Gravity Compensation on Humanoid Robot Control with Robust Joint Servo and Non-integrated Rate-gyroscope

Abstract

While gravity plays a significant role in legged motion controls, it also affects robot's each joint and prevents the decoupling of the upper layer controller of macro states such as COG from the lower layer joint controllers. This paper proposes two techniques to compensate for the effect of gravity in the joint controllers. Firstly, a robust joint servo system with two-degree-of-freedom control cascaded to PD control is presented. It robustly compensates stationary error which is mainly due to gravity. The minor loop featuring strongly stable PD feedback is regarded as a new plant of the major two-degree-of-freedom control loop and facilitates the nominal model identification. Secondly, the transient response of the trunk attitude is improved by a non-integrated type of feedback of rate-gyroscope outputs. Since it feeds back the output signal without estimating the trunk attitude explicitly, it doesn't suffer from amplified error due to drift and noise in the course of numerical integration of signals. Both are even applicable to systems with less adjustable control parameters, such as embedded PD controllers