Activity scheduling and derived trip departure time choice are essential components of an Activity-Based travel demand Model (ABM). Trip departure time choice directly impacts the temporal distribution of flows in the network. In many ABMs, trip departure times within the tour are assigned based on a discrete choice model conditional upon the overall tour schedule. This model is formulated as joint trip departure time choice (normally, in 30-min intervals) for all trips on the tour. However, the number of alternatives with such an approach increases exponentially with the number of trips on the tour and with number of time intervals.
For this reason, in some ABMs, a sequence of regression or hazard duration models is applied to model activity duration and trip departure in continuous time for intermediate stops on the tours. Such an approach resolves the problem of temporal resolution but it is not behaviorally appealing since the scheduling decisions for different activities and corresponding trip departure times are modeled strictly chronologically.
Improvement of trip departure time models both in terms of temporal resolution and behavioral integrity recently obtained an additional focus in the context of integration of ABMs and DTA (Dynamic Traffic Assignment). For such an integration to be coherent, the trip departure time is preferred to be continuous. The trip departure model in the Jerusalem ABM is improved by using a time allocation approach in which the available time on the tour is allocated to different activities or stops given the expected travel time for all trips. Tour start and end times are predicted by a discrete choice model with a temporal resolution of 15 min.
The budget allocation model used in the Jerusalem ABM is Multiple Discrete-Continuous Extreme Value (MDCEV) model originally developed by Bhat in 2008. An original and very efficient algorithm for model implementation was developed by the authors. The validation results are based on the recent large-scale Household Travel Survey in Jerusalem. Model sensitivity testing is demonstrated for a highway pricing scenario. This approach is behaviorally more appealing and also paves a way for seamless ABM-DTA integration.