Grasping Planning of Multi-fingered Robot Hand by Bundle Trust Region Method

摘要

How to plan a proper set of contact points on an object so as to maximize the grasp capability is a common but generally difficult problem, which is mainly because of its nondifferentiable complexity. Subgradient methods are commonly used for finding solution, which converge slowly and require many iterations to reach an optimum. In this study, we adapt the bundle trust region method to the grasp planning problem for achieving fast convergence. The effectiveness of the algorithm to the planning problem is tested by numerical examples.