Kukic, M., and Bierlaire, M. (2025)

Simulation Framework for Longitudinal Synthetic Population Generation

25th Swiss Transport Research Conference, Ascona, Switzerland

This paper introduces a novel framework for generating longitudinal synthetic populations that track individuals over time, addressing limitations of traditional snapshot-based synthetic population methods. We propose a Gibbs sampler-based approach that combines models and cross-sectional data to generate universal, time-independent variables, which enable the consistent derivation of time-specific synthetic populations at any point in time. A key advantage of this framework is that any changes to the universal dataset are automatically reflected in derived datasets, allowing for efficient scenario testing. The methodology is demonstrated using Swiss Mobility and Transport Microcensus data, by simulating the impacts of hypothetical events such as pandemics. This approach ensures temporal consistency, captures individual-level dynamics, and reduces the computational burden of regenerating populations, showcasing its potential for activity-based modeling and long-term policy analysis when real longitudinal data is unavailable.