Adaptive Dynamic Programming-Based Fault Tolerant Control for Nonlinear Systems with Multiple Failures

Abstract

This paper investigates the fault tolerant control (FTC) scheme against multiple failures (i.e., both actuator and sensor failures occur simultaneously) for nonlinear systems via adaptive dynamic programming (ADP). A descriptor observer is designed to estimate the system states and failures concurrently. Next, a critic neural network (NN) is used to solve the Hamilton-Jacobi-Bellman (HJB) equation for the nominal system, i.e., the failure-free system, and the approximate optimal control policy is obtained. The FTC law is achieved by combining the estimated system states and failures with the approximate optimal control policy. By using the Lyapunov's direct method, we conclude that the closed-loop system is uniformly ultimately bounded. An example is employed to illustrate the effectiveness of the present FTC method.