Data Fusion for ITS: Techniques and Research Needs

Abstract

Intelligent transportation system (ITS) infrastructures contain sensors, data processing, and communication technologies that assist in improving passenger safety, reducing travel time and fuel consumption, and decreasing incident detection time. Multisource data from Bluetooth® and IP-based (cellular and Wi-Fi) communications, global positioning system (GPS) devices, cell phones, probe vehicles, license plate readers, infrastructure-based traffic-flow sensors, and in the future, connected vehicles enable multisource data fusion to be exploited to produce an enhanced interpretation of the monitored or observed situation. This occurs by decreasing the uncertainty present in individual source data. Although demonstrated for more than two decades, data fusion (DF) is still an emergent field as related to day-to-day traffic management operations. Data fusion techniques applied to date include Bayesian inference, Dempster-Shafer evidential reasoning, artificial neural networks, fuzzy logic, and Kalman filtering. This paper provides a survey of ITS DF applications, including ramp metering, pedestrian crossing, automatic incident detection, travel time prediction, adaptive signal control, and crash analysis and prevention, and indicates directions for future research. The encouraging results so far should not conceal the challenges that remain before widespread operational deployment of DF in transportation management occurs.