A probabilistic framework for learning kinematic models of articulated objects

摘要: Robots operating in domestic environments generally need to interact with articulated objects, such as doors, cabinets, dishwashers or fridges. In this work, we present a novel, probabilistic framework for modeling objects kinematic graphs. Vertices graph correspond object parts, while edges between them model their relationship. particular, set of parametric and non-parametric edge models how they can robustly be estimated from noisy pose observations. We furthermore describe estimate the structure use learned prediction robotic manipulation tasks. finally generalized new previously unseen objects. various experiments using real robots different camera systems well simulation, show that our approach is valid, accurate efficient. Further, demonstrate has broad applications, particular emerging fields mobile service robotics.