Modeling Day-to-day Trip Choice Evolution under Network Disruption

In this paper we propose a “prediction-correction” framework to model traveler’s perception evolution under network disruption. Distinctive from existing models, the proposed framework assumes drivers make their trip choices according to their predictions on future traffic, since previous daily experiences become invalid when network is disrupted. Drivers predict travel costs after network disruption, and then correct their predictions by comparing with their actual experienced travel costs. Traveler’s prediction is formulated as an individual dynamic process, such that qualitative network conditions could be quantified in the model. We demonstrate the proposed traffic evolution model using the data collected from the I-35W Bridge collapse in Minneapolis, Minnesota, and compare it with a day-to-day traffic assignment model without prediction behaviors. To the best of our knowledge, this is the first time that day-to-day traffic evolution models have been applied to study realistic network disruption scenarios.