Measuring Anthropomorphism of a New Humanoid Hand-Arm System

Existing perceptional and analytical techniques for quantifying anthropomorphism of robotic hand-arm systems have significant limitations. The former is based on subjective appraisals by human observers, which are prone to biases. The latter frequently disregards the manipulability of the robotic hand-arm system leading to potentially flawed estimation of anthropomorphism. Additionally, prior methodologies have predominantly focused on assessing anthropomorphism of robotic hands while paying scant attention to that of robotic arms. In this paper, we propose an analytical method to quantify the functional anthropomorphism of robotic hand-arm systems. In this method, relative workspace coverage and manipulability of the robot arm and fingers are compared with those of the human arm and fingers. The proposed methodology is then used to quantify functional anthropomorphism of several hand-arm systems, including our new full-scale humanoid robot hand-arm system, the iCub hand-arm system, and robot arms such as Atlas, H20, and Hubo2+/Sophia, as well as robot hands like Shadow, DEXMART, and SEOULTECH. Our new robot hand-arm system has a functional anthropomorphism of 23.31%, while iCub's is 35.96%. Atlas, H20, and Hubo2+/Sophia robot arm have functional anthropomorphism of 24.79%, 12.82%, and 29.64%, respectively, while Shadow, DEXMART, and SEOULTECH robot hands have functional anthropomorphism of 36.96%, 48.91%, and 25.33%, respectively. The proposed method is an effective tool for designers to quantitatively compare the human-likeness of both new and existing robotic hand-arm systems and identify possible improvements in the design. We further conducted a series of kinematic experiments to ascertain the kinematic capabilities of our humanoid robot hand-arm system.