Global Finite Time Neural Network Control Method for Nonlinear Strict Feedback Systems

Abstract

The invention relates to a nonlinear strict-feedback system and global finite-time neural network control method, belongs to the field of nonlinear system control and the field of neural network control, and is used for solving a problem of nonlinear strict-feedback system and global neural network control with unknown dynamics. The method is based on a backstepping framework; a switching mechanism is introduced to implement switching between neural network control in an active approximation domain and robust control outside the approximation domain; meanwhile, a neural network weight is updated on the basis of a tracking error and a modeling error; learning performance of a neural network is improved; on the basis, a robust design scheme is given out; finite-time convergence of the systemtracking error can be implemented. According to the invention, one type of strict-feedback system and neural network control with unknown dynamics is guaranteed to always work in the active approximation domain, global stability of a closed-loop system can be implemented, learning performance of the neural network for unknown dynamics is improved, and the performance requirement of the actual control problem is ensured.

Description

Technical field [0001] The invention relates to a nonlinear strict feedback system global finite time neural network control method, which belongs to the field of nonlinear system control and neural network control. Background technique [0002] The control problem of nonlinear strict feedback system has become a research hotspot of scholars at home and abroad. Many practical control problems, such as hypersonic vehicle reentry control, flexible manipulator control, MEMS gyroscope control, and quadrotor control, are required Use a strict feedback system to describe. Neural network can approximate unknown dynamics and model uncertainty, and is widely used in aircraft control. However, most of the current methods assume that the neural network can always be effectively approximated in the entire area for controller design, which makes the closed loop The system can only guarantee semi-global stability, which is difficult to guarantee in practical applications. "Global neural dyna...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the neural network control method of the current aircraft seldom considers whether the neural network approximation is always effective during the control process, the present invention designs a global finite time neural network control method of a nonlinear strict feedback system, which can realize effective The switching between the neural network control in the approximation domain and the robust control outside the effective approximation domain ensures that the neural network works in the effective approximation domain and realizes the global stability of the closed-loop system. At the same time, the weight of the neural network is updated based on the tracking error and modeling error to improve Based on the learning performance of the neural network, a robust design scheme is given, which can realize the finite time convergence of the system tracking error

Images