Traffic Parameters Estimation Of Urban Arterial Roads Under Connected Vehicle Environment

As an important prerequisite for traffic guidance,adaptive signal control and accident detection,traffic parameters estimation for urban roads has always been the focus of research in traffic field.The emergence of connected vehicles has brought new development opportunities to the study of traffic parameters estimation.On the basis of studying the previous research progress,this thesis analyzes the spatio-temporal evolution process of urban arterial road traffic state based on traffic flow theory and dynamic Bayesian network,and proposes a method to estimate traffic parameters by using the connected vehicle trajectory data in the simulated environment.The main work includes the following parts:(1)Based on the traffic flow theory,the spatial-temporal evolution model of urban arterial road traffic state is established.Taking the number of vehicles arriving in queue and the number of vehicles remaining in queue after the green time as two key parameters to define the traffic state,the spatio-temporal evolution process of traffic state in the urban trunk road network with signalized intersections is analyzed based on the traffic flow theory.(2)Based on the dynamic Bayesian network,the probabilistic dependence of traffic state on the time series of urban arterial road links and the probabilistic correlation between traffic state and the trajectory data of connected vehicles are analyzed.Using coupled hidden markov model(CHMM),a special dynamic bayesian network,analysis a posteriori probability distribution of the hidden variables(traffic state)and observed variables(connected vehicles trajectory data)under the given traffic parameters.Hence,the traffic parameters estimation problem is transformed into the typical "learning" problem of CHMM.(3)Expectation maximization(EM)algorithm is used to estimate traffic parameters in dynamic Bayesian networks.EM algorithm is a common method to solve CHMM’s "learning" problem.In order to solve the problem in the E step of EM algorithm,which is hard to obtain the optimal analytical solution of the posterior probability distribution of traffic state based on the observed values and current parameters,particle filtering algorithm is introduced to represent the traffic state instances in the network with weighted particles,and Monte Carlo simulation is used to approximate the optimal Bayesian estimation.(4)VISSIM simulation and method verification.Using VISSIM micro traffic simulation software,a simulation model of urban arterial road network with two signalized intersections is built.The vehicle trajectory data output by the simulation was taken as the connected vehicles trajectory data in the simulated Internet of Vehicles environment.The proposed method was used to estimate the traffic parameters of each link,and then compared with the real traffic parameters in the VISSIM simulation,the MAPE estimated by each parameter estimation result is less than 0.22,which verifies the effectiveness and accuracy of the method.At the same time,the traffic parameter estimation results based on the trajectory data of connected vehicles under different permeability are also tested,and other sensitive factors affecting the parameters estimation are analyzed.