ABSTRACT

As the precise model of most practical mechatronics system cannot be obtained, the practice of typical control method is limited. Accordingly, numerous AI (Artificial Intelligence) control methods have been used widely. Fuzzy control and Neural Network control have been an important point in the developing process of the field. However, shortcomings exist in each of these methods. For example, the fuzzy control is unable to learn, and the physical meanings of learning result of the Neural Network control are not clear. Combining the strong points of above two methods, a new control method of FNN (Fuzzy Neural Networks) is explored in this paper. Additionally, a problem concerning the traditional network learning is discussed and a solution to such a problem is obtained subsequently. The new control strategy does not depend on the classical model and the algorithm is simple. The results of the experiments applying the new strategies are discussed.
Through different researches on control system, which model is unacquainted, the reasonableness, effectiveness and applying universality of the new control strategies is proved.