IntersectionZoo is a cooperative eco-driving-based multi-agent reinforcement learning environment for benchmarking contextual reinforcement learning algorithms to assess their generalization capabilities. Additionally, it also aims to advance eco-driving research through standardized environments and benchmarking.
See our documentation for more information on the application of IntersectionZoo. A comprehensive report of benchmarking results is available in the documentation.
- Documentation
- Installation instructions
- Tutorials
- Intersection SUMO network files
- Benchmarking results
If you find a bug or are facing an issue, please open a new issue in GitHub. The team can be reached through the contact details listed here.
We welcome your contributions.
- Please report bugs and improvements by submitting GitHub issue.
- Submit your contributions using pull requests.
If you use IntersectionZoo in your work, you are highly encouraged to cite our paper:
V. Jayawardana, B. Freydt, A. Qu, C. Hickert, Z. Yan, C. Wu, "IntersectionZoo: Eco-driving for Benchmarking Multi-Agent Contextual Reinforcement Learning", 2024.
The Wu Lab at MIT actively maintains IntersectionZoo. The contributors are listed on the IntersectionZoo Team Page. The project was partially funded by the Utah Department of Transportation.