| The rapid growth of private car ownership in cities has made road traffic inadequate to meet the needs of commuters,leading to increasing traffic congestion.Traffic congestion is gradually evolving from road sections to regional congestion.In this regard,the classification of heterogeneous road networks can provide an important basis for decision making by relevant authorities.In this paper,we discuss the dynamic partitioning algorithm and the traffic state classification method based on the traffic operation characteristics and Normalized Cut(NCut)algorithm respectively,as follows.Aiming at the study of urban road network partitioning algorithm,this paper takes into account the spatial and temporal characteristics of traffic flow,and constructs a comprehensive measure of traffic operation characteristics.Based on the theory of NCut algorithm,this paper proposes a static partitioning algorithm for road networks,which solves the boundary error problem by adaptively adjusting the boundary.Considering the shortage of the static partitioning algorithm in application,the dynamic partitioning algorithm of the road network is proposed by fixing the stable blocks of the static algorithm and dividing the road sections with low homogeneity time by time.The results show that the proposed algorithm can effectively partition heterogeneous areas.The proposed algorithm can reflect the spatial and temporal propagation characteristics of the traffic flow by investigating the evolution of the dynamic partitioning of the road network from “morning peak-flat peak-evening peak” and evaluating the evolution with the Macroscopic Fundamental Diagram(MFD).In addition,all sub-regions form a stable shape of the Macroscopic Fundamental Diagram,which indicates that the sub-regions are homogeneous and the proposed algorithm achieves effective and desirable partitioning results.Aiming at a traffic state classification model,this paper first incorporates speed variance heterogeneity into the basic traffic diagram model based on a traffic flow theory model.By establishing the speed variance function,the model is able to include a wide range of information on observable and identifiable variables.Secondly,a two-stage hierarchical Bayesian inference method is used to estimate the model parameters,and the proposed model is tested and compared with the observed data of Guiyang city.Finally,the variation point parameters of the traffic state are estimated from the road sections using the unused traffic flow model,and the traffic state and congestion of the sub-regions of the road network are evaluated from a macroscopic perspective.The results show that the proposed model framework has a good performance in terms of fit and prediction accuracy.The traffic state model incorporating random effects has more stable estimation of the variation point parameters and is more consistent with the actual road traffic state scenario.This paper implements a dynamic classification algorithm for the road network.The experimental results show that the proposed algorithm can achieve effective and ideal classification results.The traffic state classification model is further proposed to discriminate the traffic state of intersections,road sections and road network sub-regions,and to assess the capacity of road network sub-regions,which provides some theoretical support and reference basis for traffic management departments. |
