Code for the paper - State-Constrained Zero-Sum Games with One-Sided Information
- Install the conda environment:
conda env create -f env.yml
- To train the models, run the following:
- first, run
train_uncons.sh
to train the unconstrained game. In the command line, enter./train_uncons.sh
- next, run
train_cons.sh
to train the constrained game.
- first, run
- To simulate the games using the pre-trained models, do the following:
- Run
validations_scripts/simulate_uncons.py
to simulate unconstrained game. - Run
validations_scripts/simulate_cons.py
to simulate constrained game.
- Run
Optionally, to train the Reachable Tube, navigate to reachability/experiment_scripts
and run train_hji_8D.py
Known Issues: If you get a directory exists error, run the .sh
file again and it should be ok.