Research On Signalized Itersection Control Delay Estimation Based On Automatic Number Plate Recognition Data

Real-time,accurate and reliable control delay estimation results are of great significance for mastering the service level of signalized intersection,evaluating the performance of signal control schemes,and providing scientific and efficient congestion solutions.In this paper,the automatic number plate recognition data are used to study the control delay estimation method based on double-loops data and single-loop data.Specific research includes the following aspects:This paper preprocessed the original data at first,then proposed approaches to solve the common questions in research of automatic number plate data,including repeated recognition,and the time difference between license plate system and signal system.Then obtained the travel time of vehicles on the road section and filter abnormal data.The quality of license plate data is analyzed in detail by recognition rate and repetition rate.In this paper,the signal cycle is used as the research time window,and the missing rate is used as an indicator to quantify the degree of data missing.The impact of missing data on travel time and control delay estimation was evaluated.The results show that the missing travel time of unmatched vehicles will affect the accuracy and stability of the travel time and control delay estimation when taking the signal cycle as the time window.Therefore,this paper proposed a method to estimate the travel time of unmatched vehicles based on the least square method.The verification results based on random sampling algorithm show that the method proposed can provide accurate and robust cycle-average travel time estimation results.At the same time,the estimation performance of single vehicle travel time of unmatched vehicles is also not bad.In this paper,on the premise that road delay increases with the increases of traffic flow,and the non-stop travel time varies throughout the day,a non-stop travel time estimation method which taking signal cycle as time window is proposed,and the application and judgment criteria of the method are described.The difference between the estimated travel time and the estimated non-stop travel time is taken as the estimation result of control delay.A simulation model is established to verify the effectiveness of the estimation method of the non-stop travel time and control delay.For the control delay estimation based on single-loop data,this paper refers to the control delay calculation formula provided in HCM2010(Highway Capacity Manual 2010),and divided the 15 direct parameters involved in the formula into three categories: directly obtainable,estimable and unobtainable,and proposes methods to estimate traffic parameters using single-loop data.Since the traffic parameters provided by single-loop data are insufficient,the control delay result calculated by HCM formula does not fit well with real data,this paper uses support vector machine(SVM)to estimate control delay.After inputting the training set,model parameters are adjusted based on K-CV(K-cross validation)method.Finally,the test set is inputted,the analysis results of estimation error show that the error can be reduced stably by using support vector machine compared with HCM formula.