A First Step of Humanoid's Walking by Two Degree-of-freedom Generalized Predictive Control Combined with Visual Lifting Stabilization

Biped locomotion created by a controller based on Zero-Moment Point (ZMP) known as reliable control method looks different from human's walking on the view point that ZMP-based walking does not include falling state. However, the walking control that does not depend on ZMP is vulnerable to turnover. Therefore, keeping the walking of dynamical motion stable is inevitable issue for realization of human-like natural walking-we call the humans' walking that includes turning over states as "natural." In our research group, walking model including slipping, impact, surface-contacting and point-contacting of foot has been developed. Although "Visual Lifting Stabilization" (VLS) strategy has been also proposed in order to enhance standing robustness and prevent the robot from falling down without utilizing ZMP, the torque generation strategy making lifted-leg step forward is derived by trial and error. Therefore, as a first step to realize humans' walking, this paper explores two degree-of-freedom generalized predictive control (GPC) method in order to generate the torque making lifted-leg step forward. Simulation results indicate that this strategy helps stabilize bipedal walking even though ZMP is not kept inside convex hull of supporting area.