Abstract
In this paper, we propose a new approach for a humanoid balance control. In previous research, a balance control of a humanoid was regarded as a method which absorbs the difference of assumed environment model and real environment. In contrast, we focus on generating gird motions using prediction of time series of external force. We introduce a disturbance preview control method to generate gird behavior against external impact. In order to reduce ZMP error at the moment of external impact, we control desired COG trajectory by using a disturbance preview control that takes future information of an external force represented as velocity disturbance of COG. We show numerical simulation results and clash experiments using a small humanoid robot HOAP-2.
