This paper studies failure prediction for robotic lift-and-place tasks in non-prehensile material handling, where objects are supported by friction rather than grasped. High-speed motions can cause objects to slip or destabilize, while slower motions reduce throughput. PREFAIL analyzes the relative motion between target objects and their carrier to identify precursors of failure. The authors also introduce a dataset that labels the latest intervention time, allowing evaluation of whether a prediction arrives early enough to prevent failure. The approach is validated on simulation and real-world datasets, with the abstract reporting improved accuracy and timeliness.
No heat snapshots are available in the last 24 hours.