ML_tetris

• Our end goal is to perfect a reinforcement library containing different algorithms applicable for different gaming scenarios.

• We’ll make our own version of Tetris using Unity, and integrate the reinforcement learning algorithms we’ve implemented so far to train a competitive gaming AI.

• The AI model would take the current game state as input and produce an action (such as moving or rotating a tetromino) to maximize its long-term cumulative reward. By training the model using reinforcement learning, it can learn to make intelligent decisions based on the feedback it receives from the game.

• Implement A2C and DQN reinforcement learning algorithms from scratch using PyTorch.

• Develop a Tetris game using Unity and integrate the implemented reinforcement learning algorithms to train a competitive AI model.

• Expand the reinforcement learning library by implementing additional algorithms such as PPO (Proximal Policy Optimization) and others suitable for various gaming scenarios.

• Document the implementation process, provide clear explanations, and create tutorials to facilitate the usage of the reinforcement learning library for other developers and researchers.