Experiments on torque pattern learning for static balance with respect to unknown periodic external force

This paper considers torque pattern learning for balance control with respect to un- known periodic external forces. To cope with uncertain factor of environment, the feedback control is essential. For balancing problem, we propose a control method based on the ground reaction force feedback. When external force is periodic, the torque pattern becomes regular. Learning this torque pattern, the feedback information is less important. We show this learning process by the experiment. For the static balance, we proposed a control method under unknown constant external force and a learning method for unknown periodic external force based on the above scenario 1, 0, 0) . However, in these works, the period of periodic external force had to be known. In this paper, we extend the above methods to cope with unknown periodic external force with unknown period. As a result, the static balance is kept with acquired torque pattern that contains no feedback information of the ground reaction force. This paper is organized as follows: in the second sec- tion, we review our studies for static balance control and its torque pattern learning, and clarify the problems we left. In the third section, we propose a new method to solve it. The fourth section shows robot experiments and the last fifth section gives concluding remarks.