UniContact:A Basic Model for Robotic Manipulation of Contact Synthesis on Rigid and Articulated Rigid Bodies with Arbitrary Manipulators

21 Sept 2023 (modified: 11 Feb 2024)Submitted to ICLR 2024EveryoneRevisionsBibTeX
Supplementary Material: pdf
Primary Area: applications to robotics, autonomy, planning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: manipulation, grasping, robotics
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: We posit that one fundamental, core component of robotic manipulation is inferring contacts with the environment, enabling the agent to exert control. In this work, we study a fundamental problem of contact synthesis in robotic manipulation to choose a set of contact positions and forces on a random rigid or articulated rigid object for an arbitrary robot manipulator to produce a specified external wrench. Our framework first segments the point clouds with normals into feasible contact region sets. For each feasible contact region set, a model is trained to produce the feasible contact point within these region sets by taking as inputs the robot description, the target wrench, the object point cloud with normals, and the contact region set. After gathering the contact positions from the neural network model, we develop an optimization process to fine-tune the contact points and contact forces and generate the joint values for the robotic manipulator to exert contact forces on the object's surface without penetration. We perform extensive experiments to verify the effectiveness of our proposed framework both in simulation and in real-world experiments. Supplementary and Videos are on the website.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 3011
Loading