Adaptive Neural Network Control Of A Class Of Unknown Nonlinear Systems And Their Applications

Nowadays, nonlinear systems are broadly applied in many production process. Therefore, we must aim at the model of nonlinear systems and use nonlinear control theory to analysis this nonlinear systems. However, for nonlinear systems, it is difficult to obtain the precise mathematical model that is a major problem for us. Even if we can establish its mathematical model, the model is often too complex to make it difficult to achieve the desired control effect if we use traditional control method. In recent years, the rise of artificial neural networks, the research of numerical approximation methods and unknown parameters processing is developed quickly. Artificial neural network is used as a real predictor, and it is more widely used to solve the problem of unknown nonlinear system control. For neural network, it has simple structure, parallel processing, high fault tolerance capabilities, and it can adaptive learning. Thus, in the control field, artificial neural network demonstrates its strong vitality and broad prospects.In this paper, we aim to solve the nonlinear systems with uncertain parameters and their trajectory tracking control. For these different nonlinear systems, we need to achieve output constraints according to their actual condition. In this paper, a barrier Lyapunov function is proposed to cope with output constraint. To handle the system uncertainties, we apply adaptive neural network to approximate the unknown model parameters of these systems. In addition, the control law is designed by using backstepping, Moore-penrose pseudoinverse and other methods. Based on logarithmic barrier Lyapunov function, adaptive neural network control with both full-state feedback and output feedback is designed. When all states are known, that is, for the full state feedback, the controller design by using Moore-penrose pseudoinverse, and the output constraint is never violated; when only output is known and other states cannot be measured, that is, for the output feedback, the high gain observer is designed to estimate these unknown states, the signals of the closed loop system are semiglobally uniformly bounded(SGUB), the asymptotic tracking is achieved, and the multiple state constraints are never violated.For our paper, the total of three types systems controller are designed. First, a class of single input single output(SISO) nonlinear systems preliminary theoretical study, design proved by the stability of the controller system, and to achieve the output constraint; secondly, more complex robotic systems multiple input multiple output(MIMO) by application of research, and through the controller design proves system stability, and output constraint in never violated; finally, MIMO systems research ship, and proved by the controller design system stability, and ultimately achieve output constraint. The method of Lyapunov stability of the closed loop system is analyzed at the same time the entire state of the system volume to achieve semi-globally uniformly bounded control performance expected control effect and the controller also be verified through simulation section. And it confirmed the feasibility and effectiveness of the control algorithm in this paper.