Cai, Bai-gen (Beijing Jiaotong University), Cao, Yan (Beijing Jiaotong University), Chai, Linguo (Beijing Jiaotong University), ShangGuan, Wei (Beijing Jiaotong University), Cao, Yue (Beijing Jiaotong University), Gao, Jin (Beijing Jiaotong University), Chen, Junjie (Beijing Jiaotong University)

A Graph Reinforcement Leaning-Based Approach for Merging of Mixed Traffic Vehicles on On-Ramp

Scheduled for presentation during the Invited Session "Modeling, Optimization and Game Control of Human-Machine Interaction Behavior in Intelligent Transportation Systems" (ThAT10), Thursday, September 26, 2024, 11:30−11:50, Salon 18

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 14, 2024

Abstract

On-ramp merging congestion is the main reason for the inefficiency of expressway operation, the emergence of CAV can effectively improve the problem, but it also faces the challenge of CAV cooperative merging in mixed traffic environment. This paper proposes a method for collective intelligent CAV decision-making at the on-ramp merging scenario under mixed traffic conditions on graph reinforcement learning. Firstly, the topological relationship of the interaction influence between vehicles is constructed based on graph theory and the markov decision process (MDP) for the collaborative merging control of CAVs is constructed to facilitate the application of deep reinforcement learning (DRL) for decision-making. Then, a dynamic mask is applied to assess the effective driving decisions of CAVs during the training process. The proposed method is able to consider the collaborative relationships of CAVs and the impact of CAV decisions on HDVs during the process of updating decision-making strategies. The numerical results indicate that CAVs under the proposed model make more collaborative driving decisions, leading to increased traffic speed and reduced queuing time, effectively alleviating merging congestion issues.