Urban Road Network Traffic State Perception Based On Trajectory Data Fusion

The timely and accurate acquisition of urban road traffic state is the basis for evaluating the road network state,analyzing the operation efficiency of urban road network,and designing refined traffic control strategies.It is also an important part of intelligent transportation system.The existing research on road network state is mostly based on the data collected by fixed detectors such as loop detector and microwave detector.Fixed detector has the disadvantages as high installation and maintenance costs and limited road network coverage,which can not provide traffic state estimation for the whole road network.In recent years,with the wide application of GPS and the rise of LBS(location-based service),the trajectory data as floating car data with abundant information and wide coverage is widely used for urban road traffic state estimation,which fills the gap of traditional detectors.However,due to the limitation of data storage and transmission,the sampling frequency and penetration rate of trajectory data is low.The low frequency and sparse characteristics cannot guarantee the accuracy and stability of the estimated urban road network state.To solve the above problems,this thesis proposes to study the urban road network operation state by fusing different types of trajectory data.The main innovative work of this dissertation is embodied as follows:In order to better clarify the applicability of trajectory data in different application scenarios,the data quality of different trajectory data is evaluated,and five indicators including positioning accuracy,sampling frequency,coverage,completeness and range validity are proposed to evaluate the trajectory quality.Aiming at the problem of insufficient accuracy and stability caused by the low-frequency characteristics of trajectory data,a vehicle trajectory reconstruction algorithm for signalized intersection based on low-frequency trajectory data is proposed.By resampling high-precision trajectory data,it is verified that the proposed method could reconstruct the vehicle's trajectory with sparse sampling points,which could enhance the availability of trajectory data.Different types trajectory data fusion could improve the penetration rate of the trajectory data.Traffic kinematic wave estimation and link speed estimation methods are designed fusing multi-source trajectory data,which overcomes the limitations of the trajectory data application and accurately estimates link speed and control delay of the intersection.A multi-source data fusion model is proposed for control delay estimation based on the traffic kinematic wave.The method solves the problem of control delay estimation under the condition of sparse trajectory data.Firstly,the multi-source trajectories are reconstructed by the proposed method,and then the locations where vehicles join the queue and leave the queue are extracted.According to the traffic kinematic wave theory,the positions of vehicles joining the queue are all on the aggregation wave,and the positions of vehicles leaving the queue are on the dispersive wave.By using the optimization theory and the least square method,the aggregation wave and the dissipated wave are modeled respectively,and the estimation of the vehicle queue profile at the intersection is completed.According to the relationship between the control delay experienced by the vehicle and the arrival time of the vehicle and the traffic wave parameters,the control delay value of each cycle can be obtained.Through VISSIM simulation experiment,the performance of the proposed algorithm under different sampling frequencies,different trajectory penetration rates,different saturations,and different proportions of multi-source trajectory are tested.The results show that the proposed method could provide accurate control delay when the number of trajectories in above 3.The field experiments also verify this conclusion.The estimation model of intersection control delay is designed based on probability graph model.The model solves the problem of trajectory uneven distribution in urban road network and realizes the comprehensive and accurate speed perception for regional road network.Based on the historical trajectory data,Gaussian process model is constructed as a priori,and the relation between travel time distribution and road speed is investigated by using taxi trajectory data in passenger state,and the relationship between travel time distribution and link speed is constructed by using empty taxi trajectory data,bus trajectory data and bus IC card data.The variational parameters and model parameters are estimated.The multi-source trajectory data of Harbin city are used to evaluate the proposed algorithm.It is found that among the three kinds of trajectory data,the order of the impact on the final speed prediction accuracy is the passenger state trajectory data,the empty state trajectory data,and the bus trajectory data.When the proportion of bus trajectory is 10%,the proportion of empty taxi trajectory data is 20%,and the taxi trajectory data of passenger state is 30%,the probability graph model could accurately predict the speed.Based on control delay and link speed data,an urban road network operation state perception method based on dictionary compression theory is proposed.This method not only uses the traffic parameter,but also adds the road level classes.The main evaluation indicators include spatial anomaly degree and temporal anomaly degree.The degree of abnormality can provide a more refined and comprehensive evaluation of the state of the urban road network.This method is suitable for the evaluation of link,intersection and regional traffic status.The thesis conducts exploratory research on data fusion in the context of big data,and builds a complete urban road network operating state perception framework with different trajectory data fusion methods.The framework contains the processing of trajectory data,including the trajectory data quality evaluation,map matching,low-frequency trajectory reconstruction,the traffic state parameters estimation such as link speed and control delay and urban road network state perception based on dictionary compression theory.This thesis performs multi-angle and multi-level perception of the urban road network operation status from the three perspectives of urban road network intersection delay,urban road network segment speed,and spatio-temporal anomaly degree of urban road network operation status.The results of the thesis solve the problems existing in the current application of trajectory data,and can use sparse,low-frequency trajectory data to provide a low-cost,high-temporal-accuracy urban road network status perception.The framework constructed in this thesis is open,and both loop detector data and video data can be added to the current framework.The research in this paper enriches the current theoretical system of traffic big data fusion research,and provides a framework for urban road network operation status perception based on the fusing different trajectory data and has the potential to integrate multiple types of traffic detection data.