Motorway incident detection using probe vehicles

Vehicles that collect data from a traffic stream of which they are part (i.e. probe vehicles) have been shown to provide reliable and cost-effective real-time travel information. This paper presents a probe-vehicle-based algorithm designed to detect incidents on motorways. The algorithm is based on a bivariate analysis model (BEAM) using two variables: the average travel times of probe vehicles, and the travel time differences between adjacent time intervals. The premise of the model is that link travel times increase more rapidly as a result of a change in capacity (i.e. when an incident occurs) than as a result of a change in demand. The statistical principles of bivariate analysis have been used to study the relationships between the two variables in incident and non-incident conditions. Four motorway road links with field data have been used to test the feasibility of the model with 99% and 99.9% coverage contours respectively. The model achieved an average incident detection rate of 89.5% with a false alarm rate of 0.71% using the 99% coverage contour. The average incident detection rate was 73.7% with a false alarm rate of 0.57% using the 99.9% coverage contour