Time-independent, spatial human coordination for humanoids

The work presented here focuses on humanoid gait generation. More precisely we deliberately separate the motion generation in two distinct parts: the coordination and the stability. The goal is to highlight the human coordination, and then reduce its complexity to a minimal set of relations (twelve gait descriptors with respect to an unique gait parameter). Polynomial approximations of these relations are done and implemented into a kinematic simulation of an anthropomorphic humanoid model coming from a real humanoid robot. This simulation allows the leg coordination through the contact with the environment. After the validation of the coordination relations, they are injected into an actual humanoid robot NAO. We choose the small humanoid NAO from Aldebaran society because its size and set up permit to keep the gait coordination (in three dimensions) part primarily in gait generation by comparison with the stability part (in the frontal plane). The results are promising and authorize to plan an application of this work on a more complex humanoid, plus the study of other gaits, and more important on the gait transition.