This Neural network bot for Rocket League.

In order to use to use this bot you have to install this plugin and python of course. And Epic Games version of the game is preferred. This bot uses reinforcement learning and especcaly Proximal Policy Optimization (PPO) for training.

Run

python run.py

You are able to configure launch by changing the file rlbot.cfg.

Training

python ppo_training.py --tag "<tag for the run>" -n <a number of instances of the game>

All settings are in the file "ppocket_rocket\ppo_cfg.yaml"

Also this project consists the training with simulator of rocket league. But it has bugs and it's not debbuged. assets.path, sim_requirements.txt, SDL2.dll, rlviser.exe, umodel.ext are files for the simulator. assets.path should consist a path to your folder with Rocket League.

dqn_training.py - is training by the algorithm DQN (Deep Q-Network). But it didn't work well and the code is not supported. I leave this code for the case if i will use an another off-policy method.