Combination Optimization Model of Urban Key Intersections Based on Q-Learning Algorithm

Macro-traffic organization measures have shortcomings in traffic congestion control at specific city intersections. From the perspective of micro-governance, based on Q-reinforcement learning a combined optimization model of “channelization + signal control” was built to comprehensively manage key urban intersections in this study. The intersection of Daxue Road and Ziwei Road in Jinan city of Shandong Province is taken as an example. Vissim simulation combined with actual data was used to compare and analyze the model effect before and after optimization. The results show that the average delay at the intersection is reduced from 55.01 to 43.15 s, and the average queue length is reduced from 27.08 to 13.92 m. The combined optimization model can significantly improve the traffic efficiency of bottleneck intersections, and the traffic operation effect is better than the traditional signal timing method. It can provide a theoretical reference for the congestion management of key intersections in large cities.