RL_Soccer

This is a reinforcement learning based project. You play a 1v1 with an AI agent which has been trained for 10 hours to master this simple game.

In the Play tab you can play against 10 different opponents. Each opponent has been created during the same training session. With each new opponent 100 episodes of training are added. The agent gradually increases in difficulty to the point where it outperforms most humans (On second thought, it outperforms me).  

In the Replay_game tab you can see the progress the agent makes. In this tab you will find two agent. These are two different agents which have learned the game via playing each other. Both getting better and raising the level of play.

This is not supposed to be a full blown game. It initially was just a project to test the TD3 algorithm. However, since it was a 1v1 I figured it would be fun to make a small game out of it.