Betterment Process of Contact Motion by Learning Method

Abstract

A contact motion of a robotic manipulator to a workpiece is a fundamental operation which occurs frequently. The contact motion is a very short time phenomenon and a so-called impact force arises. If we use only a feedback control scheme, it is difficult to realize a desirable contact motion. In this paper, a new control method of the contact motion by a learning control is proposed. The learning control is based on an iterative operation, and is one of the feed-forward controls which do not need precise values of system parameters or a model of the impact force. Therefore it is useful for the contact motion control. Proposed learning controller has two features as follows. Firstly, it observes an input signals and output signals at the different period so as to improve a transient response of a system. Secondly, it can learn an initial state of the system, which includes an approaching velocity of the manipulator. By learning the approaching velocity, an fast contact motion can be achieved, because the magnitude of impact force depends on the approaching velocity. Simulation results are shown to verify an effectiveness of the proposed method.