Code for the paper - State-Constrained Zero-Sum Games with One-Sided Information

Training and Simulation

1. Install the conda environment: conda env create -f env.yml

Training

2. To train the models, run the following:
   1. first, run train_uncons.sh to train the unconstrained game. In the command line, enter ./train_uncons.sh
   2. next, run train_cons.sh to train the constrained game.

Simulation

3. To simulate the games using the pre-trained models, do the following:
   1. Run validations_scripts/simulate_uncons.py to simulate unconstrained game.
   2. Run validations_scripts/simulate_cons.py to simulate constrained game.

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.