Identification Of Typical Traffic State At Signalized Intersection Approach Based On Vehicle Identification Data

Traffic flow state identification at signalized intersections is an important basis for traffic signal control optimization.Traditional identification of traffic state at signalized intersections is mainly based on section detector data to realize identification of congestion state and oversaturated state.It cannot accurately capture the spatio-temporal characteristics of traffic flow state.This study relies on the vehicle identification data,especially the vehicle identification data collected by bayonet electronic police system,which is widely built on urban roads at present.According to the arrival-departure characteristics of traffic flow,the typical traffic state identification of signalized intersection approach is studied.Through the identification of the spillover state of road section lane and the blocking state of channelized section lane,the existing traffic signal control timing and traffic organization channelization problems are found,providing support for intelligent urban traffic management and control.In order to obtain accurate traffic demand information,the paper proposes a method to reconstruct the vehicle cumulative arrival curve at urban road intersections based on vehicle identification data.Based on the vehicle identification data collected by bayonet electronic police equipment,accurate traffic flow departure information can be obtained.However,this kind of data has the problem that some vehicle plates at adjacent intersections do not match,some arrival information of traffic flow is missing.Considering the real-time performance of vehicle cumulative arrival curve reconstruction and the influence of the accuracy of vehicle identification data,this study constructs two cumulative arrival curve reconstruction models based on exponential smoothing method and Kalman filter principle respectively,and conducts case analysis.The results show that the reconstruction method based on Kalman filter has higher accuracy when the matching rate is between 30%and 70%.When the matching rate is higher than 70%,the two reconstruction methods have higher accuracy.When the matching rate is less than 30%,the accuracy of the two reconstruction methods is poor.On the basis of reconstruction of vehicle cumulative arrival curve,identification models of lane spillover state and lane blocking state in channelized section are constructed respectively,and case analysis is carried out based on SUMO traffic simulation software.In the aspect of road section lane spillover state identification,the number of vehicles stranded in the road section lane and the headway of vehicles leaving the upstream intersection are selected as identification parameters,and a road section lane spillover state identification model is constructed based on the above parameters.In the aspect of lane blocking state identification in channelized section,it mainly includes:①Estimating the vehicle cumulative departure curve in the lane non-blocking state according to the reconstructed vehicle cumulative arrival curve and traffic signal control timing scheme;② Based on the dynamic time bending distance,periodically comparing the similarity between the estimated departure curve under the condition of lane non-blocking and the measured departure curve;③ Based on the time series segmentation algorithm,the dynamic time bending distance between left-turn lane and adjacent straight-ahead lane is compared to realize the identification of left-turn blocking straight-ahead state,straight-ahead blocking left-turn state and lane non-blocking state.The results show that the method can effectively identify the two states of road section lane spillover and channelized section lane blocking.