DecNashPlanning

Decentralized game-theoretic planning with v0.0.2 of the Interaction Simulator, which uses the Interaction Dataset scenarios to determine vehicle paths and spawn times.

Code for "Multi-Vehicle Control in Roundabouts using Decentralized Game-Theoretic Planning" at IJCAI 2021 AI4AD Workshop.

If you find this repository useful, please cite the paper:

@inproceedings{jamgochian2021multivehicle,
  author = {Arec Jamgochian and Kunal Menda and Mykel J. Kochenderfer},
  title = {Multi-Vehicle Control in Roundabouts using Decentralized Game-Theoretic Planning}
  series = {Artificial Intelligence for Autonomous Driving Workshop},
  booktitle = {International Joint Conference on Artificial Intelligence (IJCAI)},
  year = {2021}
}

Getting started

Clone the DecNashPlanning repository

git clone https://github.com/sisl/DecNashPlanning.git
cd DecNashPlanning

Clone InteractionSimulator with the decnash tag and install requirements using poetry.

git clone --branch decnash https://github.com/sisl/InteractionSimulator.git
cd InteractionSimulator
poetry install
poetry shell
cd ..
export PYTHONPATH=$(pwd):$PYTHONPATH

Copy INTERACTION Dataset files: The INTERACTION dataset contains a two folders which should be copied into a folder called ./InteractionSimulator/datasets:

• the contents of recorded_trackfiles should be copied to ./InteractionSimulator/datasets/trackfiles
• the contents of maps should be copied to ./InteractionSimulator/datasets/maps

Install additional requirements for DecNashPlanning:

pip install -r requirements.txt

Copy INTERACTION Dataset files: The INTERACTION dataset contains a two folders which should be copied into a folder called ./InteractionSimulator/datasets:

• the contents of recorded_trackfiles should be copied to ./InteractionSimulator/datasets/trackfiles
• the contents of maps should be copied to ./InteractionSimulator/datasets/maps

Download and install Julia . We use version 1.5.3.

Running experiments

You can run all experiments to generate trajectories and videos using python experiments/exp_wrapper.py (which makes successive calls to experiments/experiment.py with appropriate settings.

To calculate metrics from already generated trajectories, run python experiments/metrics.py.

To generate plots used in paper for single game solution, run python experiments/single_game.py.

You can manually make animations from saved states, graphs, lengths, and widths (post-process) using python experiments/animate.py.