Mobility Anomaly Generation using LLM-Driven Behavior with Kinematic Constraints

In three linesGenerative framework to create annotated human trajectory anomalies. Uses LLM agents to inject semantic behavioral anomalies (out-of-distribution check-ins, skipped routine visits) into simulated trajectories, with map-constrained routing reconstruction and context-aware spatial noise model to bridge simulation-to-reality gap.

Summary generated by Claude — human-verified