Bayesian inference for vehicle speed and vehicle length using dual-loop detector data

A dual-loop detector consists of two connected single-loop detectors placed several feet apart. Compared with a single-loop detector, it is able to provide more useful information on traffic flow with a higher precision. In this paper we investigate statistical inference for vehicle speed and vehicle length using dual-loop detector data. A Bayesian analysis is performed to combine current observations on traffic flow with prior knowledge, which results in a set of simple formulas for the online estimation of both vehicle speed and vehicle length. As a by-product, vehicle classification is also investigated on the basis of posterior classification probabilities. The computational overhead of updating the estimates is kept to a minimum when new information on traffic flow becomes available. The method is illustrated using real traffic data.