Induction Control Optimization Method Based on Dynamic Sensitivity

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: ① The selection of dynamic sensitivity is affected by the space occupancy. ② Under 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.