An Agile Design of Activity-based Mobility Intervention using Large Language Models
Joon-Seok Kim, Jooyoung Yoo, Su Han, Andreas Züfle
Human mobility simulations are essential tools for evaluating policies and planning effective interventions in dynamic environments. Recently, a conceptual framework for Activity-Based Mobility Interventions (ABMI) has been proposed to support such simulations by facilitating intervention design across heterogeneous simulation models and capabilities. However, current approaches to designing these interventions rely heavily on manual effort, limiting scalability, responsiveness, and overall productivity. In this study, we propose using large language models (LLMs) to automate and streamline the ABMI design process. Our method enables the rapid generation of intervention scenarios from high-level descriptions and contextual information, significantly reducing the time and expertise required. We demonstrate the effectiveness of our approach in creating intervention scenarios for emergency response scenarios, including wildfire evacuations and epidemic containment. By integrating LLMs into the mobility simulation workflow, this work opens new avenues for adaptive, data-driven policy design and enables more timely and effective responses to real-world crises.