Trip-timing decisions with traffic incidents

This paper analyzes traffic bottleneck congestion when drivers randomly cause incidents that temporarily block the bottleneck. Drivers have general scheduling preferences for time spent at home and at work. They independently choose morning departure times from home to maximize expected utility without knowing whether an incident has occurred. The resulting departure time pattern may be compressed or dispersed according to whether or not the bottleneck is fully utilized throughout the departure period on days without incidents. For both the user equilibrium (UE) and the social optimum (SO) the departure pattern changes from compressed to dispersed when the probability of an incident becomes sufficiently high. The SO can be decentralized with a time-varying toll, but drivers are likely to be strictly worse off than in the UE unless they benefit from the toll revenues in some way. A numerical example is presented for illustration. Finally, the model is extended to encompass minor incidents in which the bottleneck retains some capacity during an incident.