Pougala, J., Hillel, T., and Bierlaire, M. (2022)

OASIS: Optimisation-based Activity Scheduling with Integrated Simultaneous choice dimensions

Activity-based models offer the potential of a far deeper understanding of daily mobility behaviour than trip-based models. However, activity-based models used both in research and practice have often relied on applying sequential choice models between subsequent choices, oversimplifying the scheduling process. Recent work (Pougala et al., 2022) has established a new modelling approach for activity scheduling which integrates the different choice dimensions (activity participation, ordering, scheduling, etc) into a single optimisation framework, based on utility maximisation principles. In this paper we introduce a complementary methodology for the estimation of model parameters from historic data, allowing for the generation of realistic and consistent daily mobility schedules. In combination, the optimisation-based scheduling simulation and the parameter estimation framework, is hereby referred to as OASIS (Optimisation-based Activity Scheduling with Integrated Simultaneous choice dimensions). The estimation framework introduced in this paper consists of two main elements: (i) choice set generation, where we generate a sample of competitive alternative schedules by applying the Metropolis-Hasting algorithm to historic schedules, and (ii) discrete choice parameter estimation, where the scheduling process is formulated as a discrete choice problem, in which each individual chooses a full daily schedule from a finite set of possible schedules. We validate our approach by estimating parameters for a sample of individuals from the 2015 Swiss Mobility and Transport Microcensus (Office fédéral de la statistique and Office fédéral du développement Territorial, 2017), and evaluating the output of the OASIS model against realised schedules from the data. The results demonstrate the ability of the new framework to estimate stable and significant parameters from historic data that are consistent with behavioural theory. Furthermore, the schedules outputted by the OASIS model are evaluated as being both feasible and realistic at a disaggregate level, and correspond well to the aggregate trends in the historic data. This work opens the way for future developments of activity-based models, where a great deal of constraints can be explicitly included in the modelling framework, and all choice dimensions are handled simultaneously.