微信里点“发现”,扫一下
二维码便可将本文分享至朋友圈
演讲摘要:Conventional grasp planning algorithms are based on model-free sampling-based techniques that ignore the underlying structures of the problem. We propose a line of model-based grasp planners that are aware of object shapes, material properties, and various constraints. We first formulate a stress-minimization grasp metric that minimizes the probability of causing brittle fracture. Next, we formulate a deep-differentiable grasp planning algorithm that uses gradient information to guide deep grasp planning policy optimization. Finally, we propose an optimization-based grasp planner that unifies precision and power grasp. In all these techniques, we use the model-based approach either accelerates grasp planning or improves the quality of grasps.
讲者简介:Zherong Pan is a senior researcher at Tencent America. He was a postdoctoral researcher at the Intelligent Motion Lab of the University of Illinois Urbana Champaign. He obtained Ph.D. degree from the University of North Carolina at Chapel Hill. His research is focused on numerical analysis, motion planning, and physics-based modeling for high-dimensional deformable objects.