Disturbance Adapting Walking Pattern Generation Using Capture Point Feedback Considering CoM Control Performance

To ensure that a humanoid robot can walk in complex environments, a robust walking pattern that responds to disturbances is essential. When generating the walking pattern, zero moment point (ZMP)-based walking pattern is most widely used, as it generates a center of mass (CoM) trajectory to locate the ZMP in the supporting polygon of the robot. However, tracking the predesigned reference ZMP may not result in a flexible response to disturbances. Hence, walking patterns that use a capture point (CP) were introduced to improve the robustness of walking, which typically control to track the reference CP. In this study, a method is adopted that updates the reference ZMP with the feedback of the CP. This method was used in our previous study on a ZMP-based preview approach that accounts for CoM tracking performance. The stability analysis of the proposed approach is derived, and its physical meaning is explained. The proposed approach is implemented on a humanoid robot referred to as DYROS-JET, demonstrating its improved robustness against disturbances.