This folder contains an RL environment based on open-source game Gameplay, A3C and PPO agents
Useful links:
Certain algorithms are taken from online github or inspired by online works.
This project focuses on MARL in Google football/soccer environment, the algorithms used are: PPO and A3C.
The A3C folder contains MARL for A3C, while the standalone A3C file in this folder is for manual control against the A3C model (previously trained).
The models must be loaded using user's respective path. Existing path is done for this current project and would not work on other user's computer. Please load checkpoint or model path to your own respectively.
Please refer to our report for the final results and analysis of our project MARL.
Note: The actions from trained model are the players in yellow.
This is modification to the game play in order to test against the trained model. Hence, users (Blue Team) can play against the actions from the trained model (Yellow Team)
