Abstract
In this research, we propose a method to acquire expert's control characteristics by CMAC learning algorithm which is a method of neural networks with teaching signals of the expert operation. In this method, we propose an efficient CMAC learning algorithms using the adaptive learning gain controlled dynamically. Furthermore, we aim to acquire multiple control characteristics of expert using by multiple CMAC learning algorithms, and to execute the experiment of automatic operation based on the control skill of expert acquired in this method.
Moreover, in order to confirm the efficiency of this method, we used a radio controlled car on the market as a learning objects. In this experiment, an expert controls a radio controlled car with the motion image obtained by the ceiling camera. At this time, some voltage values obtained by the radio controller is used as teaching signals of CMAC learning algorithm. By this experiment, we confirmed that the expert's operation characteristics is obtained by the proposed method.
