Aggregate calibration of travel demand sub-models using opportunistic data: a sequential hybrid optimization approach

This paper investigates methods and algorithms for the aggregate calibration of travel demand sub-models, focusing on destination and mode choice models under data-limited conditions. Smaller planning agencies often lack access to advanced Decision Support Systems, such as Activity-Based Models (ABMs), due to the substantial data, time, and resource requirements. As a practical alternative, model transfer from other estimation contexts or the specification of models based on existing studies is commonly adopted. Within this research, calibration strategies for destination and mode choice models are explored using synthetic data generated by an Activity-Based Model, in a controlled parameter-recovery setting. The use of synthetic data is motivated by the limited availability of real-world contexts in which both a fully specified ABM and suitable opportunistic data sources are simultaneously accessible, as well as by the substantial effort required to pre-process and validate real opportunistic data within typical research timelines. The calibration problem is formulated as a simulation-based optimization task, and its implementation is tested on a small network using an ABM demand model. Three optimization algorithms are evaluated in a controlled environment, assessing their ability to recover known model parameters. The application of the ADAM optimizer in this context represents a novel contribution, extending its use to the aggregate calibration of travel demand models. Finally, a sequential hybrid optimization framework is proposed, combining global exploration, efficient convergence, and local refinement. The results demonstrate the framework’s potential for advancing transportation modeling research and provide practical insights for improving Decision Support Systems in data-constrained planning environments.