Fault Tolerant Control for Nonlinear Systems Based on Adaptive Dynamic Programming with Particle Swarm Optimization

This paper develops a fault tolerant control (FTC) scheme based on adaptive dynamic programming(ADP) employing the particle swarm optimization (PSO) for nonlinear systems with actuator failures. Using the well-known ADP method, the solution of Hamilton-Jacobi-Bellman equation (HJBE) is approximated by constructing a critic neural network (CNN) which is trained by the PSO algorithm. Compared to the existing gradient descent-trained CNN, the PSO-trained CNN has a higher success rate in solving the HJBE. In order to eliminate the impact of the actuator failure, the ADP-based FTC strategy is developed to guarantee the closed-loop system to be ultimately uniformly bounded (UUB). Finally, a simulation example is provided to demonstrate the effectiveness of the developed method.