Traffic Signal Timing Optimization For Congested Hot Spots In Multi-source Data Environment

Under the supply of limited urban road network,refined traffic management and control is one of effective techniques to solve congestion problem caused by the imbalance of supply and demand.Data-driven traffic signal timing optimization is of great significance for improving the quality of traffic management and control,easing congestion and improving the efficiency of traffic operation.The traditional fixed-point detector has limitations of limited coverage,high construction cost.With the rapid development of connected technology,the emerging mobile detectors provide new data support for traffic signal optimization.How to combine various data sources for intelligent optimization of traffic signal timing is a cutting-edge topic in the field of traffic control.In this context,based on the integration of vehicle location tracing trajectory data,intersection license plate recognition data and signal timing data,this study proposes signal timing optimization models for congested areas in the multi-source data environment.Details of the research work can be divided into two parts: traffic state estimation and signal timing optimization.Firstly,utilizing queue length as the parameter to quantify traffic state,this study proposes a queue length dynamic estimation method combining vehicle location tracing trajectory and license plate recognition data.When the penetration rate is relatively high,a discharging shockwave reconstruction-based method for queue length estimation is developed,in which discharge points are extracted from vehicle trajectories and Bayesian method is used to estimate the posterior distribution of queue length.When the penetration rate is relatively small and vehicle trajectory samples are limited,this study introduces license plate recognition data and constructs a bi-level Random Forest model to classify vehicles into two categories and estimate the stop location of queueing vehicles,exploring the relation of two data sources and increasing the number of discharging points.NGSIM dataset,taxi GPS trajectories and license plate data in Kunshan are used to examine the performance of proposed models and conduct sensitivity analysis,with the results showing that the estimation accuracy is satisfying in various penetration rate conditions.Secondly,based on the acquisition of approach queue lengths for network intersections,this study uses the ratio of queue service time and effective green duration as the indicator to quantify the saturation degree of intersection approaches and determine the critical intersections needing signal optimization.Next,for the critical and their neighboring intersections,a queue backpressure-based distributed signal optimization algorithm is developed with the objective of maximizing the output flow of critical intersections,optimizing the effective green duration for each phase.The algorithm is tested in the simulation platform constructed by the actual downtown area in Kunshan.The result demonstrates that the residual queue phenomenon is eliminated,with average vehicle delay reduced and the overall traffic efficiency of road network improved after optimization.