Risk-aware multimodal choice during disturbances: a computational graph-based hybrid choice model

Understanding multimodal travel behavior is critical for resilient urban transportation planning, particularly under disturbances. This study proposes a Multidimensional Risk-Enhanced Hybrid Choice Model (MR-HCM) that systematically integrates objective and subjective risk factors. The COVID-19 pandemic serves as a case study, with the Wells–Riley equation used to quantify infection probability as an objective risk and survey-derived latent variables representing psychological safety perceptions. Leveraging a Computational Graph (CG) framework, the MR-HCM enables efficient and robust parameter estimation. Empirical validation using the 2022 pandemic-era National Household Travel Survey (NHTS2022) shows that incorporating dual risks improves performance, increasing log-likelihood by 4.84% over baseline models. Beyond health crises, the MR-HCM’s modular design supports extension to other disruptions, such as delays or environmental hazards, through adjusted objective and latent components. Looking ahead, this framework facilitates risk-based metro scheduling and adaptive multimodal coordination, enhancing resilience and passenger-centric urban transportation planning.