Benchmarking the Dexterity and Performance of Grippers and Hands with Modular, Sensorized Objects

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 robot system. For in-hand manipulation, no complete and unified basis for comparison is available and this lack of suitable benchmarking methods should be addressed. This work focuses on the development of modular, sensorized objects that will facilitate benchmarking of the grasping and dexterous, in-hand manipulation capabilities of both human and robot hands. The proposed objects aim to solve three major issues in this context; pose tracking, the size vs diversity problem, and accessibility. Along with the objects, a set of benchmarking protocols and procedures is provided with the aim of establishing proper foundations for a comprehensive hand dexterity assessment. The manufacturing instructions of the objects, the descriptions of the protocols and the benchmarks and appropriate code that facilitates the experimentation will be made publicly available through the creation of a corresponding repository.

More details can be found at the following publication:

Geng Gao, Gal Gorjup, Ruobing Yu, Patrick Jarvis, and Minas Liarokapis, "Benchmarking the Dexterity and Performance of Grippers and Hands with Modular, Sensorized Objects," IEEE Robotics and Automation Letters, 2019 (under review).