Abstract
Research in biomedical field has shown that light contact between a man and environment reduces sway of his body. That means a man uses sensed contact point and force to estimate and control his body attitude. In this study, a method to estimate attitude of a humanoid robot by using contact information as a man does is proposed. To realize light contact between a humanoid robot and wall, admittance control is implemented in the hand of the robot. Through light contact with wall, a humanoid robot observes gravity acceleration and normal vector of the wall and estimates attitude of the robot using extended Kalman filter. We realized light contact and experimental result showed effectiveness of the proposed estimation method. Legged locomotion with touching a wall is also realized, but the proposed estimation method could not be applied due to impulse force.
