A Multi-Agent Cooperative Attention Framework for Joint Control of Traffic Signal and Vehicles

This letter presents a novel Traffic Signal and Vehicle Joint Control (TSVJC) framework, which seamlessly integrates traffic signal regulation and vehicle speed control to enhance road traffic efficiency. Within this framework, each intersection operates a dedicated signal control process managed by specialized signal agents responsible for adjusting traffic signals. Simultaneously, vehicle agents in the vicinity dynamically adjust their speed as part of a coordinated optimization process. To facilitate effective collaboration between these heterogeneous agents, we introduce the Multi-Agent Cooperative Attention Optimization (MACAO) algorithm, which generates weight vectors based on observed traffic states across multiple intersections through Graph Attention Networks (GAT). Simulation results demonstrate that the proposed algorithm substantially reduces vehicle travel time and enhances traffic efficiency compared to existing solutions.