Roadway travel times: maximum likelihood estimation based on floating car data intervals

Abstract

Massive Floating Car Data (FCD) datasets have become available for roadway networks, which contain travel time information on short spatial intervals between pairs of successive observations along individual trips. This paper brings about a stochastic model of travel times with a Maximum Likelihood estimation method to exploit FCD material. Probabilistic specifications are put forward for link travel times as Gaussian random variables along with standard error of each estimator. This allows for simple estimation of link attributes based on “Link FCD intervals” and their confidence intervals. An application instance is dealt with for one motorway and one urban avenue in the Grand Paris area with results showing better accuracy than automotive methods based on pointwise average speed.