Modeling Leisure-related Agent-Place Relationships in Agent-based Travel Demand Models

Understanding spatial behavior is essential for accurately modeling travel behavior, as many activities are repeatedly conducted at a limited set of preferred locations. This paper presents a novel approach for synthesizing agent-specific place relationship networks and integrating them into destination choice modeling within the agent-based framework mobiTopp. Using multi-week activity tracking data from the MOBIS study, we develop sequential models that estimate the number of leisure-related relationships, characterize their behavioral and temporal structure, and assign spatially plausible destinations based on attractiveness and accessibility. The resulting agent-place relationship networks extend conventional short-term destination choice models by embedding long-term spatial preferences. An exemplary analysis of measure sensitivity reveals that incorporating agent-place relationships substantially reduces the responsiveness of agents with strong preferences to accessibility changes, while agents without such relationships remain highly sensitive.