Agent-Based Travel Scheduler: Decomposing OD Data for Predicting Individual Travel Schedules through Agent-Based Modeling

Abstract

This study introduces the Agent-Based Travel Scheduler (ABTS) model, designed to predict individual travel schedules by decomposing GPS-based aggregated Origin-Destination data according to demographic features and trip purposes. ABTS improves upon traditional models by generating detailed individual travel schedules based on agent-based modeling within an activity-based framework. The model's accuracy is refined through robust validation by calibrating key parameters that influence individual travel patterns in both spatial and temporal dimensions. Applied to Milwaukee, Wisconsin, USA, the model demonstrated 83% accuracy in estimating travel patterns, although it showed reduced performance in densely populated downtown areas and locations where special events that attract large crowds occurred. ABTS offers advantages in computational efficiency, flexibility in parameter adjustment, and adaptability to various urban scenarios, making it a valuable tool for policy analysis and urban planning.