Research And Realization On Dynamic Data Driven Traffic Simulation

With the increasing propulsion of intelligent transportation system (ITS), microscopic trafficsimulation technology has turned into an effective tool to describe traffic behaviors, playing agrowing role in solving traffic problems. Although many achievements on traffic simulation modelestablishment and system development were made, the existing pre-established traffic systems basedon historical data tend to ignoring unexpected emergencies in the process of operation, which poses alow accuracy in real-time conditions. On account of the development of advanced technology, theapplication of traffic simulation based on the real-time data becomes possible. The characteristic ofdynamic data driven application system (DDDAS) is to effectively combine the simulation and realdata, which enables the simulation to dynamically incorporate real data and make adjustments, thusdrawing the simulation results to approaching convergence quicker and more credible. Currently,DDDAS has been widely used in fields with adequate real data like the crisis management,engineering science and disaster forecast.In view of the above background, aimed at applying DDDAS paradigm to microscopicsimulation system named Movsim, a dynamic data driven traffic simulation method that gives realmeasured data feedback to traffic state forecasting is put forward.First of all, based on the analysis ofthe DDDAS paradigm, the logical process of Movsim, vehicle model and road network model, thedynamic data-driven traffic simulation framework is advanced. After that, accompanied by exploringthe combination of key technologies such as parallel processing, the operation mechanism offramework is dissected. Secondly, the traffic simulation model based on particle filter (PF) is built. AsPF provides a solution to state estimation of nonlinear and non-Gaussian system, it is introduced toimplement the data assimilation module. To this end, random noise translationmodel, segmentedvehicle density noise model, data mapping model based on the middleware named JSON, two weightcalculation models based on sliding window and sensor intelligent choice respectively and resamplingmodel is designed. Also, the implementation process based on PF is presented. And then, the system isdeveloped and implemented. Combined with the traffic simulation model, the system is divided intomodules and the main modules such as the data assimilation injection, multithreading, dynamic datamanagement, human-computer interaction are carried out. Finally, combined with established trafficsimulation system, the linear and circular road simulation scenarios are constructed to validate theapplication effect and simulation precision of traffic simulation of dynamic data driven framework.