Using probe data to estimate OD matrices

Abstract

Probes provide an entirely new kind of data source for estimating OD matrices. Indeed, if probes were the Universe of vehicles, then the OD matrix would be known with certainty, failing limitations of the way the probes are sensed and variations from day to day. Consequently, the question is: what tradeoffs exist between probe density and the quality of the estimated matrix? In addition, do different methodologies produce matrices of different quality and how does the availability of other data, like link volume counts, influence the quality of the matrix produced? Moreover, what type of estimation methodology works best? These questions are addressed in this paper, through a literature review, a discussion of the issues involved, and an example problem. One conclusion reached by the authors is that it is possible under certain conditions to obtain a mean absolute percent deviation of 9% for the OD flows if the percentage of probe trips is 10%.