Agent Trajectory Analysis
Overview
Agent trajectory analysis examines sequences of actions or movements, from autonomous vehicles and pedestrians to language-model agents using tools. Methods include social and physics-guided motion prediction, trace and graph representations, step-level evaluation, and attribution of failures or reusable behaviors. Benchmarks and trajectory-level measures reveal inefficiency, unsafe steps, and procedural differences that final success scores can miss. A central challenge is reliably connecting events across long histories to outcomes, while balancing detailed diagnosis with scalable, efficient evaluation.
Latest papers
51 papers3 papers in the last four weeks, down 40% on the four weeks before. 0.0% of all new papers.