Mixed Vehicle Platoon Forming: A Multiagent Reinforcement Learning Approach
Yujie Shi, Haoxuan Dong, Chaozhe R. He, Yuxiao Chen, Ziyou Song
Abstract
The emerging connected and automated vehicle (CAV) technologies present opportunities to improve traffic safety, economy, and efficiency. However, the diverse uncontrollable human-driven vehicles (HDVs) will continue to predominate traffic for a long time, resulting in coexisting CAVs and HDVs in the form of mixed vehicle platoons. This study proposes a mixed vehicle platoon forming method based on a two-stage control framework to adapt to dynamic mixed traffic environments. The platoon formation generation stage creates feasible formation (i.e., a spatially coordinated mix of CAVs and HDVs ensuring safe and efficient platoon control) appropriate for mixed traffic based on the empirical formation method. The multiagent reinforcement learning is used in the second stage for realizing the platoon forming control safely and efficiently with the guidance of feasible formation. Finally, extensive simulations are conducted using the Highway-env simulator to evaluate the effectiveness of the proposed method. The results demonstrate that the proposed method can effectively control CAVs in conjunction with HDVs to form a mixed vehicle platoon, achieving an average energy efficiency improvement of up to 10.69% and a reduction in travel time by up to 2.73% compared to benchmark strategies.