From most to least basic:
In the folder basic-GameTheory-SetUp contains some basic (first pass), very simple, versions of the 2-player interaction with game theoretic dynamics.
In the folder DEMO is a pygame interface of an OpenSpiel game set-up for the DARPA CASTLE PI meeting March 2024.
In the folder RL is the current worl towards a trained PPO RL agent integration into CC2: https://github.com/cage-challenge/cage-challenge-2. In this dynamic 2-player environment of a consistant topology, an approximate course correllated equilibrium is to be computed "on top of" the PPO Agent, taking the adversarial invader into account as part (stochasticity) of the environment.
In the folder MARL is the current work toward a trained MARL integration into both cotraining adversarial RL agent in CC2 or in the 5 cooperative RL agents in CC4: https://github.com/cage-challenge/cage-challenge-4. In this complex, dynamic environment that is a random size (initialized randomly at start) and provides an asyncronous setting due to assymetric time-ticks for actions, an approximate course correllated equilibrium will be computed "on top of" the MARL agents, taking the adversarial invaders into account as part of the environment.