A pattern recognition approach to data fusion in Intelligent Vehicle Highway Systems : Palacharla, Prasad V., Ph.D. University of Illinois at Chicago, 1995. 78 pp.

Abstract

This research addresses two issues--traffic congestion data fusion, and travel time data fusion and estimation. Congestion level data fusion involves fusing data from multiple sources into an estimate of the current congestion level. Such information is needed for Advanced Traveler Information Systems (ATIS) applications which provide travelers with high-level information on the transportation network. Travel time data fusion involves fusing data from multiple sources into an estimate of current link travel times. Such fusion is needed for ATIS applications making use of dynamic route guidance. The current methods for solving the data fusion problem either uses simple selection and aging technique or weighted averaging technique. These techniques have many limitations such as ignorance of data from one or more sources, lack of learning capabilities, and lack of tolerance for uncertainty. This research work targeted towards overcoming some of these limitations using pattern recognition approach. Research results showed that both the Dempster-Shafer theory of evidence and fuzzy four-valued logic provide interesting mechanisms for fusing uncertain traffic congestion data. We have demonstrated the ability of a counterpropagation neural network for travel time data fusion. Our research results show that a method based on both fuzzy logic and neural networks in combination can significantly improve the accuracy of travel time estimation and reduce the time and effort needed to extract the traffic engineering knowledge and devise fuzzy if-then rules.