Queue Prediction Of Heterogeneous Traffic Flow At Intersection Under Video Detector

In today's urban road network in China,vehicle congest ion has beco me an important traffic problem.The inconvenience it brings has aroused widespread concern in society.Traffic congest ion is t ypically characterized by the long queue of vehicles at intersect ions.Therefore,vehicle queuing problem has beco me an important branch of traffic flow theory and an important direct ion of so lving urban traffic congest ion problem.The essence o f vehicle congest ion is vehicle delay,which manifests as vehicle queuing.In or der to study the problem of congestion,it is necessary to clarify the problem of vehicle queuing.With the development of intelligent transportation techno logy,more and more attent ion has been paid to the driving characterist ics of buses.Vehicle detection technology,dynamic short-term traffic forecast ing,real-t ime queuing and other aspects should have a new development.Therefore,this paper will take real-t ime queued vehicles as the research object,and do some research on upstream intersect ion discre t ization,Kalman filter predict ion and the format ion mechanism of intersect ion queuing.Firstly,the definit ion o f discrete phenomena generat ion and fitt ing and vehicle queuing are introduced.The format ion process of queuing at signalized intersections,queuing characteristics at supersaturated intersect ions and their basic parameters are analyzed.On the basis o f exist ing research,the theoret ical basis of queuing research is obtained.Secondly,through the development and applicat ion of existing video de tectors,using the informat ion obtained fro m the upstream and downstream intersect ion detectors for in-depth use,the upstream intersection vehicle departure is depicted in each t ime element.In order to accurately describe the heterogeneous traffic flow dispersion,the key is to accurately predict the vehicle travel t ime,considering the difference of different vehicle types travel t ime,Kalman filter is used to calculate the travel time.The truncated bimodal normal distribut ion predict ion model is obtain ed by t ime predict ion,which can accurately predict the arrival flow rate of vehicles at downstream intersect ions.Then,start ing fro m the queuing mechanism of downstream signalized intersections,aiming at the real-t ime prediction of the number o f queued vehicles,the necessit y and method o f real-time predict ion o f lane-by-lane queuing are proposed.Through the arrival o f vehicles at downstream intersect ions and the truncation effect of signalized lights on arrival,the unsaturated and saturated states are discussed separately by using the basic theory o f traffic waves,and the backward extensio n of queuing waves and the start ing waves are discussed.Start,dissipate wave leave,get real-t ime queuing prediction of intersect ion.Finally,using the north entr ance road of Huancheng South Road and Xinwe n Road intersect ion in Kunming Cit y,Yunnan Province as the verificat ion object,the established queuing predict ion method is verified by the exist ing video detectors.The queuing results obtained by the actual detectors are co mpared and analyzed,the predict ion error is tested,and the maximum queuing length error and the sensit ivit y of the model are verified,which verifies the research results.Effect iveness of results.The queuing predict ion method in this pape r can provide a new theoret ical basis for intelligent traffic control and provide effect ive pract ical guidance for road users.