Baud, C., and Bierlaire, M. (2026)

A Flexible and Realistic Synthetic Panel Population Generation

29th IEEE International Conference on Intelligent Transportation Systems (ITSC), Naples, Italy

This paper proposes a general framework for constructing synthetic populations whose panel structure is specified by design. Most existing methods are snapshot-based or rely on pseudo-panels, neither of which tracks individuals consistently over time. A time-independent approach has been illustrated through a proof-of-concept example but not formal- ized as a general, model-agnostic framework. The main contribution of this paper is to formalize and generalize the time-independent perspective into a unified framework capable of accommodating arbitrary probabilistic specifications. Unlike the prior example-based implementation, the proposed framework explicitly enforces internal consistency and realism through structural constraints embedded in a life- based representation. Individuals are defined independently of calendar time, and a deterministic mapping recovers their state at any time, so that panel data are obtained by construction. The approach is model-agnostic, and applicable at arbitrary temporal and spatial resolutions. We implement the approach and show that it enables the generation of fully specified synthetic populations from which both individual trajectories and aggregate population distributions can be derived at any time of interest.