Research on short-term traffic conditions prediction has been largely concerned with parameters such as flow, occupancy, and speed, ignoring predictions during congestion, a period when predictions are needed the most. Emanatingg from the practical need to predict traffic parameters during congested periods, this paper uses the principles of duration modeling to address an important question: given the onset of congestion, how long will it last? As such, the goal of this paper is to propose an approach for estimating the duration of congestion on a given road section and the probability that, given its onset, congestion will end during the following time period. The results indicate that the Loglogistic functional form best describes congestion duration, and that the probability of congestion ending within a specified time period is likely if it has lasted up to approximately 12 min (with a peak at 6 min). Further, it was found that if congestion lasted over 21 min it was probably caused by something external to the traffic system events.
Abstract