Grasp planning of multi-fingered robot hands

Abstract

This paper presents our latesl: research results in stability analysis and synthesis of grasping of multi-fingered robot hands. A new approach, based on the ray-shooting technique, is developed for formulation of stability analysis and synthesis in both continuous and discrete spaces. This novel formulation leads to several simple and efticient algorithms for checking and planning of form-closure grasps in real-time. The algorithms have been implemented and simulation results will be also presented. further extended the approach to 3-D 4-finger grasps. The authors (3), (7) have recently developed an algorithm for computing form-closure grasps by iteratively moving the convex hull of the primitive contact wrenches towards the origin of the wrench space. In this paper, we present a new framework using the ray-shooting technique for stability analysis and planning of 3-D form-closure grasps. The ray-shooting technique is a powerful tool for extracting geometric information in Computational Geometry and Computer Graphics. Based on the duality between convex hulls and convex polytopes, a ray-shooting problem can be transformed to a Linear Programming (LP) problem (14), which can be solved in real-time when the size of the problem is not huge. It will be demonstrated that the qualitative test and computation of a form-closure grasp can be formulized as ray-shooting problems. Based on the ray-shooting technique, a simple and efficient heuristic method is developed for searching for a form-closure grasp on the surface of a 3-D object in the direction of reducing the distance between the convex hull and the origin of the wrench space. To overcome the local minimum problem in the heuristic search, a recursive problem decomposition strategy is introduced so that a complete and efficient algorithm is developed for finding a form-closure grasp in a discrete domain. It is proved that the algorithm takes 0(-) time, where K is the number of the local minimum points in the grasp space and n is the number of fingers. The . complexity of the algorithm does not depend on the number of points representing the object, but on the number of the local minimum points (grasps), i.e. the geometry of the discrete surface of the object. This algorithm works much more efficiently than the combinatory method whose computational complexity is Ow), where N denotes the number of points representing the surface of the object. Compared with other heuristic methods, this algorithm can always find a solution or report no solution. The heuristics used here is just to guide the local search procedure. KlnK n