Deep Reinforcement Learning for Local Path Following of an Autonomous Formula SAE Vehicle

With the continued introduction of driverless events to Formula:Society ofAutomotive Engineers (F:SAE) competitions around the world, teams areinvestigating all aspects of the autonomous vehicle stack. This paper presentsthe use of Deep Reinforcement Learning (DRL) and Inverse Reinforcement Learning(IRL) to map locally-observed cone positions to a desired steering angle forrace track following. Two state-of-the-art algorithms not previously tested inthis context: soft actor critic (SAC) and adversarial inverse reinforcementlearning (AIRL), are used to train models in a representative simulation. Threenovel reward functions for use by RL algorithms in an autonomous racing contextare also discussed. Tests performed in simulation and the real world suggestthat both algorithms can successfully train models for local path following.Suggestions for future work are presented to allow these models to scale to afull F:SAE vehicle.