The Anthropomorphic Hand Assessment Protocol (AHAP)

The progress in the development of anthropomorphic hands for robotic and prosthetic applications has not been followed by a parallel development of objective methods to evaluate their performance. The need for benchmarking in grasping research has been recognized by the robotics community as an important topic. In this study we present the Anthropomorphic Hand Assessment Protocol (AHAP) to address this need by providing a measure for quantifying the grasping ability of artificial hands and comparing hand designs. To this end, the AHAP uses 25 objects from the publicly available Yale-CMU-Berkeley Object and Model Set thereby enabling replicability. It is composed of 26 postures/tasks involving grasping with the eight most relevant human grasp types and two non-grasping postures. The AHAP allows to quantify the anthropomorphism and functionality of artificial hands through a numerical Grasping Ability Score (GAS). The AHAP was tested with different hands, the first version of the hand of the humanoid robot ARMAR-6 with three different configurations resulting from attachment of pads to fingertips and palm as well as the two versions of the KIT Prosthetic Hand. The benchmark was used to demonstrate the improvements of these hands in aspects like the grasping surface, the grasp force and the finger kinematics. The reliability, consistency and responsiveness of the benchmark have been statistically analyzed, indicating that the AHAP is a powerful tool for evaluating and comparing different artificial hand designs