SAFE REINFORCEMENT LEARNING-BASED GREEN LIGHT OPTIMAL SPEED ADVISORY FOR MIXED-TRAFFIC PLATOONS

This thesis develops a platoon-centric, safe RL-based Green Light Optimal Speed Advisory (GLOSA) system to optimize the CAV speed profile of a mixed-traffic platoon. First, we design a multi-agent RL algorithm to achieve a balance between the energy and travel efficiency of a mixed-traffic platoon, where the impact of the CAV speed policy on its following vehicles is considered. Second, to enhance robustness against unobservable Human-driven Vehicles (HVs), we use the concept of privileged information to train the RL agent. Specifically, we leverage both HV and CAV information to train the RL agent while retaining access to only CAV information during implementation. Third, we integrate Control Barrier Functions (CBFs) into the RL-based policies to ensure car-following and red-light safety. Fourth, we address signal timing undertainty by leveraging CP to estimate a confidence interval of the signal timing prediction results. We present the training and testing results of our proposed RL-based GLOSA under different penetration rates of CAVs. These results suggest our algorithm can achieve a balance between energy and travel efficiency while ensuring car-following safety and red-light safety.

Author dc:creator

• YANG RUINING

• Control Barrier Function
• Multi-Agent Reinforcement Learning
• Signal Timing Uncertainty
• Mixed-Traffic Platoon
• Green Light Optimal Speed Advisory

Licence

• https://scholarbank.nus.edu.sg/bitstreams/77fab9c4-43ab-40f8-af8c-e48bf5397f9f/download