Deriving Incident Management Measures Using Incident Probability Models and Simulation

Abstract

Incidents cause a large part of congestion and subsequent costs. Incident management is aimed at preventing the occurrence of incidents as well as minimizing the impact of incidents that occur. Two elements of incident management are addressed: the modeling of the incident occurrence probability for freeway sections and the modeling of the logistics of incident removal by various emergency services (tow trucks, ambulances, etc.). Models are presented that relate incident probabilities to road and traffic variables such as number of lanes, type of section (e.g., weaving sections), and traffic volumes and composition. These models give insight into the factors determining incident frequencies and may be used to design incident prevention measures. The results indicate that, for example, the splitting up of weaving sections into a series of joins and forks may be a very effective way to reduce the number of incidents. Traffic characteristics and the number of lanes also play important roles. These incident probability models can be used to generate the input for a logistics model. Such a model is presented. It was developed to test the effectiveness of various procedures aimed at minimizing incident duration. Examples of such procedures are the optimal location of emergency and recovery vehicles, given the spatial pattern of incident occurrence; optimal assignment of such vehicles to incidents; and various strategies for going from the vehicle's station to the incident site. These strategies include the permission to use the shoulder lane for emergency services. Experiments with the models for a heavily trafficked part of the Dutch freeway network indicate that a combination of such strategies can reduce incident congestion costs by nearly one-half.