Control Strategies for Pursuit-Evasion Under Occlusion

Overview

Project link

Models (Optional)

The MPT models included here are pretrained from the work described in Motion Planning Transformer. These models will be placed under pursuer/planner/models during the setup process.

The Motion Planning Transformers code base is currently for reviewing purpose alone. Please do not distribute.

Setup

1. Ensure Anaconda (or Miniconda) is installed on your system and available in your terminal.
2. From the root of this repository, make the setup script executable with: chmod +x setup.sh
3. Run:
4. ./setup.sh

This creates a new Conda environment named erl_vc, installs the necessary packages and downloads the models.

Usage

After running setup.sh, activate the environment: conda activate erl_vc

You can then run: python main.py to start the simulation, and press q to quit simulation.

Notes

• If gdown is missing, the script installs it automatically to facilitate model downloads.
• If requirements.txt is not present, the script skips dependency installation.
• The environment can be removed by running: conda remove --name erl_vc --all
• Path planning and CARLA-ROS integration is still under migrating from old code base and not available at this time.

Reference

If you find this work useful, please cite it as follows:

• M. Zhou, M. Shaikh, V. Chaubey, P. Haggerty, S. Koga, D. Panagou and N. Atanasov Control Strategies for Pursuit-Evasion Under Occlusion Using Visibility and Safety Barrier Functions, IEEE International Conference on Robotics and Automation (ICRA), 2025.

@inproceedings{Zhou2025VisibilityControl,
	author = {Minnan Zhou and Mustafa Shaikh and Vatsalya Chaubey and Patrick Haggerty and Shumon Koga and Dimitra Panagou and Nikolay Atanasov},
	title = {Control Strategies for Pursuit-Evasion Under Occlusion Using Visibility and Safety Barrier Functions},
	booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
	year = {2025}
}