Research On Queue Length Estimation Of Oversaturated Signal Intersections Based On Networked Vehicle Trajectory Data

As an information input and evaluation index for traffic control and signal timing,queue length can visually reflect the traffic operation condition of intersections.As the number of networked vehicles grows,vehicle trajectory data is available in a variety of ways,and the content of trajectory data is gradually improved,the use of networked vehicle trajectory data to estimate intersection queue lengths has also received increasing attention from researchers.Most of the current studies use traffic wave models or probabilistic statistical models to estimate queue lengths at intersections,but most of them require assumptions on vehicle arrival patterns and cannot be fully consistent with the actual traffic operating conditions at intersections;on the other hand,most of the studies are only applicable to under-saturated traffic conditions and do not perform well in over-saturated traffic conditions.Therefore,in this paper,we use the trajectory data of networked vehicles to achieve the estimation of intersection vehicle arrival rate and saturation flow rate,and on this basis,we construct the initial queue length and maximum queue length estimation models for oversaturated intersections,avoiding the assumption of vehicle arrival mode and giving the intersection queue length estimation method under traffic oversaturation conditions.The main work of the paper includes intersection vehicle arrival rate estimation model construction,oversaturated intersection queue length estimation model construction,and method performance evaluation.In terms of intersection vehicle arrival rate estimation modeling,a vehicle arrival rate estimation model is constructed for different scenarios based on the number and type of network-linked vehicle trajectories observed during the signal cycle,and the corresponding traffic density is calculated in conjunction with traffic flow theory,on the basis of which a saturation flow rate estimation model is constructed.In the modeling of queue length estimation at oversaturated intersections,by analyzing the generation and propagation process of traffic shock waves at oversaturated signal intersections,the initial queue length and maximum queue length estimation models for oversaturated signal intersections are established using the trajectories of networked vehicles observed during the signal cycle;for the possible overflow backflow phenomenon at intersections under oversaturation,this paper detects whether the downstream of the intersection occurs through the trajectory data The queue length estimation model is further extended to the intersection overflow backflow scenario by detecting whether overflow occurs downstream of the intersection through trajectory data.In particular,for the case where the penetration rate of net-linked vehicles is low,the number of available net-linked vehicle trajectories is insufficient.At this time,the study takes advantage of the fact that queue lengths at intersections have continuity over adjacent cycles and proposes a method based on fitting historical queue data.In terms of method performance evaluation,this paper selects the intersection of Ganjiang Middle Avenue and Luyin Road in Nanchang City as the research object and uses the microscopic simulation platform SUMO to establish an experimental model.After determining the reliability of the simulation model,simulation experiments with different permeability rates and different saturation degrees are designed,and MAE and MAPE are selected as evaluation indexes.The experimental results show that the initial queue length MAE value obtained by the method based on the networked vehicle trajectory data is less than 8.5m and the MAPE value is less than 11.0%,and the MAE value of the maximum queue length is less than 12.0m and the MAPE value is less than 8.5% at a networked vehicle penetration rate of not less than 20% and an intersection loading degree greater than 1.0;at a networked vehicle penetration rate of not more than 10% and an intersection load factor is greater than 1.0,the MAE value of the initial queue length obtained based on the historical queuing method is less than7.0m and the MAPE value is less than 9.5%,and the MAE value of the maximum queue length is less than 10.5m and the MAPE value is less than 8.5%.It is determined that the method in this paper can better achieve the estimation of initial queue length and maximum queue length for oversaturated signal intersections.