Filters








2 Hits in 2.1 sec

OmniHang: Learning to Hang Arbitrary Objects using Contact Point Correspondences and Neural Collision Estimation [article]

Yifan You, Lin Shao, Toki Migimatsu, Jeannette Bohg
2021 arXiv   pre-print
Our system learns to estimate the contact point correspondences between the object and supporting item to get an estimated stable pose.  ...  To this end, we train a neural network based collision estimator that takes as input partial point clouds of the object and supporting item.  ...  Our primary contributions are: (1) proposing a contact point matching representation for object manipulation tasks and applying it to learn how to hang arbitrary objects (2) proposing a neural motion  ... 
arXiv:2103.14283v1 fatcat:sfi2tmdfineqnh5bigcdck3tuu

Learning Models as Functionals of Signed-Distance Fields for Manipulation Planning [article]

Danny Driess, Jung-Su Ha, Marc Toussaint, Russ Tedrake
2021 arXiv   pre-print
To demonstrate the versatility of our approach, we learn both kinematic and dynamic models to solve tasks that involve hanging mugs on hooks and pushing objects on a table.  ...  We show that representing objects as signed-distance fields not only enables to learn and represent a variety of models with higher accuracy compared to point-cloud and occupancy measure representations  ...  Omnihang: Learning to hang arbitrary objects using contact point correspondences and neural collision estimation. In 2021 IEEE International Conference on Robotics and Automation (ICRA).  ... 
arXiv:2110.00792v1 fatcat:lvs52m6nbrd4fop6gis2c74i3y