On the Variation in Mode Choice Behavior in Agent-based Travel Demand Models

Past research shows that individual mode choice preferences and, thus, taste variation of mode choice play an essential role and are observable even for one day. Both multiand monomodal behavioral patterns with different degrees of mode choice variation are the subject of investigation. Hence, agent-based travel demand models (AB-TDMs) must account for this taste variation, which is expected to affect model sensitivity. To assess the impacts of mode choice model configuration on the resulting variation, we apply an approach based on mixed logit models and implement them in an AB-TDM simulation. We analyze the mode choice behavior regarding variation indicators for the simulation period of one day and one week and compare it to observed behavior. We show that classic MNL models cannot appropriately account for mode choice variation in AB-TDM, both for one week or one day. We show that mixed model approaches can bridge this gap by better capturing heterogeneity and suggest using mixed models in an agent-based context. This prevents models from overestimating multimodal mode choice behavior.