Research On Neural Network Control Of A Class Of Pure-Feedback Systems With Output Constraints And Prescribed Performance

In the control process of early industrial production,linear system control methods were widely used,but with the development of the times and changes in control requirements,as well as the nonlinearity and uncertainty of the system,people's control requirements can't be satisfied by linear system control.Nonlinear system control was proposed and began to be used extensively.In the nonlinear system,from the perspective of mathematical structure,the pure-feedback system with non-affine input is more general than the strict-feedback nonlinear system with affine input,and its application range is wider in practice.In real life,due to the physical failure of the system,saturation,and some of its safety specifications,the output of the system will be limited.In addition,only the steady-state performance of the system can be guaranteed by the control of nonlinear system generally,the system's transient performance and steady-state performance are not considered at the same time.Therefore,considering the above factors,a class of neural network control problem of pure-feedback system with output constraints and prescribed performance is studied in this article,it is based on the backstepping method.The research content is as follows:1.The neural network control problem is considered for the pure-feedback system that has output constraints and input saturation.First,the non-affine pure-feedback system is transformed into a nonlinear system with explicit input.The system controller is constructed based on Lyapunov's second method and backstepping method,also consider the barrier Lyapunov function.Finally,the control system stability and signal boundness are illustrated by using the stability analysis.The justifiability of the controller is demonstrated by two simulation results.2.The neural network control problem is studied based on event-triggered for the pure-feedback system with finite-time prescribed performance.The system form is transformed,from a non-affine pure-feedback system to a nonlinear system with explicit input.Then the neural network is adopted to approximate the unknown function,and the system controller is constructed based on an improved event-triggered control strategy.An improved event-triggered control strategy is proposed to obtain a larger threshold,and also the proposed controller can avoid the Zeno-behavior.Boundness of all dynamic signals can be ensured by the controller design in this chapter,the tracking error of the system can be limited to a pre-given boundary by prescribed performance in a finite time.The feasibility of the controller is demonstrated by two simulation results.3.The neural network control problem is studied based on event-triggered for the pure-feedback system with output constraints.A pure-feedback system with non-affine inputs is transformed into a nonlinear system that has the explicit inputs by applying the mean value theorem.In addition,the unknown system nonlinear function is approximated by utilizing the radial basis function neural network.The tracking error of the controller is limited to a small constraint boundary by using the positive obstacle Lyapunov function.An adaptive controller for the pure-feedback systems is established,which based on the backstepping control theory and event-triggered control theory,it can ensure all signals are bounded and avoid the Zeno-behavior.The accuracy of the controller is proved by two simulation results.