Abstract
A robot with a large number of degree of freedoms (DOFs) may have the flexibility to cope with disturbances and irregular factors exist in a real environment. In this paper, we propose a control method for such a robot where the prediction using k-nearest neighbors from past experiences (stored instances) of a similar situation is used to determine the control signal. We apply our method to the control task of a robot with complicated structure, driven by several elastic actuators. Experimental results show that the control of a robot with a large number of DOFs can be achieved by the proposed method.
