Skill acquisition and rule extraction method of expert´s operation

Generally, CMAC (Cerebellar Model Articulation Controller) is used for learning of robot action, because of high speed and low memories. However, results obtained by learning are hard to understand for human because they are numeric data. We research skill acquisition of an operator by using CMAC in our laboratory. In this research, we conducted the rule extraction by Fuzzy Neural Network (FNN) from the learning result of Adaptive Learning CMAC. By using this method, the extracted rule makes easy to understand for human. In addition, an operator´s skill is acquired from the operators whose a level of skill is different. By comparing the characteristic of expert´s operation with that of amateur´s one, we extract an expert´s particular skill. In this experiment, we use a radio controlled car operated by human, and report the effectiveness of the proposed method.