Within-day And Day-to-day Dynamic Traffic Assignment Theory And Their Applications In Road Congestion Pricing

Dynamic traffic assignment(DTA)is one of the cornerstones for the development of intelligent transportation systems.It has played a significant role in dynamic traffic control and route guidance.By influencing travelers' behaviors and decision-makings on travel mode,departure time,and route choice,DTA aims at optimizing the network flow distributions,realizing the efficient use of the road network,so as to achieve the purpose of easing traffic congestion.Besides,as one of the economic means of traffic demand management strategies,road congestion pricing has played an important role in alleviating traffic congestion,reducing resource consumption,and improving air quality.Therefore,systematically studying on the DTA and dynamic congestion pricing problem,establishing a concise but physically consistent dynamic traffic flow model to describe dynamic traffic flow propagation process,and implementing a fair and effective congestion pricing scheme,have a significant role in alleviating traffic congestion and improving the happiness of urban commuting.This thesis was supported by the Distinguished Young Scholar Project of the NSFC(71822007),Jiangsu Graduate Research and Innovation Project(KYCX17?0142)and Scientific Research Foundation of Graduate School of Southeast University(YBPY1885).This work studies dynamic traffic demand and travel behavior under dynamic conditions,and investigates DTA models and their solution algorithms.Some congestion pricing schemes are proposed based on the DTA theories,and the corresponding solution algorithms are also designed in this study.The thesis contains the following contents:Firstly,in view of the defect that the link-based cell transmission model(CTM)requires an exogenous proportional parameter to distribute traffic flows at the merging and diverging intersections,this paper establishes a path-based CTM to avoid the exogenous proportional parameter.In the path-based CTM,the distribution of traffic flows is determined endogenously by the upstream and downstream traffic demand.Based on the Newell's cumulative curve approach,this paper establishes a more concise route travel time calculation method;on this basis,a dynamic user equilibrium(DUE)model based on the path-based CTM is established,and the self-adaptive gradient projection(SAGP)algorithm is used to solved the DUE problem.Secondly,while considering the dynamic characteristics of traffic flows,this paper establishes a bi-level programming model for the dynamic congestion pricing problem.Compared with traditional bi-level programming models for the congestion pricing problem,the lower level is a variational inequality for the DUE problem based on the path-based CTM,and the upper level is a joint distance and time delay toll(JDTDT)scheme to optimize the overall system performance.A hybrid SAGP and artificial bee colony(ABC)algorithm is developed to solve the bi-level programming model.In numerical experiments,different toll schemes are compared,including the dynamic JDTDT vs.static JDTDT,dynamic JDTDT vs.dynamic JDTT(joint distance and time toll),JDTDT vs.pure distance-based toll,etc.,which validates the effectiveness of the proposed dynamic JDTDT scheme.Thirdly,a path-based day-to-day dynamic traffic assignment model is established,and a surrogate-based simulation optimization method is proposed to calibrate the parameters in the day-to-day dynamics model.In the surrogate-based simulation optimization,it does not need to calculate the time-consuming nonlinear non-convex objective function,which can significantly save the computational efforts.In addition,by introducing the process of maximizing the expected improvement,it can also approximate the global optimal solution.Based on the license plate recognition data,the parameters in the day-to-day dynamics model are calibrated,which validates the effectiveness of the proposed model and solution algorithm.Finally,considering that traffic flows and costs are changing during the whole planning period,there is no one pricing scheme that has smallest total travel costs for each day.Under each toll scheme,this thesis defines the value of regret for each day,which is equal to the difference between the total travel cost under this toll scheme on this day and the smallest total travel cost under this toll scheme during the entire planning period.On this basis,a robust optimization model minimizing the maximal regret is proposed,which takes into consideration the traffic conditions on each day during the whole planning period.This study not only theoretically deepens people's understanding of the complex transportation network systems,but also provides decision support for traffic flow prediction,route guidance,traffic management and control in practice.