Variable Universe Fuzzy Control Algorithm

With the rapid development of science and technology, modern industries and projects become more and more complex. These systems have nolinear characteriscs of multivariable, strong-coupling, disturbance and the exact mathematical models of the controlled plants are hardly built. So, the design methods of trational control that based on exact mathematical models can not be applied to these systems. Fuzzy controller is widely used by its merits, such as, simple structure, strong robust, no mathematics model of the controlled object. But fuzzy control also have the low control accuracy, rule explosion and other issues, variable universe fuzzy control algorithm solves these problems to a certain extent.In traditional variable universe control, contraction-expansion factors are based on function models. The design of this kind of variable universe fuzzy controller faced the choice of function models of contraction-expansion factors and parameters of models and the contraction-expansion change of universe was described by a simple function model. Furthermore, contraction-expansion factor of input universe was described by single variable error or change ratio of error, the interaction of error and change ratio of error, will cause adjustment of universe repeatedly. In allusion to these problems, this paper makes some exploratory reseach in variable universe fuzzy control algorithm, attempts to realize the adjustment of the universes by fuzzy inference and fuzzy neural network respectively. The main contents of this paper are as follows:1. This paper mainly makes a study of trational variable universe fuzzy control algorithm and does research on the feasibility of contraction-expansion factors of input universes are decided by the error and the change of error.2. In allusion to the problem how to choice function models of contraction-expansion factors and their parameters, the paper designs a variable universe fuzzy control algorithm based on fuzzy inference by analysing the change law of universe. In the algorithm, the idea of fuzzy inference is introduced to describe the contraction-expansion change of universes. It uses a fuzzy controller of contraction-expansion factors to adjust the input and output universes of system online to avoid selecting function models of contraction-expansion factors and their parameters.3. To the generic variable universe fuzzy control system, contraction-expansion factors realized by fuzzy inference is convenient. However, to the complex or high precision system, it has the problem that it is hard to confrim fuzzy rules of contraction-expansion factors and fuzzy partition, which may bring on unideal effect. In allusion to the problem, this paper presents that fuzzy neural network is used to realize the function of contraction-expansion factors. The algorithm has the language of fuzzy inference and learning ability of neural network, which can optimize fuzzy partition of contraction-expansion factors fuzzy controller and make the change of universes in reason.The inverted pendulum system is a natural instability system with multivariable, nonlinear and strong-couping, which is the ideal means to test the new control theory. So we choice the one level inverted pendulum, the triple inverted pendulum and the quadruple inverted pendulum to test the proposed algorithms in this paper respectively. At the same time, the proposed algorithms compare with variable universe fuzzy control algorithm based on function model proposed by Prof. Li Hongxing. Experiments prove the feasibility and validity of these algorithms.