Estimation And Prediction Of Link Travel Time For Urban Trunk And Secondary Streets

Travel time is the most important, the most widely concerned, and can bestreflect the traffic status. To compare with the spot traffic parameters, it can betterbe used as indication of the overall road system performance, real-time measure ofcongestion. Travelers, traffic authorities and researchers often use it to evaluate thetraffic situation or be a basis for traffic decision-making.Not only it is very important to get real time travel time, but also it is also veryimportant to predict the future travel time. Traffic demand in future short timetrend determines the change of traffic status, therefore, link travel time predictioncan facilitate traffic administrators and travelers to manage better decision-makingand travel decision-making. Predicting the travel time is more meaningful thanestimating the current information, it can predict the development of travel timechanges during the future period, and can speculate the traffic state, then canprovide a basis for traffic decision-making..Estimation and prediction of link travel time for urban trunk and secondarystreets is more complex than estimation and prediction of link travel time for thecontinuous traffic flow facilities. Combined with the project of the NationalNatural Science Fund, Estimation and prediction methods of link travel time forurban trunk and secondary streets are studied. The main research work includes thefollowing aspects:(1)The characteristic analysis of link travel time for urban trunk andsecondary streets and design of experimental environment Compared to continuous traffic flow facilities, link travel time characteristicof urban trunk and secondary streets must be different, because of the signalcontrol. When the loop detectors are laid on the different position, the correlationbetween the spot traffic parameter data of being collected and link travel time is notonly more difficult, but also will be different. Aiming at these problems, on thebasis of the correlation analysis between the spot traffic parameter data of beingcollected and link travel time, the concept of fuzzy critical interval and theacademic thinking of designing respectively the link travel time estimation andprediction methods under the different traffic state is put forward, and on the basisof mainstream professional simulation system the overall experimentalenvironment about estimation and prediction of link travel time for urban trunk andsecondary streets is built.(2) Study on link travel time estimation method based on the spot trafficparameter dataIn order to further improve link travel time estimation effect based on SCATSsignal control system detector data, on the basis of the spot speed data, a kind ofestimation model of travel time based on modulus coefficient according to differenttraffic conditions is put forward, which effectively improves estimation effect ofsingle model, and has higher estimation accuracy under the congestion. Aiming atSCOOT signal control system detector data features and the problem that BPRmodel can not maintain good estimation effect in different traffic conditions, theuniform-state accumulative volume modified BPR function and the division-stateaccumulative volume modified BPR function are put forward, they cansignificantly improve the estimation accuracy of average travel time under theheavy traffic state.(3) Study on link travel time estimation method based on GPS floating cardataAiming at the problem that link travel time estimation effect of single vehiclebased on GPS data can be improved, speed-time numerical integral correction model based on GPS instantaneous speed data is proposed, in order to improvetravel time estimation results of individual floating car. Aiming at the samplevolume's adequacy or not, the link travel time estimation method for a given stateis put forward. The results show that, the proposed method can further reduce linktravel time estimation error based on GPS data.(4) Study on link travel time fusion estimation method based on multi-sourceIn order to make adequately use of the data collected by traffic dector oftraffic signal control system and the GPS floating cars, a link travel time estimationmethod which combines local integration and overall fusion for a given state is putforward, and its validity is confirmed through the comparative analysis.(5) Study on link travel time prediction method based on data fusionLink travel time prediction is very important to help administrators andtravelers carry on the management decision and travel decision, in order to furtherimprove link travel time prediction effect, a kind of link travel time predictionmethod based on the distinction traffic state ideology and data fusion technology isproposed. The basic models for fusion are the lateral prediction model based ongray theory and the longitudinal prediction model based on Genetic algorithm withvariable input gray BP neural network. The results show that, the proposed newmethod has a comparative advantage to reduce prediction error.The research content, method and conclusion related by the paper questpositively for the estimation and prediction of link travel time for urban main trunkroad and secondary trunk road, and can provide certain theoretical basis andreference works for traffic dynamic management.