A Recursive-bayesian Inference Fusion Model For Dynamic Link Travel Time Estimation

Travel time is a significant index which could reflect the traffic condition. As the important information basis of Advanced Traveler Information System (ATIS), link travel time estimation based on real-time data has become a key issue in ITS domain. Probe vehicles and loop detectors received great attentions in link travel time estimation area, which are widely used as urban traffic sensors. However, it is difficult for individual sensor to surveille traffic conditions both in time and space comprehensively. Data fusion is a method which could utilize the complementarity between probe vehicle and loop detector data and improve the estimation accuracy of link travel time. Such an improvement is significant to the operation of ATIS.During the research, the relevant conception of detected data and data fusion were clarified at first. Considering the detected data, the conceptual framework of probe vehicles and loop detectors, detected data feature and the application of detected data in link travel time estimation were emphasized on. For data fusion, the application of this technique in transportation domain, especially in link travel time area, were also key points in this session.Moreover, link travel time was estimated by using the methods based on probe vehicle data and loop detector data respectively. Then, according to the principal of Bayesian Fusion, the convergent conditions and substitution rules were introduced into the original Bayesian Fusion Model which could realize the recursive processing of Bayesian fusion.Furthermore, in order to testify the Recursive Bayesian Estimation Fusion Model, VIS SIM, which is a widely used microscopic traffic simulation software, were used to build a virtual road network and an actual road network refering to the road network of urban center area in Dalian. The research optimized the original setup of detectors according to the evaluation files of simulation at first and estimated the link travel time by using the Recursive Bayesian Estimation Fusion Model.Estimation results showed that the proposed model could improve the accuracy of estimation results and the accuracy could keep stable in different traffic flows.