Study On Traffic State Estimation Of Highway Basic Sections Based On Temporal-spatial Analysis

Real-time monitoring of traffic state is the basis of ensuring the traffic safety andtravel efficiency. But now, constrained by the network size, detection technology, costfactors and other objective conditions, the high-density layout of vehicle detectors on allsections of road network cannot be guaranteed. Therefor, how to use the limited testingequipment to realize highway traffic state estimation efficiently and accurately. That hasimportant theoretical and practical significance for improving monitoring efficiency, themanagement and service level of highway.At present, the reseach methods are maily through threshold comparison orabsolute standards, ignoring the intuitive sense of the traffic participants under differenttemporal-spatial background. So, this paper chooses the highway basic sections as thereseach object. After analyzing the traffic flow distribution characteristics in time andspace, a characterization getting method of site and road traffic flow characteristicparameters from fixed detectors is proposed, and based on that, the methods of trafficstate fuzzy estimation, deviation estimation and temporal-spatial correlation estimationof highway basic sections are proposed. The main contents include:①The meaning of traffic state fuzzy estimation, deviation estimation andtemporal-spatial correlation estimation of highway basic sections, and its relativemetrics, classification are described. Meanwhile, after analyzing the temporal-spatialdistribution characteristics of traffic flow, the site and road traffic flow characterizationparameters are got from fixed detectors.②Considering the shortcomings of traditional traffic state absolute estimationstandards, a relative estimation standard based on week-similarity of time series oftraffic flow characterization parameters under certain space is proposed. By using thefuzzy clustering (FCM), the traffic state fuzzy estimation of highway basic sections isrealized, showing the matching degree between the real-time traffic flow parameters andthe recent historical traffic patterns based on week-similarity. The stability of theproposed method is verified through an example.③From the feeling of traffic participants, based on the week-trend of time seriesof traffic flow characterization parameters under certain temporal-spatial conditions, abase-pattern of traffic flow based on wavelet analysis is proposed, the deviation degreeof traffic state based on the theory of Z score is quantified, and then a traffic state deciation estimation method on the temporal-spatial analysis is put forward, showingthe extent of real-time traffic flow parameters deviate from the week-trend. Theeffectiveness of the proposed method is verified through an example.④Considering the propagation characteristic of the traffic flow in lengthwayspace, a change point recognition method based on vector similarity and a patterndivision method of traffic flow time series on change point are proposed, and thenthrough the correlation analysis of traffic state under diffenrent patterns, thetemporal-spatial correlation of traffic state is got. The feasibility of the proposed methodis verified through an example.Synthesize all of above, this paper has developed a new way of highway basicsections traffic state estimation based on temporal-spatial analysis. The experimentresults show that this research considers the intuitive feeling of traffic participants underdifferent temporal-spatial background, methods are stable, effective and feasible. Andthe results have great value of application.