TrajGenAgent: A Hierarchical LLM Agent for Human Mobility Trajectory Generation

TrajGenAgent is a newly proposed hierarchical LLM agent designed for generating human mobility trajectories without the need for model fine-tuning. It employs a two-stage orchestrator-worker architecture, where an LLM synthesizes activity chains based on historical data and a deterministic workflow grounds these activities using personalized points of interest and kinematics-aware travel time estimation. The framework demonstrates improved spatiotemporal fidelity and semantic coherence compared to existing models, making it a valuable tool for practitioners focused on realistic synthetic trajectory generation for applications in transportation and urban planning.