Improving Travel Time Estimates from Inductive Loop and Toll Collection Data with Dempster–Shafer Data Fusion

Dempster–Shafer data fusion can enhance travel time estimation for motorists and traffic managers. In this paper, travel time data from inductive loop road sensors and toll collection stations are merged through Dempster–Shafer inference to generate an improved estimate of travel time. The technique captures travel time data from the two sources and combines them by using Dempster's rule and belief values (also called probability mass) calculated from a confusion matrix. The most probable travel time over the monitored road section is selected as that with the largest belief. A case study is provided to illustrate application of the fusion technique with data gathered on winter Saturdays for 2 years: 2003 data are used to compute the confusion matrices and belief values, and 2004 data are used for validation.