Cooperative Traffic Signal Control with Bus Priority: A Multi-Agent Reinforcement Learning Approach

Intelligent traffic signal control plays a crucial role in reducing the escalating problem of traffic congestion. However, traditional methods of traffic signal control struggle to effectively adapt to the ever-changing road conditions and demands in complex urban traffic environments. This becomes particularly problematic when considering the punctuality of buses, as it can lead to traffic congestion for other vehicles. To address this issue, this paper introduces a multi-agent reinforcement learning method that prioritizes bus punctuality while minimizing travel time for regular vehicles. We established the SUMO simulation environment with dedicated bus lanes and designed a reward function based on bus arrival and vehicle waiting times. By employing a deep reinforcement learning approach, we aimed to reduce average travel time. The simulation results demonstrate significant improvements, including a 74.44% reduction in vehicle waiting time at intersections and a 78.06% decrease in bus delay time. Notably, even when focusing solely on optimizing vehicle travel time, our approach enhances bus punctuality, resulting in a 57.03% reduction in bus delay time. Thus, we believe that our approach offers a viable solution for addressing bus priority policy challenges.