Modal Gait Analysis: On the Use of POD and MAC to Extract the Fundamental Differences Between a Human and a Robot Walking

The capabilities of humanoid robots have greatly increased in recent years, especially when it comes to bipedal locomotion. However, there is a lack of standardized definitions of human-like gait for robots and little attention has been paid to how human-like the gaits of such robots are. In this contribution, we aim to contribute to such a standardized benchmark. We perform a Proper Orthogonal Decomposition of the kinematic gait data during one step of our humanoid robot Lola to extract the underlying gait pattern. We then use the Modal Assurance Criterion to quantitatively compare the gait pattern of Lola with the gait pattern extracted from human gait data. The results show that our approach allows for a quantitative assessment of gait similarity, and thus of human-likeness of gait. We observe some overall similarities in the gait patterns of Lola and the human subjects, however, the gait patterns between different human subjects are much more similar than when comparing them to Lola.