Online DNN-Driven Nonlinear MPC for Stylistic Humanoid Robot Walking with Step Adjustment

Giulio Romualdi, Paolo Maria Viceconte, Lorenzo Moretti, Ines Sorrentino, Stefano Dafarra, Silvio Traversaro, and Daniele Pucci
Paolo Maria Viceconte and Giulio Romualdi are co-first authors

📅 This paper has been accepted for publication at the 2024 IEEE-RAS International Conference on Humanoid Robots (Humanoids), Nancy, France 🤖

📚 Paper     🎥 Video     🖼️ Poster     🔧 Experiments     🌐 Website     📂 Dataset

Online.DNN-driven.Nonlinear.MPC.for.Stylistic.Humanoid.Robot.Walking.with.Step.Adjustment.mp4

Reproducing the Experiments

To reproduce the experiments, we provide a conda environment for easy setup. Follow the steps below:

1. Install the Environment

Run the following command to create the conda environment:

conda env create -f environment.yml

2. Activate the Environment

Activate the newly created environment:

conda activate dnn-mpc-env

3. Run the Simulation

Start the experiment with:

./run_simulation.sh

This script will:

• Launch the Gazebo simulator
• Start the YARP server (for simulator communication)
• Initialize the DNN-driven MPC controller

When prompted, type y and press Enter to start the simulation. The humanoid robot will begin walking, and you can observe its behavior in the Gazebo simulator.

⚠️ Known Issue: The Gazebo real-time factor is scaled by a factor of 10. This is due to the use of the MUMPS linear solver in the IPOPT docker image. Alternative solvers (e.g., MA27) are available but cannot be redistributed.

---

Maintainers

👨‍💻 @paolo-viceconte

👨‍💻 @GiulioRomualdi