Task and context sensitive optimization of gripper design using dynamic grasp simulation

Abstract

In this work, we present a generic approach to optimize the design of a parametrized robot gripper including both gripper parameters and parameters of the finger geometry. We demonstrate our gripper optimization on a parallel jaw type gripper which we have parametrized in a 11 dimensional space. We furthermore present a parametrization of the grasping task and context, which is essential as input to the computation of gripper performance. We exemplify the feasibility of our approach by computing several optimized grippers on a real world industrial object in three different scenarios.

Optimizing grippers for compensating pose uncertainties by dynamic simulation

“…Simulation as a tool in the context of grasping and gripper design is mostly encountered in problems of optimal grasp planning and structural design [11], [12]. In our previous research, we have used dynamic simulation to obtain feedback in the finger geometry design phase [3], [13]. In this work, using an approach similar to [2], we aim at confirming the relevance of such feedback by presenting and comparing results obtained from the experiments performed in both simulated and the real-world settings.…”

“…The optimal gripper design for the objects in the considered industrial scenario (magnet and rotorcap) has been learned using the methods presented in [3]. We describe the main points of the method below for the sake of completeness.…”

“…] is the vector of m individual gripper quality indices and w i are the respective weights. Since in the considered industrial case, we are mostly concerned with the alignment properties of the gripper, we have selected the following weights for the objective function: The assumed limits for the volume and stress are V limit = 200cm 3 and σ limit = 25MPa respectively.…”

“…Different types of such cutouts exist, such as cutouts based on molding ("inverse" of the object) [1], convex hulls [2], or simple geometric primitives (prismatic, round). In [3], it was shown how to learn optimal finger parameters for two-finger grippers in simulation, where one of the objectives was to enhance the gripper's alignment capability.…”

Task and Context Sensitive Gripper Design Learning Using Dynamic Grasp Simulation

“…This paper is based on two conference papers, one on gripper evaluation by means of an objective function [2] and the other on the learning of the gripper shape by optimizing this objective function through gradient descent [3]. Compared to [2] and [3], in this paper we present:…”