TrajLLM: A Modular LLM-Enhanced Agent-Based Framework for Realistic Human Trajectory Simulation

TL;DR

TrajLLM tackles privacy and cost barriers in mobility data by introducing a modular, LLM-enhanced agent-based framework that simulates realistic daily trajectories. It jointly leverages persona and activity generation via LLMs with a dual destination selection strategy: memory-informed LLM-based recommendations and a physics-inspired spatial model that computes POI probabilities through $W^s_j$, $f(d)$, and $P_i$, complemented by memory-driven consistency. Key contributions include the four-module architecture (persona, activity, destination, memory), explicit equations for spatial-dynamic destination selection such as $W^s_j = \frac{V_j}{f(d_{ij})}$ with $f(d) = (d + r_0)^{-\beta} e^{-d/k}$ and $P_i = \frac{W^s_i W^f_i}{\sum_j W^s_j W^f_j}$, plus a scalable memory-summarization approach that enables long-term realism. The framework supports interactive demonstrations and shows that LLM-driven simulations align with observed patterns, offering a privacy-preserving and interpretable tool for urban planning, traffic management, and public health.

Abstract

This work leverages Large Language Models (LLMs) to simulate human mobility, addressing challenges like high costs and privacy concerns in traditional models. Our hierarchical framework integrates persona generation, activity selection, and destination prediction, using real-world demographic and psychological data to create realistic movement patterns. Both physical models and language models are employed to explore and demonstrate different methodologies for human mobility simulation. By structuring data with summarization and weighted density metrics, the system ensures scalable memory management while retaining actionable insights. Preliminary results indicate that LLM-driven simulations align with observed real-world patterns, offering scalable, interpretable insights for social problems such as urban planning, traffic management, and public health. The framework's ability to dynamically generate personas and activities enables it to provide adaptable and realistic daily routines. This study demonstrates the transformative potential of LLMs in advancing mobility modeling for societal and urban applications. The source code and interactive demo for our framework are available at https://github.com/cju0/TrajLLM.

Figures (2)

• Figure 1: Overall Pipeline of TrajLLM
• Figure 2: Real-Time Simulation of Daily Activities in Tokyo