Study On Dynamic Traffic Assignment Of Traffic Network And The Related Problems

This dissertation study around the dynamic traffic assignment of traffic network - the core theories of traffic induce of Intelligent Transportation System. Genetic Algorithm (GA) and improved GA are introduced into resolution of differential game model of dynamic traffic network equilibrium, and the suitability of the model is significantly improved.The background of this study is stated in the first chapter, the urban traffic condition and ITS study condition were analyzed, and the study importance, study object and the main content were put forward. In the second chapter, the development of urban traffic dynamic assignment models were analyzed and evaluated, and the advantages and disadvantages were concluded.In the third chapter, genetic algorithm was used to solve the differential game model of dynamic traffic network equilibrium under open loop information structure which put forward by Byung-Wook Wie. The computing results of the numerical example are compared with which using Pontryagin-type minimum principle.Under open loop information structure, the traffic state of the arc at time t is unknown to the players, and they make blind decisions. Therefore, the aggregate traffic on arc increased with time, this is not coherent with the traffic reality. Dynamic traffic assignment problem based on ATIS was put forward; the differential game model of dynamic traffic network equilibrium under close loop information structure was formulated.The influence of road capacity on assignment policy was considered in the formulation of dynamic traffic assignment model with time window restriction. The parallel GA with population immigration was introduced in this chapter.Markov analysis method is used to construct an isolate intersection traffic situation prediction model to estimate accurately what the forthcoming traffic conditions of the multi-phase intersection may be. The suitability and operability of this model was verified.Considering the limitation of the time sequential prediction model, Support Vector Machines (SVM) was adopted in the traffic status prediction model with considering time and space simultaneously. Finally the paper concludes with the main research fruits and a discussion on possible future extension of the study.