Lead Researcher: Junshan Zhang
Project Goal: To devise warm-start RL algorithms that can learn driving policy quickly to improve safety, accelerated by offline training.
Research design: Inspired by AlphaZero, we study warm-start reinforcement learning (RL), where online RL is carried out and accelerated by a prior policy trained by offline
Expected outcome: Develop RL-based behavior/trajectory planning and real-time control algorithms for AV driving while reducing the training time significantly.