Adaptive Neural Network Control For Nonlinear Systems With Constraints

In recent years,modern industrial processes have become more and more large-scale and complex,with highly non-linear characteristics and serious uncertainties.Moreover,many practical systems,such as process industry systems,robotic systems,power systems,etc.,require that the state or error of the system be constrained within a specific boundary,i.e.,"Hard Constraints" condition.If these constraints are not satisfied,the tracking performance and stability of the closed-loop system will be affected,and even serious safety accidents will occur.Therefore,the study of constrained control for uncertain nonlinear systems is of great significance both in theory and in practice.This thesis focuses on the problem of constrained control of non-strict feedback systems,and considers the influence of unmodeled dynamics caused by model errors and measurement errors,as well as the situation of input saturation,unknown control direction and unmeasurable state.Based on the existing research results,a new variable separation method is proposed,and the corresponding adaptive neural network controller is designed to achieve the control goal.The main contents are summarized as follows:1.For a nonlinear SISO system with full state constraints,the system with full state constraints is transformed into an unconstrained system by using non-linear transformation.Based on Taylor formula,a new method of variable separation is proposed.The relationship between the norm of state vector and the error function is established.Furthermore,an adaptive tracking controller is proposed by using the idea of backstepping.Under its control,the closed-loop system can not only ensure that all system states do not violate their respective constraints,but also when there are unmodeled dynamics and input saturation in the closedloop system,it can still run stably and the tracking error finally converges to a small domain around the origin.2.A class of non-strict feedback systems with unknown con trol direction and full-state constraints is considered.Based on the previous chapter,this chapter considers the influence of unknown control direction.By utilizing Nussbaum gain technology and dynamic surface control algorithm,the problem of "dimension explosion" caused by backstepping method can be avoided.We use neural networks to approximate the unknown nonlinear function terms defined and propose an adaptive neural network controller,which makes all signals in the closed-loop system are semiglobally uniformly ultimately bounded,and state constraints are not violated.Finally,Lyapunov stability theorem is used for stability analysis.3.For non-strict feedback nonlinear systems with unknown control coefficients and unmeasurable states,a linear transformation is used to transform the system with unknown control coefficients into a controlled system with control coefficients of 1.Combining backstepping method and Barrier Lyapunov Function,an adaptive output feedback controller is proposed.In the case of unknown disturbances,the closed-loop system can still run stably and the approximation error of each step can be constrained within the given boundary.