Traffic Operation Evaluation And Optimization At Signalized Intersections Based On Connected Vehicle Trajectory Data

In recent years,the vigorous development of intelligent information and communication technology has brought new data and opportunities for traffic operation evaluation and optimization at signalized intersections.Connected vehicle operation platforms represented by Amap,Baidu,and Didi companies can provide massive(a large number of sampled vehicles)and high frequency(high uploading frequency of vehicle positions)spatiotemporally continuous vehicle trajectory data.These connected vehicle trajectory data have the advantages of wide coverage and low maintenance cost,which can not only dynamically reflect the spatiotemporal distribution of road network traffic demands,but also accurately describe the operation law of traffic flows.Nevertheless,the connected vehicles are randomly distributed in the urban road network and the current overall penetration rate is still low(usually no more than 10%,and even less than 5% at local periods or intersections).Thus,how to fully exploit the abundant traffic flow information hidden in the randomly sampled connected vehicle trajectory data to achieve accurate and efficient traffic operation evaluation and optimization at signalized intersections,is a hot research direction of engineering application in the field of urban intelligent transportation,and also a frontier scientific problem that needs to be solved urgently.Regarding the above problems,this study constructs a traffic operation evaluation and optimization framework for signalized intersections with low-penetration-rate connected vehicle trajectory data as the only input.The study proposes several estimation methods for key parameters including arrival rate,queue length,and cumulative flow diagram,develops optimization methods for time-of-day based fixedtime signal control and adaptive signal control,and achieves reliable estimation of traffic operation evaluation parameters and global optimization of signal control at intersections under the condition of low-penetration-rate connected vehicle data.The main research contents and achievements of this study are as follows:(1)Connected vehicle trajectory data-driven framework for traffic operation evaluation and optimization at signalized intersections.Considering the development trend of signal control under the background of the emerging data and the deficiencies of existing related studies,to get through the key technologies of developing a new generation of signal control system based on connected vehicle trajectory data,this study takes historical and real-time observed connected vehicle trajectories as input,constructs a traffic operation evaluation and optimization framework for fixed-time and real-time signal-controlled intersections.The framework includes an operation evaluation layer and a control optimization layer.The former builds an estimation method system for key traffic operation parameters based on five core methods,which comprehensively evaluate the traffic state at fixed-time and realtime signal-controlled intersections and provides data input for the control optimization layer.Based on the input from the operation evaluation layer,the latter constructs two trajectory data-driven traffic flow characterizing models for fixed-time and real-time signal-controlled intersections respectively,and correspondingly develops optimization methods achieving global optimization of signal control at isolated intersections.(2)Traffic operation evaluation at intersections based on historical and realtime connected vehicle trajectory data.Existing studies only using real-time connected vehicle trajectory data cannot achieve a reliable estimation of traffic operation parameters under a low penetration rate environment.This study fully exploits the prior and real-time traffic flow information provided by historical and realtime connected vehicle trajectories,proposes five core methods that are model-driven or data-driven,and achieves reliable estimation of key parameters of traffic operation at signalized intersections such as arrival rate,queue length,and vehicle cumulative flow diagram in low-penetration scenarios.The simulation and empirical evaluation results have shown that,after considering the prior information provided by historical connected vehicle trajectories such as arrival distribution,queuing position distribution,and the number of connected vehicles,the proposed methods can achieve stable and reliable estimates even under low penetration rates,whose estimation accuracy is significantly higher than existing studies.Under the condition of low penetration rates of 2%-3%,the relative error of cycle-based arrival rate estimation is only about 15%and the absolute error of cycle-based queue length estimation is only about 2 vehicles.(3)Fixed-time signal control optimization at isolated intersections based on historical connected vehicle trajectory data.Considering that the assumption of uniform vehicle arrival in existing studies constrains their practical applications,this study proposes a fixed-time signal timing optimization method that considers the timedependent vehicle arrival distribution during the cycle.The method firstly proposes a cumulative flow diagram model that is purely driven by connected vehicle trajectory data to precisely describe the dynamic evolution process of the traffic operation parameters with various signal control parameters.Then,based on this model,a multiobjective optimization model suitable for both undersaturated and over-saturated traffic conditions is built to minimize the theoretical vehicle delay and exceeded queue dissipation time.Finally,given the data-driven feature of the optimization model,a bilevel solution algorithm based on particle swarm optimization is developed to globally optimize signal control parameters.The simulation test results show that the proposed optimization method is insensitive to the penetration rates and the arrival patterns.Under different penetration rates and degrees of saturation,the proposed method can achieve smaller vehicle delay and queue length compared to the signal plans by Synchro.By considering the time-dependent arrival distribution during the cycle,this method realizes more precise optimization of signal control parameters and achieves better signal control benefits.(4)Adaptive signal control optimization at isolated intersections based on real-time connected vehicle trajectory data.Considering that existing studies are oriented to a high penetration rate or fully connected environment,or to integrate the vehicle arrival information provided by fixed-location detectors,which is difficult to apply to the current low-penetration-rate and low-positioning-accuracy connected vehicle environment,this study proposes an adaptive signal control method that considers the stopping information of connected vehicles at upstream intersections.The method firstly proposes an arrival prediction model for all vehicles based on small sample vehicle data,by tracing the stopping information of connected vehicles at upstream intersections.Then,an adaptive signal control optimization model based on the predicted arrivals is constructed in a rolling optimization framework,to minimize the total delay.Finally,a dynamic programming algorithm is developed to efficiently optimize signal control parameters in real-time.The simulation test results show that the proposed method can achieve better control benefits than fully actuated control under different penetration rates and with medium and high input volume.Under the low penetration rate of 5% and the penetration rate of the current connected vehicle environment of 10%,the proposed method can reduce the total delay by 17.2% and24.9%,respectively,compared to the fully actuated control.By considering the trajectory information of connected vehicles at upstream intersections,this method effectively addresses the limitation of insufficient traffic flow information provided by trajectory on the link and solves the problem of real-time signal control optimization at isolated intersections in the scenario of connected vehicles with low penetration rate and low positioning accuracy.To sum up,this study focuses on the problem of reliable estimation of traffic operation parameters and global optimization of signal control at intersections under the condition of low-penetration-rate connected vehicle trajectory data.The key parameter estimation methods for arrival rate,queue length,and cumulative flow diagram,as well as fixed-time and real-time signal control methods for isolated intersections,are proposed based on connected vehicle trajectory data,which enriches and improves the theories and methods of traffic operation evaluation and optimization at signalized intersections under the background of emerging technologies,and provides support for the development of a new generation of signal control systems based on connected vehicle trajectory data.