基于动态灵敏度的感应控制优化方法

Induction Control Optimization Method Based on Dynamic Sensitivity

为评估雷视一体检测方式在感应信号控制中的实际应用效果和价值,结合雷视一体检测特征,本文建立基于车型的初始绿灯配时模型,同时考虑实时等待车辆数及不同车型的启动损失时间和通过路口区域时间;通过分析雷视一体机在不同车头时距(灵敏度)检测响应条件下单位绿灯延长时间的变化,构建基于动态灵敏度的感应控制模型,并结合平均通过车辆数、平均车速、平均损失时间以及平均最大排队长度建立通行收益指数,使用强化学习方法确定动态灵敏度选择方案。最终,使用SUMO(Simulation of Urban MObility)建立仿真模型,对定时控制、固定灵敏度控制(传统感应控制)以及本文提出的动态灵敏度控制进行验证。仿真结果显示:动态灵敏度的选择与空间占有率有关;在15%最大交通量条件下,本文提出的动态灵敏度控制相对于定时控制和固定灵敏度控制,使交叉口平均车速分别提高了31.69%和5.05%,平均损失时间分别降低了36.19%和7.44%,平均最大排队长度分别缩短了45.17%和7.78%;且在交通量达到27%最大交通量之前,动态灵敏度的控制效果均为最优。仿真结果表明,本文提出的动态灵敏度感应控制模型性能良好,能为雷视一体在感应信号控制中的实际应用提供理论参考。

To evaluate the practical application effect and value of the integrated detection method in the induction signal control,this paper develops an initial green light timing model based on the characteristics of the integrated detection method,and takes into account the number of real-time waiting vehicles and the starting loss time of different models and the time passing through the intersection area.By analyzing the change of unit green light extension time under different headway(sensitivity)detection response conditions,the paper proposes the induction control model based on dynamic sensitivity.The dynamic sensitivity selection scheme is determined using the reinforcement learning method and the traffic benefit combining the average number of vehicles,the average speed,the average loss time and the average maximum queue length.At last,the simulation model is established by the SUMO(Simulation of Urban Mobility)to verify the timing control,fixed sensitivity control(traditional induction control)and the proposed dynamic sensitivity control.The simulation results show that:①The selection of dynamic sensitivity is affected by the space occupancy.②Under the condition of 15%maximum traffic volume,compared with the timing control and the fixed sensitivity control,the proposed dynamic sensitivity control increases the average speed of the intersection by 31.69%and 5.05%,reduces the average loss time by 36.19%and 7.44%,and shortens the average maximum queue length by 45.17%and 7.78%.Before the traffic volume reaches 27%of the maximum traffic volume,the control effect of dynamic sensitivity is optimal.It shows that the proposed dynamic sensitivity induction control model has good performance and can provide a theoretical reference for the practical application of radar and vision integration in induction signal control.