Modular, Accessible, Sensorized Objects for Evaluating the Grasping and Manipulation Capabilities of Grippers and Hands

Abstract

The human hand is Nature's most versatile and dexterous end-effector and it has been a source of inspiration for roboticists for over 50 years. Recently, significant industrial and research effort has been put into the development of dexterous robot hands and grippers. Such end-effectors offer robust grasping and dexterous, in-hand manipulation capabilities that increase the efficiency, precision, and adaptability of the overall robotic platform. This work focuses on the development of modular, sensorized objects that can facilitate benchmarking of the dexterity and performance of hands and grippers. The proposed objects aim to offer; a minimal, sufficiently diverse solution, efficient pose tracking, and accessibility. The object manufacturing instructions, 3D models, and assembly information are made publicly available through the creation of a corresponding repository.

Electromyography-Based Decoding of Dexterous, In-Hand Manipulation of Objects: Comparing Task Execution in Real World and Virtual Reality

“…In the first experiment, subjects were given three real objects and were asked to perform in-hand manipulation motions. The objects used for the experiment were a Rubik's cube and a chips can from the Yale-CMU-Berkeley (YCB) grasping object set [25], and a custom made sensorized sphere [26]. The sphere was designed to accommodate efficiently retro-reflective markers to capture its motion.…”

Grasping Benchmarks: Normalizing for Object Size \& Approximating Hand Workspaces

“…They especially emphasize the need for reporting 'object ranges' that a hand can grasp. To this end, they recommended using object sets (a good sample: [3], [9], [10], [4], [11]).…”

“…Previously, researchers have addressed this problem by using standardized object sets (such as [3], [4]) morrowjo, nishatn, campbjos, ravi.balasubramanian, cindy.grimm at Oregon State University, *all authors contributed equally and simple task-based measures to characterize a robot hand's capabilities [5]. These standards help to normalize the object shape and task, but do little to help with comparisons across hand designs -if hand workspaces are different sizes, success or failure may simply be a result of the object's size relative to the hand, not the ability of the hand to perform the task.…”

Sensorized objects used to quantitatively study distal grasping in the African elephant

“…More recently, Gao et al. 42 developed modular sensorized objects for trajectory tracking. Although force sensors are not embedded in these objects, they can facilitate benchmarking the dexterity and performance of artificial hands and grippers.…”