Data-driven Road Network Performance Evaluation Models And Approaches For Road Network Oriented Toward Traffic Demand Management Strategies

Traffic congestion nowadays has been addressed by Traffic Demand Management(TDM),which limit the traffic demand based on actual supply of traffic capacity instead of increasing supply to accommodate demand.Since TDM has a great impact on the society and economy,TDM requires a scientific argumentation and quantitative assessment.The road network performance evaluation model and method for TDM relies on the traditional four-step traffic model for a long time,which is difficult to reflect the dynamic time-varying characteristics of traffic operation.The cumulative effect of traffic congestion step by step over time cannot be reflected.Therefore,the precision of the simulation result cannot be ensured.Furthermore,the modeling work is complicated and needs much time cost.Another method is dynamic traffic model.It needs modeling based on the detailed road sections,which is difficult to meet the requirements of large-scale network.This dissertation addresses problems and limitations of existing models and methods and aims to develop data-driven models which can output quantified effect promptly and reliable.The relationship between factors related with traffic demand and the road network performance indices has been found based on big data.Furthermore,the dynamic iterative recursion model has been developed to estimate the effect of different traffic demand management strategies.Finally,the models and methods have been verified throughreal world case studies.The main research of the dissertation is as follows.(1)Based on the classification of TDM strategies,a multi-dimensional modeling framework for different types of TDM strategies are formulated.For the five TDM strategies including total trip amounts control,departure time control,trips spatial distribution control,trip mode change,and road network capacity constraints,an impact analysis is carried out on demand element such as OD amounts,OD distribution,departure time distribution,and traffic assignment.Three model methods are therefore identified to start modeling respectively by distinguishing the one that can be directly solved by the data,and the ones that can be solved by data coupling with other traffic models.(2)The first modeling approach is a data-driven model for TDM strategies of trip amounts control and departure time control.For the control of trip amount and departure time,assuming that other factors such as traffic assignment are unchanged,this dissertation proposes a recursive evaluation model for road network performance based on two-level demand separation.The model divides the travel demand into two levels including the basic demand of free-flow and the actual demand affected by congestion.The basic demand is used to adjust the total demand and departure time.The actual demand is used for the recursion of the road network operation step by step over time.The recursive process uses the relation function of the on line vehicles as actual demand and the average speed of the road network as a conduction function to come up with an estimation for 24-hour network operation status under different demand patterns and different TDM strategies.The model is relatively simple and can reflect the dynamics of traffic operation as well as the cumulative impact of congestion over time.It can quickly and accurately evaluate TDM strategies effect such as driving restriction and off-peak commuting.After the actual data verification,the model has an accuracy of more than 90% for the average speed of the road network.(3)The second model is the modules of model-driven and data-driven approaches for TDM strategies related with trips spatial control.The trips spatial control represented by Job-housing balance and Congestion Charging has an impact on OD distribution and traffic assignment.A data-driven method is not adequate,which needs to be integrated with model-driven method.In this dissertation,the big data for multi-source traffic is integrated into the four-step traffic model.A traffic demand estimation method based on the computational graphs neural network learning framework is proposed.An evaluation model for TDM effect is developed based on computational graphs.The model minimizes travel demand estimation errors through forward and backward propagation,and then performs TDM cost-benefit analysis.(4)The third modeling method is an estimation based on the data correlation of the percentage of car trips and road network capacity.The method is based on the data-driven model,combines data correlation to solve limit values or boundary values.The percentage of car trips under the optimal efficiency of multiple trip modes and the road network capacity under the boundary speed of the road network are obtained,which solves the problem of TDM constraint target and boundary.(5)Based on mobile phone data and car trips data,this dissertation proposes a trip characteristics extraction method for TDM analysis model,including trip OD,departure time,travel distance,time cost,travel route,etc.The data-driven management can therefore be provided with good data conditions.Based on the car trip data,a pattern identification method based on clustering analysis is proposed,which lays a foundation for the impact of TDM model under different traffic demand patterns.(6)Taking Beijing as an example,three model methods are applied to real work case studies in the dissertation.The effects of different TDM strategies under the application scenarios such as license plate control,tail number restricted driving,off-peak commuting,and job-housing balance.The reliability of the model is verified by comparing the simulation results of motor vehicle license plate tail odd-even rationing policy with the floating car data monitoring results.The road network capacity and the optimal car trip percentage in the central urban area of Beijing are also estimated.