Fixed-Time Robust Neural Learning Control for Nonlinear Strict-Feedback Systems With Prescribed Performance

Abstract

The fixed-time robust neural learning control for nonlinear strict-feedback systems with output constraint and unknown dynamics is investigated in this article. The system nonlinearity is identified using neural network (NN) while the prescribed performance design is employed to avoid the output constraint to be violated. Considering the intelligent approximator only working in an active domain, the smooth switching mechanism is introduced to indicate its effectiveness. Based on the designs of neural approximation and switching signal, a robust adaptive controller is constructed where the NN works in the active domain and the fixed-time robust design works outside of that domain. Especially for the neural update law, the estimation error is constructed based on a state observer and a serial-parallel identification system, even though the real system nonlinearity is not available. The effective neural approximation is achieved while the error signals are ensured to be practical fixed-time stable. Simulation tests are ultimately conducted to demonstrate the validity of the design.