New methods for travel time estimation on freeway sections

Abstract

In this paper we present two novel approaches to estimate the travel times between subsequent detector stations in a freeway network, with long distances between detector stations and several unobserved onand off-ramps. The network under investigation is a two-lane freeway. The maximum distance between detector stations, for which travel times were estimated is about 20 km with four unobserved onand off-ramps in between. The algorithms were applied on real data sets, which has led to reasonable estimates. However, due to unknown actual ('true') travel times, a performance assessment was not possible. The algorithms were also applied on simulated data with known travel times. This allowed the verification of the estimated travel times. The simulated data were generated by the microscopic traffic simulation tool AIMSUN NG®. The detector stations were assumed to be equipped with widespread double loop detectors, i.e., for each vehicle, the only information used was its length (with a superimposed measurement noise) and the arrival time at the detector stations. The estimated travel times show that with both methods all relevant travel time characteristics were correctly identified for the investigated scenarios. Moreover, a comparison of the estimates with the actual travel times has shown very good accuracy. Besides the fact that the methods work well even under hindered conditions (long distance, unobserved ramps), some additional practical benefits are: provided that single car data are available with sufficient accuracy, no additional investments are required; both methods work fully anonymous; extensions to more sophisticated detection technologies that provide additional vehicle features are straightforward; the travel time estimates form a good basis for any travel time prediction method.