Finite element-based grasp analysis using contact pressure maps of a robotic gripper

Conventionally, a robotic grasp is represented by point contacts with rigid body description for object and gripper. With these assumptions, the contact forces and moments are considered to act at points, and a grasp wrench space is constructed to determine the stability of the grasp. However, wrench space formulations ensure mechanical equilibrium only; it does not guarantee grasp stability. Moreover, the material properties of object and grippers are not taken into account, as a result of which the underlying deformation and the contact pressure are neglected. In this work, the contact pressure maps of robotic grasps, that accounts for deformability occurring at the gripper-object interface, are used for robotic grasp analysis and database generation. Using a conformable pressure sensing sheet, the contact pressure maps of typical household objects such as cans, bottles and cubes, while grasping and perturbation are measured and explored. Apart from these objects, the contact pressure maps of fruits such as apple and orange, while grasping is also studied. The stability of each grasp is determined based on the contact pressure maps, and a contact area-based metric, π\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pi $$\end{document}, is used to identify the most stable grasp. These pressure maps are then used to validate the FE framework designed for robotic grasp analysis. π\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pi $$\end{document} obtained from experiments is then compared with that of FE simulation outputs, and it is observed that the relative error between them is less than 4%, which, in turn, demonstrates the efficacy of the proposed framework. As the FE framework is not suited for real-time grasp synthesis due to the computational time involved, we show how to exploit the framework for generating an offline grasp experience database, which can then be deployed for grasp synthesis real-time.