Observer-Based Decentralized Adaptive NNs Fault-Tolerant Control of a Class of Large-Scale Uncertain Nonlinear Systems With Actuator Failures
针对一类大尺度不确定非线性系统，提出一种基于观测器的分散自适应神经网络容错控制策略，仅依赖输出信息，无需精确参数，能处理执行器故障并保证系统稳定。
In this paper, we are concerned with the fault-tolerant control (FTC) issue for a class of large-scale uncertain multi-input and multioutput nonlinear systems. The features of such class of systems are that virtual control variables are in nonaffine pure-feedback form and actuator failures consist of both lock-in-place and loss of effectiveness. We develop an adaptive observer to reconstruct unavailable state information for this class of systems taking advantage of the universal approximation property of neural networks (NNs). And then, an observer-based decentralized adaptive FTC strategy is designed recursively by combining backstepping methods with NNs, FTC theory as well as the dynamic surface control (DSC) technique. The superiorities of this proposed strategy are that it is only dependent on output information of the system and there is no requirement for accurate parameters of the system. It is also hardly inevitable that repeat differentiation calculations of virtual functions with the help of DSC technology. In theory, the stability of the resulting closed-loop system is rigorously investigated, and it is proven that all signals remain uniformly ultimately bounded and tracking errors converge to a small neighborhood around the origin by suitable choice of design parameters. Finally, simulation results, both practical and numerical examples, are illustrated to verify the feasibility of the theoretical claims.