Design and experimental evaluation of a human skill-based controller

Abstract

In recent years owing to the young generation does not willing to do hard works the skilled workers decrease rapidly. Leaded by the above situation, it is becoming a big issue to whom to transfer the techniques and know-how of these skilled workers. Therefore, the controller based on human-skill is needed. Moreover various neural networks (NNs) have been proposed these years, the appearance of these NNs make it easy to deal with nonlinear system and as time goes by they are going to play an important role in control engineering field. In addition there is a controller belongs to artificial NNs which is named as cerebellar model articulation controller (CMAC) has been proposed at 1975 by Albus. And a lot of controllers which is based on CMAC have been proposed. In CMAC the cerebellar is considered as a controller of the body motion system and CMAC is a kind of highly abstracted mathematic. The advantage of CMAC compared with other NNs is CMAC costs shorter learning time. Moreover, PID controller has been widely used in control engineering due to its simple structure and clearly physical meaning. And the way how to adjust PID parameters is well known by skilled workers. In this paper a human skill-based PID controller using CMACs has been proposed. On the basis of proposed method, the PID parameters are tuned by some CMACs which have learned human-skill. Furthermore, the experiment in order to illustrate the proposed scheme was performed by using a beam and ball control model.