Urban Network-wide Traffic Capacity Estimation And Traffic State Identification Using Multi-source Data

Optimize the temporal and spatial distribution of traffic demand through accurately characterizing the traffic state and traffic capacity of urban road network is the main issue to alleviate traffic congestion.As a road network-level traffic flow model,Macroscopic Fundamental Diagram（MFD）provides a new way to identify network traffic state and estimate regional traffic capacity based on real traffic flow data.Related studies based on MFD lack consideration of the evolution characteristics of the traffic state and dynamically changes of multiple influencing factors.In view of this,this paper intends to conduct traffic state identification and traffic capacity estimation for road network based on MFD.The paper consists four parts:construct MFD with multi-source data considering the consistency between parameters,analyze the influencing factors of MFD construction for complex scenes of real data,and identify the road network traffic state considering the continuous evolution of traffic flow and estimate the road network traffic capacity based on data-driven method,the specific research work is summarized as follows.First,for MFD construction,related studies uses different source of data to obtain the parameter of MFD separately,and there is a problem of inconsistent between the two parameters of MFD.To this end,this article proposes a MFD construction method using multi-source data based on DS-evidence theory considering data reliability.Considering the data availability for urban road traffic,vehicle trajectory data represented by floating car data and cross-sectional data represented by vehicle detectors are selected in this paper.For vehicle trajectory data,a MFD construction method is proposed that takes into account the difference in permeability of floating cars with different origins and destinations（OD）supplemented by cross-sectional data.Based on this,the parameters of MFD are estimated using vehicle trajectory data and cross-sectional data,respectively.A fusion method for MFD construction is proposed based on DS-evidence theory using multi-source data through quantifying the reliability of estimated MFD parameter from different data sources.The results of case study under real data and simulated data show that compared to the trajectory data based and cross-sectional data based MFD construction,the accuracy of the fused MFD construction is improved,the maximum estimation error is reduced by 22.3%.Second,as most studies of factors that affect MFD construction focus on single source of data or single influencing factor,a systematic analysis method of comprehensive influencing factors of MFD construction accuracy is designed based on orthogonal experiment method.Considering the data characteristics for urban road traffic in china,the penetration rate of floating cars,spatial distribution equilibrium degree of floating cars,coverage rate of fixed vehicle detectors,spatial distribution equilibrium degree of fixed vehicle detectors,recognition rate of Automatic vehicle license plate equipment（ALPR）equipment,traffic demand level were considered comprehensively.Firstly,Consider the independent effects of each factor,the orthogonal table of L₃₆（2¹3³6³）is adopted,and the influence degree of each factor is determined by range analysis to determine the key factors;Secondly,considering the interaction among the key factors,the orthogonal table L₂₇(3¹³)was adopted,and the significance of each factor was further quantified by F test.The following is the result of case study under simulated data for real road network show that the sorted factors by degree of influence.The equilibrium of spatial distribution of floating cars,the coverage rate of fixed vehicle detectors,the penetration rate of floating cars,the level of traffic demand,the recognition rate of ALPR equipment,and the spatial distribution equilibrium degree of fixed vehicle detectors.In the case the coverage rate of fixed vehicle detectors or the penetration rate of floating cars reached 45%and 20%,with the increase of coverage rate and penetration rate,the accuracy of MFD construction can be ensured.When the coverage rate of the fixed vehicle detectors is relatively low and the penetration rate of floating cars is relatively high,the interaction between the factors can be manifested significantly.Third,the traffic state of road network is mostly classified by aggregated traffic flow parameters in temporal level,and insufficient attention is given to the temporal changing trend of traffic state.This paper proposes a network-level traffic transition state identification method that considers the continuous evolution process of traffic flow.Based on this,the traffic state classification and static traffic capacity estimation for road network are realized.And the influencing factors of road network traffic capacity was quantitatively analyzed.Road network traffic state present different evolution characteristics（such as flow-density monotone increasing,continuous fluctuations within the range of a state,etc.）in a day,which corresponds to the different mutation of traffic state.Therefore,first,the sub-time series before and after a network traffic flow-density point in MFD of a certain period of time under a certain length of sliding time window is extracted to describe the evolution process of the road network traffic flow pattern.Second,the degree of similarity between the two sub-time series is quantified based on the dynamic time wrapping algorithm.Third,the Brent’s method is used to extract the extreme value of the fitted distance to identify critical transition point（CTS）of the road network state.Fourth,the Gaussian mixture model is used to identify different types of CTSs,then the traffic state identification and network traffic capacity estimation were realized with CTSs.Finally,the factors affecting the traffic capacity of the road network are analyzed based on the orthogonal test method.The results of case study under simulated data of real road network show that the CTSs can effectively capture the sudden changes of road network state evolution characteristics.Based on this,it can distinguish traffic state into three classifications（under-saturated,saturated,and over-saturate）and estimate static traffic capacity reasonably.The case study of the influencing factors of traffic capacity of road network show that the influencing factors are in order of traffic demand distribution,signal timing plan,road network structure,and regional traffic guidance strategy.Final,studies on capacity estimation of network traffic focuses on the static capacity of road network,which did not consider dynamic factors that affect capacity.A data-driven method for traffic capacity’s dynamic estimation and a traffic state identification model considering dynamic traffic capacity for road network were proposed.It is assumed that the MFD curve is a concave function,and the maximum flow corresponds to the road network capacity.The dynamic traffic capacity of the road network is estimated by adaptive Kalman filter method based on the static traffic capacity and the average traffic flow density interval of the road network in the saturated state.Considering the dynamic changing characteristics of traffic capacity,the traffic state of the road network is identified at real-time based on the hidden Markov model.The results of case study under real data and simulated data show that the traffic capacity near the morning and evening rush hours has increased in different ranges.The road network traffic operation state identification method considering the dynamic traffic capacity can effectively identify the continuous evolution and change of the traffic state.The proposed method can achieve a reasonable and stable estimation of the road network dynamic traffic capacity and an accurate judgment of the traffic state of the road network.