Reinforcement Learning Race Car

I implemented a reinforcement learning approach for autonomous race car control using the Soft Actor-Critic algorithm.

Key achievements:

• Achieved 100% lap completion and 141 second lap time on a 4.35 km long track
• Demonstrated viability of RL for autonomous race car control by implementing the Soft Actor-Critic RL algorithm from Stable-Baselines3
• Enabled successful learning through the use of reward functions
• Implemented an observation space with simulated LiDAR readings in a gymnasium wrapper class
• Utilized BeamNG.tech simulator for testing and training