light information. Build your own scenes with the following suggested requirements and perform a quantitative analysis of your algorithm: -a. Traffic lights obstructed by trees -b. Various types of traffic lights like hanging, on a pole, multiple traffic lights in different orientations at an intersection -c. Scenes with different weather conditions -Work with [Automated Driving Toolbox™](https://www.mathworks.com/help/driving/index.html), [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html), [Lidar Toolbox™]( https://www.mathworks.com/help/lidar/getstarted.html), and [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html) for this project. - -Suggested steps: -1. Become familiar with MATLAB and Simulink based sensor fusion using the [examples in Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/help/fusion/examples.html), send scenario information from Simulink to Unreal simulation environment using the traffic light negotiation with Unreal visualization example, scenario design using Automated Driving Toolbox, and vehicle dynamics modelling examples listed in Background Material section below. -2. Load and visualize pre-built [US city block](https://www.mathworks.com/help/driving/ref/uscityblock.html) scene in Unreal and Simulink. This scene already contains traffic lights and intersections, along with ego and other vehicles. -3. Develop an algorithm in Stateflow to three traffic light states (red, yellow, green) for all traffic lights in the scenario and change their colors accordingly. -4. Add sensors to the ego vehicle. -5. Use the camera output and perception to detect the traffic light color state. -6. Fuse the sensor outputs to obtain the distance between the ego vehicle and the nearest traffic light at that timestep, and poses of other vehicles nearest to/in front of the ego vehicle. -7. Program the ego-vehicle to react to the nearest traffic light pose and state ie stop at a certain distance from the light on red, go on green. The ego should also react to other vehicles in the scenario ie the ego should stop behind a lead vehicle on red. -8. Perform a quantitative analysis of algorithm performance on various scenarios (some requirements are suggested in the project description) and collate the results. Refer to the background material on how to customize Unreal scenes. - -Project variations: -1. Design a controller for the ego-vehicle that takes velocity and path information from the planner. -2. Add decision logic for a planner that changes ego vehicle state based on the color of the traffic light and positions of other vehicles in the scenario. Use the planner and controller in the ‘Customize Unreal Engine Scenes for Automated Driving’ example model as a starting point. -3. Try adding multiple types of cameras (like a fisheye) and other sensors (like a LIDAR and radar) specifically for traffic light detection. - -Advanced project work: -1. Try creating your own camera sensor to model a specific camera. -2. Try vehicle dynamics models of various fidelity to help achieve - -## Background Material - -- [Customize Unreal Engine Scenes for Automated Driving](https://www.mathworks.com/help/driving/ug/customize-3d-scenes-for-automated-driving.html) -- [Traffic Light Negotiation with Unreal Engine Visualization](https://www.mathworks.com/help/mpc/ug/traffic-light-negotiation-with-unreal-engine-visualization.html) -- [US city block 3D environment](https://www.mathworks.com/help/driving/ref/uscityblock.html) -- [Adaptive Cruise Control with Sensor Fusion](https://www.mathworks.com/help/driving/ug/adaptive-cruise-control-with-sensor-fusion.html) -- [Automated Driving Using Model Predictive Control](https://www.mathworks.com/help/mpc/ug/automated-driving-using-model-predictive-control.html) -- [Obstacle Avoidance Using Adaptive Model Predictive Control](https://www.mathworks.com/help/mpc/ug/obstacle-avoidance-using-adaptive-model-predictive-control.html) -- [Visualize Sensor Data from Unreal Engine Simulation Environment](https://www.mathworks.com/help/driving/ug/visualize-3d-simulation-sensor-coverages-and-detections.html) - - -## Impact - -Contribute to the advancement of autonomous vehicles traffic coordination in intersections through simulation. - -## Expertise Gained - -Autonomous Vehicles, Computer Vision, Automotive, Control, Deep Learning, Image Processing, Modeling and Simulation, Sensor Fusion and Tracking - - -## Project Difficulty - -Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/55) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -223 diff --git a/projects/Underwater Drone Hide and Seek/README.md b/projects/Underwater Drone Hide and Seek/README.md deleted file mode 100644 index 0b281e62..00000000 --- a/projects/Underwater Drone Hide and Seek/README.md +++ /dev/null @@ -1,59 +0,0 @@ -**Project 27:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Underwater Drone Hide and Seek

-

After robots conquered ground, sky and space, they are going deep sea next. Explore the frontier of autonomous underwater vehicles by doing a project on robot collaboration and competition underwater.

-
- -## Motivation - -Artificial intelligence and sensor technologies have pushed the boundary of how human explore uncharted spaces both above our heads and below our feet. -Underwater drones have been mapping the global seafloor and constructing undersea internet cables for years. -Now they are getting more collaborative; we are seeing increasing amount of robot fleets perform a single task in a coordinated way. -What is the next frontier following collaborative robots? Could it be competitive robots? - -## Project Description - -This project requires a scenario where robots are competing against each other. Design algorithms for underwater drones to compete in a hostile situation, where a first drone is trying to stay stealth and a second drone is actively searching for the first drone. Develop an algorithm to navigate a stealth underwater drone through a radar region (treat radar sweeps as if dynamic obstacles) without being detected. Develop another algorithm to navigate a second underwater drone, which is equipped with a fixed scanning frequency radar, to search for the stealth drone. - -Suggested steps: -1. Draw a virtual underwater environment in your chosen simulation software ([UAV scenario designer](https://www.mathworks.com/help/uav/ug/uav-scenario-tutorial.html) in MATLAB®, [Gazebo](http://gazebosim.org/), or [Unreal Engine](https://www.unrealengine.com/)) -2. Adjust the Simulink® model of underwater drone provided by MathWorks to meet your maneuverability requirement, and make a duplicate so you have two agents -3. Add a sonar sensor to the Simulink model using Sensor Fusion and Tracking Toolbox™ -4. Implement your own control strategy for each vehicle in Simulink (some ideas: either build a state machine using defined “if-then” or [Stateflow](https://www.mathworks.com/products/stateflow.html), or use reinforcement lerning via the [Reinforcement Learning toolbox™ ](https://www.mathworks.com/products/reinforcement-learning.html) to train your hide-and-seek algorithm on the entire Simulink model) -5. Show your hide-and-seek simulation in your virtual environment. - -## Background Material - -- [Modeling and Simulation of an Autonomous Underwater Vehicle](https://www.mathworks.com/videos/modeling-and-simulation-of-an-autonomous-underwater-vehicle-1586937688878.html) -- [Control of an Autonomous Underwater Vehicle](https://www.mathworks.com/videos/matlab-and-simulink-robotics-arena-lqr-control-of-an-autonomous-underwater-vehicle-1543831839770.html) -- [Radar System Design](https://www.mathworks.com/discovery/radar-system-design.html) -- [Multi-Agent Hide and Seek video](https://www.youtube.com/watch?v=kopoLzvh5jY) - -## Impact - -Ocean engineering, underwater constructions, underwater exploration. - - -## Expertise Gained - -Artificial Intelligence, Robotics, AUV, Embedded AI, Machine Learning, Reinforcement Learning, Sensor Fusion and Tracking, SLAM - - -## Project Difficulty - -Bachelor, Master’s level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/5) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[dryouwu](https://github.com/dryouwu) - -## Project Number - -27 - diff --git a/projects/Vibration Detection and Rejection from IMU Data/README.md b/projects/Vibration Detection and Rejection from IMU Data/README.md deleted file mode 100644 index 2e906e11..00000000 --- a/projects/Vibration Detection and Rejection from IMU Data/README.md +++ /dev/null @@ -1,81 +0,0 @@ -**Project 231:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Vibration Detection and Rejection from IMU Data

-

Remove vibration signals from inertial measurement units.

-
- -## Motivation - -Inertial measurement units (IMUs) are used in many navigation applications including UAVs, ground robots, and underwater vehicles. In particular, IMU is key sensor to allow stable flight of micro aerial vehicles (MAVs). However, data from IMU can be affected by high vibration level. Vibrations can come from motors, quadcopter rotors, and the surrounding environment. Accelerometer and gyroscope data can be negatively impacted by vibration of the vehicle, which can in turn degrade the vehicle’s ability to navigate accurately. To build systems that tolerate vibration, designers must have a way of simulating IMUs subject to vibration. In addition, algorithms are needed to detect the vibration signal in the IMU data. - -## Project Description - - This project has two main components: developing a simulation model for an IMU subject to vibration, and compensating for those signals in an IMU. The [Navigation Toolbox™](https://www.mathworks.com/products/navigation.html) and [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html) both contain a high fidelity IMU model – imuSensor. In the first part of the project, you will analyze vibration signals and determine how to drive the imuSensor to accurately simulate an IMU subject to vibration. You can do this using classical techniques like those in [Signal Processing Toolbox™](https://www.mathworks.com/products/signal.html) (see [Vibration Analysis](https://www.mathworks.com/help/signal/vibration-analysis.html?s_tid=CRUX_lftnav)) or with Machine Learning and Deep Learning techniques. In the second part of the project, you will develop techniques to detect and remove vibration from the IMU output. - -Suggested steps: - -Part 1: - -1. Become familiar with the MATLAB IMU simulation model [imuSensor](https://www.mathworks.com/help/nav/ref/imusensor-system-object.html). Simulate IMU signals for a stationary device and for one in motion using [waypointTrajectory](https://www.mathworks.com/help/fusion/ref/waypointtrajectory-system-object.html) (available in the the Navigation Toolbox and Sensor Fusion and Tracking Toolbox, respectively). -2. Become familiar with what IMU signals look like when the device is subject to vibration. You can see a simulation of this in [2] or look at actual IMU datasets in [ds1] and [ds2]. -3. Develop a Vibration Model to be used with the imuSensor as in the diagram below. The Vibration Model should cause the output of the imuSensor to mimic the output of an IMU under vibration. Your Vibration Model can be created with classical signal processing techniques or using a generative AI technique. Can you use this model in conjunction with the waypointTrajectory to simulate a moving device which is subject to vibration? - -| ![vibrationModel ](vibrationModel.png) | -|:--:| -| ***Figure 1**: IMU + vibration model* | - -Part 2: - -4. Develop a Vibration Compensation algorithm for use after the imuSensor as in the diagram below. The Vibration Compensation can be as simple as detecting if vibration is present and setting a Boolean flag, or more a sophisticated algorithm that attempts to filter or remove the vibration signal from the IMU output. You can do this with classical filtering techniques available in Signal Processing Toolbox, Wavelet Toolbox, or with ML/DL approaches. - -| ![vibrationCompensation](VibrationCompensation.png) | -|:--:| -| ***Figure 2**: Vibration compensation* | - -Advanced project work: - -The [MATLAB Support Package for Arduino](https://www.mathworks.com/matlabcentral/fileexchange/47522-matlab-support-package-for-arduino-hardware) allows you to record IMU data in MATLAB. Mount an Arduino with an IMU to an object whose vibration you are modeling (a vehicle, a motor, a bridge) and store that data in MATLAB. Use this real data to compare against your vibration model. - - -## Background Material - -- Navigation Toolbox: [Introduction to Simulating IMU Measurements](https://www.mathworks.com/help/nav/ug/introduction-to-simulating-imu-measurements.html) -- Signal Processing Toolbox: [Vibration Analysis](https://www.mathworks.com/help/signal/vibration-analysis.html?s_tid=CRUX_lftnav) - -Datasets: - -- [ds1] [Kaggle Accelerometer Data Set for “Prediction of Motor Failure Time”](https://www.kaggle.com/datasets/dhinaharp/accelerometer-data-set) - -- [ds2] [Bearing Vibration Data under Time-varying Rotational Speed Conditions](https://data.mendeley.com/datasets/v43hmbwxpm/2) - -Suggested readings: - -- [1] Capriglione, D., et al. "Experimental analysis of IMU under vibration." 16th IMEKO TC10 Conference. 2019. - -- [2] Güner, Ufuk, Hüseyin Canbolat, and Ali Ünlütürk. "Design and implementation of adaptive vibration filter for MEMS based low cost IMU." 2015 9th International Conference on Electrical and Electronics Engineering (ELECO). IEEE, 2015. - -- [3] Zaiss, Curtis. IMU design for high vibration environments with special consideration for vibration rectification. MS thesis. Graduate Studies, 2012. - - -## Impact - - Improve navigation systems by making them robust against vibrations. - -## Expertise Gained - -Drones, Autonomous Vehicles, Robotics, Modeling and Simulation, Sensor Fusion and Tracking, State Estimation, Signal Processing - - -## Project Difficulty - -Doctoral, Bachelor, Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/65) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -231 diff --git a/projects/Vibration Detection and Rejection from IMU Data/VibrationCompensation.png b/projects/Vibration Detection and Rejection from IMU Data/VibrationCompensation.png deleted file mode 100644 index de68e970..00000000 Binary files a/projects/Vibration Detection and Rejection from IMU Data/VibrationCompensation.png and /dev/null differ diff --git a/projects/Vibration Detection and Rejection from IMU Data/vibrationModel.png b/projects/Vibration Detection and Rejection from IMU Data/vibrationModel.png deleted file mode 100644 index 7829e275..00000000 Binary files a/projects/Vibration Detection and Rejection from IMU Data/vibrationModel.png and /dev/null differ diff --git a/projects/Visual - Inertial Odometry for a Minidrone/README.md b/projects/Visual - Inertial Odometry for a Minidrone/README.md deleted file mode 100644 index 3ef098d7..00000000 --- a/projects/Visual - Inertial Odometry for a Minidrone/README.md +++ /dev/null @@ -1,76 +0,0 @@ -**Project 234:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Visual-Inertial Odometry for a Minidrone

-

Design and implement a visual/visual-inertial odometry system using onboard camera for a Minidrone.

-
- -## Motivation - -Using an aerial vehicle to investigate an indoor environment is an emerging field – whether it is for inspection of machinery or for monitoring environmental conditions in a large greenhouse [1]. Indoor aerial vehicles can play a crucial role when the situation does not allow accessibility to humans. - -However, navigating an aerial vehicle in an indoor space is a challenging task – especially since the GPS sensor cannot help with the information about its position. This is where the odometry technique can come in handy - it helps to estimate the change of position by processing the change in sensor values. The technique uses the sensor data like Inertial Measurement Unit (IMU), camera, etc. to estimate the change in position as the vehicle moves in the space. - - -## Project Description - -Use MATLAB and Simulink to design a visual-inertial odometry system for a micro aerial vehicle. Use the downward-facing camera on the Parrot Mambo Minidrone along with the 6-axis IMU data to develop the algorithm to improve the state estimation and replace the currently adopted optical flow simulation algorithm. - -Suggested steps: - - Become familiar with MATLAB and Simulink using resources listed in the Background Material section below. - - Install the [Simulink Support Package for Parrot Minidrones](https://www.mathworks.com/matlabcentral/fileexchange/63318-simulink-support-package-for-parrot-minidrones) from MATLAB->Add-Ons. - - Use the [Parrot Minidrone Competition](https://www.mathworks.com/help/supportpkg/parrot/ref/color-detection-and-landing-parrot-example.html) model as the baseline controller. - - Add various types of [image noise](https://www.mathworks.com/help/images/ref/imnoise.html) to the camera model to simulate real-world image noise. - - Understand the present approximate camera sensor model used to calculate optical flow. - - Create the sensor model to process the information from the simulated monocular downward facing camera image using the quad body’s angular and linear velocity to calculate optical flow. You can use the [already existing blocks](https://www.mathworks.com/help/vision/referencelist.html?type=block&s_tid=CRUX_topnav) from the Computer Vision Toolbox. Use the information for position velocity estimation of the aerial vehicle [2], [3] -- Update the controller and state estimator in the baseline model to account for any possible changes due to the new perception block and sensor simulation model, if needed. - -Next Step – Hardware Deployment: - - If you have the hardware available with you, deploy the controller on the Parrot Mambo hardware. Log the ‘opticalFlow_data’ on the hardware using ['To Workspace' block](https://www.mathworks.com/help/simulink/slref/toworkspace.html). [Obtain the MAT file](https://www.mathworks.com/help/supportpkg/parrot/ug/using-flight-control-interface-to-obtain-the-log-files.html) for the logged optical flow data using the Flight Control Interface. Compare it with the results obtained from simulations. Check the Background Material section for details. - -Project Variations: - - Design a deep learning model to estimate a vehicle’s displacement to correct the IMU only estimation [4] - - -## Background Material - - - Getting started self-paced courses - [MATLAB Onramp](https://www.mathworks.com/learn/tutorials/matlab-onramp.html), [Simulink Onramp](https://www.mathworks.com/learn/tutorials/simulink-onramp.html), [Control Design Onramp](https://www.mathworks.com/learn/tutorials/control-design-onramp-with-simulink.html), [Image Processing Onramp](https://www.mathworks.com/learn/tutorials/image-processing-onramp.html) - - Deploy to hardware using [Simulink Support Package for Parrot Minidrones](https://www.mathworks.com/help/supportpkg/parrot/) - - Video on [obstacle avoidance with a camera sensor](https://www.youtube.com/watch?v=YTmq13xGnLg) that uses optical flow information from the simulated camera image for a different task - - Video series on [Drone Simulation and Control](https://www.mathworks.com/videos/series/drone-simulation-and-control.html) that explains the workflow for developing a control system for the Parrot Mambo Minidrone and explains how to deploy the algorithms on the hardware - - Check out [Optical Flow with Parrot Minidrones](https://www.mathworks.com/help/supportpkg/parrot/ug/optical-flow-with-parrot-minidrones.html) on the documentation page - - -Suggested readings: - -[1] Roldán, J.J., Joossen, G., Sanz, D., Del Cerro, J., Barrientos, A. Mini-UAV Based Sensory System for Measuring Environmental Variables in Greenhouses. Sensors 2015, 15, 3334-3350. https://doi.org/10.3390/s150203334 - -[2] Ho HW, de Croon GC, Chu Q. Distance and velocity estimation using optical flow from a monocular camera. International Journal of Micro Air Vehicles. September 2017:198-208. doi:10.1177/1756829317695566 - -[3] B. Herisse, F. Russotto, T. Hamel and R. Mahony, "Hovering flight and vertical landing control of a VTOL Unmanned Aerial Vehicle using optical flow," 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2008, pp. 801-806, doi: 10.1109/IROS.2008.4650731. - -[3] Wenxin Liu, David Caruso, Eddy Ilg, Jing Dong, Anastasios I. Mourikis, Kostas Daniilidis, -Vijay Kumar, and Jakob Engel, “TLIO: Tight Learned Inertial Odometry,” IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 5, NO. 4, OCTOBER 2020 - - -## Impact - - Advance aerial vehicle control in contracted spaces with unforeseen environment conditions. - -## Expertise Gained - -Autonomous Vehicles, Computer Vision, Drones, Robotics, Aerospace, Control, Image Processing, Low-cost Hardware, Modeling and Simulation, Signal Processing, State Estimation, UAV - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/68) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -234 diff --git a/projects/Voice Controlled Robot/README.md b/projects/Voice Controlled Robot/README.md deleted file mode 100644 index a019af45..00000000 --- a/projects/Voice Controlled Robot/README.md +++ /dev/null @@ -1,66 +0,0 @@ -**Project 30:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the rewards - - - -

Voice Controlled Robot

-

Smart devices and robots are nowadays part of our everyday life and human-robot interaction plays a crucial role in this rapidly expanding market.

-
- -## Motivation - -AI-powered virtual assistants can buy us dinner while a robotic vacuum cleans our home as we seat on the couch talking to our smart TV. -Clearly, our lives are changing and adapting to the omnipresence of these devices. -The need to interact with such devices, not only at home but also at work and even on the streets, has made ‘human-robot interaction’ a crucial and rapidly increasing field. -Get prepared to this new trend and for a career in this field by acquiring the skills to build a voiced controlled robot. - - -## Project Description - -Design and program a wheeled robot to respond to voice commands, visually identify specified targets, and navigate around obstacles to reach the target. -Use a low-cost platform such as Arduino or Raspberry Pi to build the robot and design the system including robot mission, control, and perception using Simulink. -Eventually deploy the designed system to the robot. - -Suggested steps: - -1. Use MATLAB or Simulink to establish a framework of the vision-guided robot, and make sure if you hand code a target, the robot will search for the target visually, and go there. -2. Add voice input, pass it through voice categorizer to the robot planner (you can either do it by defining specific features in the sound signal, or do it the deep learning way, find an example [here](https://www.mathworks.com/help/audio/ug/Speech-Command-Recognition-Using-Deep-Learning.html#d123e9007)) -3. At minimum, you should have “return”, “turn left”, “turn right”, “red target”, “blue target”; as voice commands -4. Connect the whole system and play with it. - - -## Background Material - -- [Speech Command Recognition Using Deep Learning](https://www.mathworks.com/help/audio/ug/Speech-Command-Recognition-Using-Deep-Learning.html#responsive_offcanvas) -- [Raspberry Pi Support from Simulink - Hardware Support - MATLAB & Simulink (mathworks.com)](https://www.mathworks.com/hardware-support/raspberry-pi-simulink.html#:~:text=Supported%20Hardware%20%20%20%20Raspberry%20Pi%20Model,R2016a%20-%20Current%20%203%20more%20rows) -- [Simulink Support Package for Arduino Hardware Documentation (mathworks.com)](https://www.mathworks.com/help/supportpkg/arduino/index.html#:~:text=The%20support%20package%20includes%20a%20library%20of%20Simulink,by%20entering%20it%20in%20the%20MATLAB%20Command%20Window) -- [Control a Raspberry Pi powered robot with MATLAB and Simulink](https://www.mathworks.com/matlabcentral/fileexchange/47376-control-a-raspberry-pi-powered-robot-with-matlab-and-simulink) -- [Building Smart Robots Using Simulink and Arduino](https://www.youtube.com/watch?v=lo7UK84Lto0) -- [Voice controlled robot](https://www.mathworks.com/matlabcentral/fileexchange/57528-voice_controlled_robot) -- [Simulating Mobile Robots with MATLAB and Simulink](https://www.youtube.com/watch?v=7p2McZCKvus) -- [Detect People with Raspberry Pi and MATLAB Online](https://www.mathworks.com/videos/detect-people-with-raspberry-pi-and-matlab-online-1563770971238.html) -- [Neurorobots for Education](https://www.mathworks.com/products/connections/product_detail/backyardbrains-neurorobots.html) - -## Impact - -Open up the opportunities to create robots that can be an intuitive part of our world. - -## Expertise Gained - -Artificial Intelligence, Computer Vision, Robotics, Signal Processing, Human-robot Interaction, Natural Language Processing, Mobile Robots, Low-cost Hardware - -## Project Difficulty - -Bachelor level - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/7) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[dryouwu](https://github.com/dryouwu) - -## Project Number - -30 - diff --git a/projects/Voice Controlled Robot/student submissions/voice-controlled-robot b/projects/Voice Controlled Robot/student submissions/voice-controlled-robot deleted file mode 160000 index ad6d2238..00000000 --- a/projects/Voice Controlled Robot/student submissions/voice-controlled-robot +++ /dev/null @@ -1 +0,0 @@ -Subproject commit ad6d2238c47f76340c4e6534754d48911bb69b9e diff --git a/projects/Warehouse Robotics Simulation/README.md b/projects/Warehouse Robotics Simulation/README.md deleted file mode 100644 index 50e20f04..00000000 --- a/projects/Warehouse Robotics Simulation/README.md +++ /dev/null @@ -1,60 +0,0 @@ -**Project 212:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Warehouse Robotics Simulation

-

Simulate multirobot interactions for efficient algorithm design and warehouse operations.

-
- -## Motivation - -Online shopping and fast delivery have been the focus of ecommerce giants. Large warehouses storage and retrieval of multiple items have been the key for fast delivery of items to the customers. Robots (ground robots, aerial robots, manipulators robots) play a major role in automating this process for many warehouse applications. Simulation is the key to understanding the different scenarios, determining the suitability of the algorithms, and optimizing the warehouse design and robot requirements. - -## Project Description - -This project focuses on the design and simulation of warehouse robotic operations. Using the functionality available in perception, navigation, fusion, and applications products listed here, [Computer Vision Toolbox™](https://www.mathworks.com/products/computer-vision.html), [Navigation Toolbox™](https://www.mathworks.com/products/navigation.html), [Sensor Fusion and Tracking Toolbox™](https://www.mathworks.com/products/sensor-fusion-and-tracking.html ), [UAV Toolbox](https://www.mathworks.com/products/uav.html), and [Robotics System Toolbox™](https://www.mathworks.com/products/robotics.html), simulates the warehouse operations and ascertains their efficiency and suitability for the task at hand. The [Gazebo-Cosim](https://www.mathworks.com/help/robotics/ug/perform-co-simulation-between-simulink-and-gazebo.html) Co-simulation capability can be leveraged to ease the integration of Gazebo simulation into the rest of the processing workflow. Here are some suggested steps to get you started on the project. - -Suggested Steps: -1. Start with the existing MATLAB example [A* Path Planning and Obstacle Avoidance in a Warehouse](https://www.mathworks.com/help/robotics/ug/a-star-path-planning-and-obstacle-avoidance.html) where a single robot is used to move around the warehouse avoiding obstacles. -2. Extend that example by placing multiple robots in the warehouse -3. Develop a multirobot mobility and a task scheduling algorithm to simulate coordinated tasks, see reference [1] for further details on a sample algorithm to get you started. - -Recommended Project Variations: - -- Vary the type of robot model used in the simulation using the robot library available in [Robotics System Toolbox](https://www.mathworks.com/help/robotics/ref/loadrobot.html) -- Simulate multirobot pick and place tasks. [Pick and Place Example](https://www.mathworks.com/help/robotics/ug/pick-and-place-workflow-using-stateflow.html ) -- Warehouse operation involving moving material using coordinated robots - - -## Background Material - -Suggested reading and viewing: - -- [1] G. Wagner and H. Choset, "M*: A complete multirobot path planning algorithm with performance bounds," 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, San Francisco, CA, USA, 2011, pp. 3260-3267, doi: 10.1109/IROS.2011.6095022. -- [2] A. Cowley, B. Cohen, W. Marshall, C. J. Taylor and M. Likhachev, "Perception and motion planning for pick-and-place of dynamic objects," 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, Tokyo, Japan, 2013, pp. 816-823, doi: 10.1109/IROS.2013.6696445. - -- [Autonomous warehouse robot with 3D movement capabilities (video)](https://www.youtube.com/watch?v=PC-9HYJ1nCI) -- [Amazon Warehouse Order Picking Robots (video)](https://www.youtube.com/watch?v=Ox05Bks2Q3s) -- [Multi-Agent Path Finding (slides)](http://idm-lab.org/slides/mapf-tutorial.pdf) - - -## Impact - -Advance the automation of warehouse applications and reduce associated time and energy consumption. - -## Expertise Gained - -Autonomous Vehicles, Robotics, Human-Robot Interaction, Humanoid, Mobile Robots - - -## Project Difficulty - -Bachelor, Master's, Doctoral - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/41) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Project Number - -212 diff --git a/projects/Wind Turbine Predictive Maintenance Using Machine Learning/README.md b/projects/Wind Turbine Predictive Maintenance Using Machine Learning/README.md deleted file mode 100644 index 7c5d0afa..00000000 --- a/projects/Wind Turbine Predictive Maintenance Using Machine Learning/README.md +++ /dev/null @@ -1,64 +0,0 @@ -**Project 197:** Fill out this [form](https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5lOTzSa_A6lPqq2LJKzvpM5UMTBZRkc4UTRETjFERVRDWllQRE40OUFSQS4u) to register your intent to complete this project and learn about the reward - - - -

Wind Turbine Predictive Maintenance Using Machine Learning

-

Improve the reliability of wind turbines by using machine learning to inform a predictive maintenance model.

-
- -## Motivation - -Wind turbines are a fast-growing source of renewable energy worldwide. When turbines break down unexpectedly, repairs can be costly and energy production can be disrupted. While regular maintenance can prevent some of these issues, unnecessary down time can be expensive and disruptive. By using properly trained models, predictive maintenance can ensure that the right interventions are done at the right time, keeping the flow of electricity high and costs low. -Properly constructed models can help predict behavior in physical systems, especially if trained using machine learning. Running simulations through varying situations can provide data to teach expected behavior. - -## Project Description - -Work with [Simscape™ Driveline™](https://www.mathworks.com/products/simscape-driveline.html) to create a plant and fault model for wind turbines using MATLAB® and Simulink®. -The model should be detailed enough to capture important dynamics. -Demonstrate that out of spec performance and component failure can be predicted by identifying data that varies but is still within spec. -Determine the most useful set of sensors to ensure that predictive maintenance is possible without adding too much cost and complexity. - -Suggested steps: -1. Perform a literature search to understand wind turbines and predictive maintenance. -2. Create a dynamic wind turbine model. -3. Determine fault conditions that could cause a turbine to be damaged or taken out of service. -4. Model different scenarios including manufacturing differences, wind loads, and wind direction. -5. Determine if you can predict faults before they occur by observing key simulation data. -6. Train a predictive maintenance model using many runs of the simulation. - -Advanced project work: - -Find real data on failed turbines and determine if the model could have predicted it. - -## Background Material - -- [Simscape Driveline documentation](https://www.mathworks.com/help/physmod/sdl/index.html) -- [Predictive maintenance with MATLAB and Simulink](https://www.mathworks.com/videos/predictive-maintenance-in-matlab-and-simulink-1498594477325.html) -- Wang, Jinjiang, et al. "An integrated fault diagnosis and prognosis approach for predictive maintenance of wind turbine bearing with limited samples." Renewable energy 145 (2020): 642-650. - - -## Impact - -Contribute to providing the world with reliable green energy. - - -## Expertise Gained - -Industry 4.0, Sustainability and Renewable Energy, Machine Learning, Electrification, Modeling and Simulation, Predictive Maintenance, Wind Turbines - - -## Project Difficulty - -Master's - -## Project Discussion - -[Dedicated discussion forum](https://github.com/mathworks/MathWorks-Excellence-in-Innovation/discussions/28) to ask/answer questions, comment, or share your ideas for solutions for this project. - -## Proposed By - -[MUdengaard](https://github.com/MUdengaard) - -## Project Number - -197