Skip to content

Latest commit

 

History

History
32 lines (20 loc) · 2.62 KB

README.md

File metadata and controls

32 lines (20 loc) · 2.62 KB
Raw

Code repository for the paper: How AI from Automated Vehicles Can Contribute to the Assessment of Human Driving Behavior

This repository contains the code for the paper How AI from Automated Vehicles Can Contribute to the Assessment of Human Driving Behavior by Tom Driessen, Olger Siebinga, Thomas de Boer, Dimitra Dodou, Dick de Waard, and Joost de Winter.

The purpose of this repository is to:

  1. Provide access to the recordings created for the demonstration.
  2. Reproduce the figures from the paper using the demonstration data.
  3. Provide access to a modified version of Openpilot so it can be used to analyze your own driving recordings.

Reproducing the analysis figures

AI Driving Assessment Plot

The analysis.py code only requires pandas and matplotlib and uses the files in the recordings folder. The full recordings of the scenarios (including video and vehicle data) are in this folder as well. The file naming convention is found in the readme located in that folder.

Cloning the Openpilot submodule from this repository is not required to reproduce the images.

Using Openpilot to analyze other data

To analyze human driving data, we made a fork of the Openpilot repository, which is included in this repository as a submodule. You can clone it directly from the fork's repository (recommended) or from this repository using the --recurse-submodules flag in the git clone command.

Please follow the installation instructions in openpilot/tools/README.md and the replay instructions in openpilot/tools/sim/README.md.

Recording virtual scenarios with JOAN

To record scenarios in the CARLA simulator using the JOAN framework (an extension that allows for human input), please consult the JOAN repository. Our scenarios were implemented in CARLA Town 1. The experiment configuration and trajectory for the bus in the surprise scenario can be found in the joan_experiments/ directory.

Citations

If you use the data or code presented here, please cite:

Driessen, T., Siebinga, O., De Boer, T., Dodou, D., De Waard, D., & De Winter, J. (2024). How AI from Automated Driving Systems Can Contribute to the Assessment of Human Driving Behavior. Robotics, 13(12), 169. https://doi.org/10.3390/robotics13120169