路径规划及信号灯控制的智能优化

Intelligent Optimization of Path Planning and Signal Light Control

由于全智能化的交通控制系统尚未得到完全普及,道路交通资源占用时空二维不够均衡、城市交通规划策略与通行车辆规模不够匹配,导致城市交通拥塞、车辆准点率低等问题。针对上述问题,首先构建了基于服务员休假的多队列多服务员排队模型和级联队列的多跳排队模型,分析了在十字路口处及一般道路情况下车辆流量、信号灯时长与通勤时延的理论关系;然后提出了基于近端策略优化的路径规划及信号灯控制的智能优化方法,通过动态感知城市实时交通拥塞状态,联合规划车辆最优路径和控制信号灯最优时长,从而最大化车辆准点率。仿真结果表明,所提方法比于自适应高效联邦对比方法在时延性能方面提升了3倍以上,在车辆准点率方面提升了18%左右。

The problems of urban traffic congestion and low vehicle punctuality rates result from the incomplete popularization of fully intelligent traffic control systems,imbalanced spatiotemporal and twodimensional occupation of road traffic resources,and insufficient matching between urban traffic planning strategies and the scale of vehicles.In response to above problems,a multi-queue multi-server queuing model based on server vacation and a multi-hop queuing model for cascaded queues are constructed.The theoretical relationship between vehicle flow,signal duration,and commuting delay in intersections and general road conditions is analyzed.Then,an intelligent optimization method for path planning and signal control based on proximal policy optimization is proposed.The method can jointly plan the optimal vehicle path and control the optimal signal duration by dynamically perceiving the real-time traffic congestion status of the city.Thereby the vehicle punctuality rate is maximized.The simulation results show that the proposed method improves the delay performance by more than three times and vehicle punctuality by about 18%compared with the adaptive and efficient federated comparison method.