Intelligent Neural Network Robot Control Theory Methods

In this paper, the concept of intelligent neural network (INN) is proposed after reviewing the development of robotic control technology and neural network technology. The definition of INN is summarized. Several types of INN are presented and applied to control robotic manipulators. The proposed INN models can be divided into two classes, one is the neural network with the abilities of fuzzy inductive inference and self-learning, and the other is the neural network with the stability of whole system. In this paper, the abilities of fuzzy inductive inference and self-learning are realized by constructing fuzzy neural network (FNN) that combines fuzzy control theory and neural network technology together. And the stability of whole system is realized by designing robust neural network learning algorithm based on stability theories from the viewpoint of whole system stability. FNN uses neural network to realize fuzzy inference. This make it has ability of fuzzy inductive inference and ability of tuning the way of inference. Since the construction of FNN has clear meaning, the design and initialization of FNN are also very easy. With regard to the researches of FNN, at first, a fuzzy guassian basis neural network is proposed in this paper that uses guassian function as membership function. Secondly, for tuning the shape of membership function, a fuzzy B-spline basis neural network is presented after analyzing the characteristics of B-spline function. Thirdly, by fuzzifizing the space division method of cerabellar mode articulation controller , a fuzzy cerabellar mode articulation controller is proposed. Fourthly, a fuzzy wavelet basis neural network is developed based on the wavelet neural network. Finally, a recurrent fuzzy neural network is given by introducing feedback units into FNN. With regard to the researches of robust neural network, a robust neural network control system is designed based on the stability theories and robust control technology. In this system, a neural network with robust learning algorithm is used to approximate the uncertainties in system and the approximation error of neural network is counteracted by a robust controller. The stability of whole robust neural network control system is analyzed. All above proposed INN models are applied on the robotic trajectory tracking control to counteract the effects of the nonlinear, coupling anduncertain factors in the robotic system. For each proposed INN control strategy, simulation experiments are made. The results of the simulation experiments show that all proposed INN models have good intelligence. When they are applied to control robotic manipulators, they are effective for achieving good performance.