TrajGenAgent: A Hierarchical LLM Agent for Human Mobility Trajectory Generation

In three linesTrajGenAgent is a hierarchical LLM-agent framework for realistic human mobility trajectory generation without model fine-tuning. An orchestrator LLM synthesizes activity chains via in-context learning, then a deterministic workflow grounds them using personalized POI retrieval, distance-aware location selection, and LLM-based duration estimation. Evaluation via anomaly-detection framework on benchmark datasets.

Summary generated by Claude — human-verified