Estimation Of Traffic State For Highway Basic Sections With Uncertain Methods

Accurate estimation of traffic state is the key to grasp the traffic condition of highway. Traffic condition can be reflected indirectly and partly with single traffic flow parameter, and the clustering analysis of multiple traffic parameters is a typical method to estimate the state for site, however, clustering results is sensitive to the sample size. On the other hand, studies on the estimation of road traffic state usually consider the travel time and velocity as input, with the limitation of data acquisition methods, there’s an uncertainty to some degree. In order to improve the effect in the application of the traffic state estimation system for highway, it makes sense to do some research for these problems.The research object is based on the basic highway section in this paper, using uncertain methods to estimate the traffic state of site and road of highway basic sections. When estimating the site traffic state, taking the spatial distribution of disequilibrium analysis of the sample point as a breakthrough. Focus on the influence of the imbalance of the spatial distribution of sample points in the traffic state estimation; when estimating the road traffic state, using the method of multi-source data fusion to solve the uncertainty problem of traffic state estimation. The main research contents include:① The traffic flow parameters characteristic analysis of highway basic sections. Firstly, analysing the temporal correlation and the imbalance of the spatial distribution of sample points of site traffic flow parameters; and then, analysing the uncertainty of using traffic flow parameters to recognize the traffic state, this will ground for the modeling of site and road traffic estimation afterwards.② Modeling of the site traffic state estimation based on GEFCM with characteristic parameter weighting. According to the defect of tradictional fuzzy clustering in traffic state estimation, combined with the imbalance of the spatial distribution of sample points and the difference of different feature parameters for clustering influence weight, building the General equalization fuzzy C means clustering with characteristic parameter weighting, by using the principal component analysis(PCA) to determine the weights of different characteristic parameters. The result shows that this method has better adaptability and reliability.③ Modeling of the road section traffic state estimation model of multi-data fusion based on DBN. For overcoming the uncertainty of using traffic flow parameters to recognize the traffic state, introducing the DBN, based on the selecting relative density, average travel time and traffic state as the node variables, the topological structure of network is determined, at last building the dynamic bayesian networks for traffic state estimation. The result shows that this method has better reliability.Lastly, introducing the design and implement of the system, and appling to the site and road traffic state estimation of Yuwu Highway. The results show that comparing with the contrast method, both of the site and road traffic state estimation methods have a higher IR and a lower FIR.