Fault Tolerant Tracking Control Through Particle Swarm Optimization Based Policy Iteration

This paper focuses on fault tolerant tracking control (FTTC) problems for nonlinear systems with actuator failure. For fault-free system, the tracking control input is derived by the policy iteration. To deal with the difficulty in choosing the weight of critic neural network (CNN), the CNN is trained by the particle swarm optimization to instead the traditional gradient descent method. To handle the actuator failure, a fault observer is constructed to compensate the tracking control input, and then the fault tolerant tracking controller is derived. The developed FTTC scheme can guarantee the tracking errors to be uniformly ultimately bounded even if the system suffers form actuator faults. A simulation study is provided to illustrate the effectiveness of the designed FTTC scheme.