Research On Composite Model Combining Multimode Split And Traffic Assignment Based On Generalized Travel Cost

With the improvement on level of Chinese urbanization and mechanization,urban traffic congestion gets more and more serious. Traffic congestion and theresulted accidents and environmental pollution have affected the efficiency andquality of urban life. In terms of congestion management strategies, most countrieshave turned mainly to traffic demand management instead of increasing traffic supply.But no matter what strategy is, before implementing, predicting the effect andassessing the feasibility of strategy are essential. In this process, two importantindicators of strategy effectiveness assessment are predicted mode split rate and trafficnetwork flow. The traditional four-step approach, which separates mode split andtraffic assignment into two separate phases and weakens the close link andinterdependence between them, is obvious unreasonable and reduces the accuracy offorecasts. Thus it is necessary to study the composite model combining multimodesplit and traffic assignment in order to improve the prediction accuracy, providetheoretical support for traffic behaviour analysis and research and provide a basis forevaluating the traffic planning and management strategies.To achieve this purpose, firstly analyse three factors which affect residents’travelchoice (namely residents’ characteristics, travel characteristics and modecharacteristics) one by one using the survey data, classify them according to thedifferent mechanisms of action and select the main factors by correlation analysis.Then, propose the construction method of urban traffic super network, describe theattribute values for different sub-network layers. On this basis, the travel cost of eachtransport mode is divided into three aspects of travel time, cost and comfort. Theexpression of each mode generalized travel cost is discussed separately. Next, thecomposite model’s basic assumptions, independent variables, dependent variables andconstraints are given. The equilibrium conditions of the model are achieves by Logitstochastic equilibrium theory. A variational inequality model is built, whose equivalency and existence of solution are proved. The concrete solving steps are alsogiven. Last, we conducted a case study with G County, Shandong Province, describedthe concrete application methods of model and analysed the advantages anddisadvantages of the model by comparing prediction results of composite model,prediction results of traditional four-step approach with revealed preference data.Results show that in respect of transport modal split, the accumulative predictionerror of composite model is13.2%lower than that of Logit model, in respect of roadsection traffic flow prediction, the prediction error of composite model is15.32%,which is6.35%lower than that of stochastic user equilibrium model, in respect of busstation section passenger volume, the prediction error of composite model is18.05%,which is4.36%lower than that of stochastic user equilibrium model. The predictionaccuracy of composite model has a certain degree of improvement in all of the threeaspects, which means the model has a good application value.