A method for evaluating humanoid robots using anthropometric and biomechanical data

Abstract

Humanoids are generalists, often lacking specific tasks or goals that can be used to evaluate or compare robot performance. Evaluating humanoids through a standardized, generalizable approach can yield a measure of robot performance on yet to be defined tasks. Anthropometric and biomechanical data from human factors design guides can be used as a baseline for evaluating humanoid performance in human-engineered environments. Past humanoid designs have used anthropometric data to define their design parameters, but no standard exists for evaluating humanoid performance. By creating a standard for humanoids, a robots capabilities can be characterized without explicitly testing every conceivable task. NASAs Valkyrie robot was tested against the anthropometric, range of motion, mass property, and strength data from NASAs Constellation Program Human-Systems Integration Requirements. The robot model offered anthropometric and range of motion data, and hardware tests were used to collect mass proportion and strength data. According to the standard used, Valkyrie fell within acceptable bounds for 59% of anthropometry, 64% of range of motion, 13% of mass property and 35% of criticality 1 strength measurements. Evaluating humanoids against the standards upon which human systems are built yields a quantitative measure of robot performance in human-engineered environments. This proposed approach can be generalized to evaluate other humanoids against a common standard. Failures or weaknesses in the humanoid design can be identified and reevaluated over time. Metric performance can be improved through advancements in robot control, operator skill, and hardware redesign.