A GAME AUTOMATION USING RL
-
data preprocessing, training: we created game first we created a game of single bullet - commit 1- Vanshika then we added more bullets and sinks - commit 2- Vamsi created environment for single bullet using gymnasium - commit 3- Vanshika train it on A2C and PPO algorithms using stable-baselines3 - commit 4- Vamsi created environment for 5 bullets - commit 5- Vanshika train it on A2C and PPO algorithms using stable-baselines3 - commit 6- Vamsi
-
Tasks for final submission:- Find a better reward function in case of multiple bullets Try some other algorithms Change grid shape and some rules Train the final model
-
Challenges we are facing:- In case of multiple bullets we were unable to train well. We will try running it for longer time to see if we can recieve desirable result. Develop a better reward function, tune hyper-parameter.