CAMS: A CityGPT-Powered Agentic Framework for Urban Human Mobility Simulation

2025-06-16

AI Summary

Conventional data-driven approaches to urban human mobility simulation struggle to simultaneously ensure individual behavioral plausibility and collective spatial distribution fidelity, while heavily relying on external GIS data. Method: This paper proposes CityGPT-Agent—a novel simulation paradigm integrating a city-knowledge-enhanced large language model (LLM) with multi-agent coordination. It comprises three tightly coupled modules: MobExtractor for template-driven modeling, GeoGenerator for collective-knowledge-guided spatial generation, and TrajEnhancer for trajectory synthesis aligned with human preferences via Direct Preference Optimization (DPO). The framework operates without external geographic inputs and supports user-profile-driven behavioral modeling and spatial-knowledge-augmented generation. Contribution/Results: Evaluated on real-world datasets, CityGPT-Agent significantly outperforms state-of-the-art baselines, producing mobility trajectories that exhibit both high individual behavioral plausibility and strong collective spatial fidelity.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Human mobility simulation plays a crucial role in various real-world applications. Recently, to address the limitations of traditional data-driven approaches, researchers have explored leveraging the commonsense knowledge and reasoning capabilities of large language models (LLMs) to accelerate human mobility simulation. However, these methods suffer from several critical shortcomings, including inadequate modeling of urban spaces and poor integration with both individual mobility patterns and collective mobility distributions. To address these challenges, we propose extbf{C}ityGPT-Powered extbf{A}gentic framework for extbf{M}obility extbf{S}imulation ( extbf{CAMS}), an agentic framework that leverages the language based urban foundation model to simulate human mobility in urban space. extbf{CAMS} comprises three core modules, including MobExtractor to extract template mobility patterns and synthesize new ones based on user profiles, GeoGenerator to generate anchor points considering collective knowledge and generate candidate urban geospatial knowledge using an enhanced version of CityGPT, TrajEnhancer to retrieve spatial knowledge based on mobility patterns and generate trajectories with real trajectory preference alignment via DPO. Experiments on real-world datasets show that extbf{CAMS} achieves superior performance without relying on externally provided geospatial information. Moreover, by holistically modeling both individual mobility patterns and collective mobility constraints, extbf{CAMS} generates more realistic and plausible trajectories. In general, extbf{CAMS} establishes a new paradigm that integrates the agentic framework with urban-knowledgeable LLMs for human mobility simulation.

Problem

• Simulating human mobility with urban spatial modeling
• Integrating individual and collective mobility patterns
• Generating realistic trajectories without external geospatial data

Innovation

• Leverages CityGPT for urban mobility simulation
• Integrates MobExtractor for pattern synthesis
• Uses TrajEnhancer for realistic trajectory generation