Research papers · Posted 2024.12.23 · Preprint
DeepMF: Deep Motion Factorization for Closed-Loop Safety-Critical Driving Scenario Simulation
DeepMF: Deep Motion Factorization for Closed-Loop Safety-Critical Driving Scenario Simulation
Overview
DeepMF factorizes traffic risk into hazardous-agent selection, opponent motion, ego reaction and accident probability using separate neural models. Current and historical observations drive interactive, closed-loop generation rather than only editing recorded scenario snippets.
Version: v1