This project uses TMInterface to control and train an AI agent for TrackMania using a Deep Q-Network (DQN). The AI learns optimal driving strategies by interacting with the game and receiving rewards for actions that improve performance, such as overtaking opponents or minimizing lap times.

Features:

• Game Control: Uses TMInterface to send inputs (steering, throttle, brake) and gather game data (position, speed, checkpoints).
• Reinforcement Learning: Trains a DQN to decide the best action in any situation.
• Performance Feedback: Rewards are based on speed, checkpoint progress, and overtaking other cars.

How It Works:

1. TMInterface connects to TrackMania to monitor the game and send commands.
2. The AI interacts with the environment, collecting data like position, speed, and track state.
3. The DQN model processes this data, predicts the best actions, and learns from feedback (reward signals).
4. Over time, the AI improves its driving to achieve better lap times and racing performance.

Requirements:

• Python 3.x
• TMInterface
• TrackMania installation

Development

• This project is currently under development, and dependencies and other packages are unstable and subject to change.