Prescribed Performance Fault-Tolerant Control for Uncertain Nonlinear MIMO System Using Actor-Critic Learning...

Area of Science:

• Control Engineering
• Nonlinear System Dynamics
• Machine Learning in Control

Background:

• Fault-tolerant control is crucial for system reliability.
• Prescribed performance ensures bounded errors and fast convergence.
• Uncertain nonlinear systems pose significant control challenges.

Purpose of the Study:

• To develop a learning-based fault-tolerant controller for uncertain nonlinear multi-input and multi-output (MIMO) systems.
• To achieve asymptotic stability and satisfy prescribed performance constraints.
• To avoid the need for a priori knowledge of system dynamics.

Main Methods:

• Introduced a novel error transformation function for constrained dynamics.
• Utilized an actor-critic learning structure with a continuous-time performance index.
• Employed a critic network for performance index approximation and reinforcement learning.
• Developed an action network-based controller using robust integral of the sign of error (RISE) feedback.

Main Results:

• The proposed controller guarantees asymptotic stability.
• Tracking errors converge to zero while maintaining prescribed performance.
• Lyapunov stability analysis confirms the theoretical guarantees.
• Simulation results demonstrate the controller's feasibility and effectiveness.

Conclusions:

• The learning-based fault-tolerant control scheme effectively addresses prescribed performance for uncertain nonlinear MIMO systems.
• The approach offers robustness against system uncertainties and faults.
• This method provides a viable solution for complex control applications requiring guaranteed performance.