I made a mini game where the agent has to cross a road with cars 🚗 without crashing, I trained an AI to play it all by itself using a brain called PPO (Proximal Policy Optimization)
No followers yet
Once you ship this you can't edit the description of the project, but you'll be able to add more devlogs and re-ship it as you add new features!
Completed fixing the game and training the model, also improved the hypertuning and then trained the model again to fit with new parameteres improved the reward and the score functions
Packed everything in the ENV -> wrapped the env in GYM -> Trained the Model( around 900_000 timesteps -> made the model play the game -> changed the UI of the game -> added variable difficulty for the game -> added reward for the model
Created the RL Game environment and wrapped it in GYM wrapper. Made it playable for a Human with 0,1,2 commands