TravelTimeEstimationfinal智能交通世界大会ITS智慧城市社区人工智能AI物联网IT报告课件教案.pdfVIP

A Bayesian Approach for Travel Time Data Fusion Zhang (Gabriel) Li*, Yan Xu, Chen Cai, Fang Chen s.will.william@, +61 4 0491 4694 National ICT of Australia, 13 Garden Street Eveleigh NSW 2015 Australia Abstract Probes with GPS devices reveal useful information for traffic conditions, but the high level of noises makes it challenging to estimate travel time from the data collected. This paper proposes a new fusion model for estimating links/routes travel time based on GPS and flow data. The focus on the study is to fully utilise the temporal data from GPS probes and the flow data from loop detectors, and demonstrate the feasibility and potential for large-scale traffic condition estimation. The model calculated the posterior probability and the expectation of travel time given the prior knowledge from GPS and flow modeled travel time respectively. The result is evaluated for different types of roads across Sydney area against travel time provided by Google Map. The approach provides capabilities to fuse broadly available data sources for the compensation of low-quality GPS data regarding traffic network condition estimation. Keywords: Travel time estimation, GPS, data fusion. 1. Introduction. Travel time estimation is an important subject for traffic planners and operators to identify key bottlenecks (both online and offline) in the traffic network for improvement. It also benefits road users with increased awareness of the traffic conditions and allows better route choice. However, the travel time of any link at a given period is the result of a complex interaction between the traffic demand (e.g. population characteristics and road users’ decision making) and the traff