ITSC 2025 Paper Abstract

Paper TH-EA-T18.1

Liu, Yifan (University of California, Los Angeles), Liao, Xishun (University of California, Los Angeles), Ma, Haoxuan (University of California, Los Angeles), Liu, Jonathan (University of California, Los Angeles), Jadhav, Rohan (University of California, Los Angeles, Mobility Lab), Ma, Jiaqi (University of California, Los Angeles)

MobiVerse: Scaling Urban Mobility Simulation with Hybrid Lightweight Domain-Specific Generator and Large Language Models

Scheduled for presentation during the Invited Session "S18b-Innovative Applications of LLM in Multimodal Transportation Systems" (TH-EA-T18), Thursday, November 20, 2025, 13:30−13:50, Southport 3

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Keywords Digital Twin Modeling for ITS Infrastructure and Traffic Simulation, Demand-Responsive Transit Systems for Smart Cities, Data Analytics and Real-time Decision Making for Autonomous Traffic Management

Abstract

Understanding human mobility patterns has long been a challenging task in transportation modeling. Due to the difficulties in obtaining high-quality training datasets across diverse locations, conventional activity-based models and learning-based human mobility modeling algorithms are particularly limited by the availability and quality of datasets. Current approaches primarily focus on spatial-temporal patterns while neglecting semantic relationships such as logical connections or dependencies between activities and household coordination activities like joint shopping trips or family meal times, both crucial for realistic mobility modeling. We propose a retrieval-augmented large language model (LLM) framework that generates activity chains with household coordination using only public accessible statistical and socio-demographic information, reducing the need for sophisticated mobility data. The retrieval-augmentation mechanism enables household coordination and maintains statistical consistency across generated patterns, addressing a key gap in existing methods. Our validation with NHTS and SCAG-ABM datasets demonstrates effective mobility synthesis and strong adaptability for regions with limited mobility data availability.