carsonquiason / ppo-rocket-league-bot

A PPO reinforcement learning Rocket League bot made in RLGym that has been converted to the RLBot framework

Carball AI RLbot

A Reinforcement Learning Rocket League Bot made with the PPO algorithm in RLGym. This current repo serves as the conversion from RlGym to the RLbot framework for easier testing against official and user created bots.
The RLGym version of the bot can be found HERE

Play Against our Bot

1. Clone this repo
2. Download RLBot HERE
3. Open RLBot, Click Add->Load Folder->Select Cloned Folder
4. Add Carball AI & Human to seperate teams
5. Click Launch Rocket League and start match

Extra Installations (If needed)

Menu->Install Missing Python Packages
rlbot==1.* torch>=1.5.1+cpu rlgym-compat>=1.0.2 rlgym-tools>=1.6.1 numpy stable_baselines3 pickle5

Changing the bot

• Bot behavior is controlled by bot.py
• Bot appearance is controlled by appearance.cfg
• Switch branches to reflect the desired model (Restart Required on RLBot after switching branches) (Main contains the 500m Model)

See https://github.com/RLBot/RLBotPythonExample/wiki for documentation and tutorials.