TrajLLM introduces a modular framework that uses large language models to simulate realistic human mobility patterns while addressing privacy and cost concerns in traditional approaches.

• Combines persona generation, activity selection, and destination prediction in a hierarchical structure
• Leveres real-world demographic and psychological data to create realistic movement patterns
• Integrates both physical models and language models for comprehensive mobility simulation
• Offers applications in urban planning, epidemic modeling, and emergency response

This engineering innovation provides a scalable approach to understanding human movement patterns without compromising individual privacy, enabling better infrastructure planning and public health interventions.

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