Reinforcement Learning for Formula 1 Strategy

Developed an AI-based strategy tool utilizing reinforcement learning to optimize pit stop scheduling and tyre compound selection during live Formula 1 races.

Customised a race simulator to replicate tyre degradation, fuel consumption, and race dynamics from past race data.

Utilized TensorFlow and TF-Agents for the reinforcement learning environment setup.

Conducted competitive analysis by assigning different algorithms (DQN, DDQN, Dueling Networks) to simulated racing drivers. Evaluated model's performance based on positions gained on track.

Analysed past f1 races and the impact of different reinforcement learning techniques on race outcomes, emphasizing the balance between exploration and exploitation.

Repo: https://github.com/PrasanthVR63/F1ReinforcementLearning.git