Reinforcement Learning-Based Fault-Tolerant Control of Uncertain Strict-Feedback Nonlinear Systems With Intermittent...

Area of Science:

• Control Engineering
• Artificial Intelligence
• Nonlinear Systems

Background:

• Fault-tolerant control (FTC) is crucial for maintaining system performance under actuator failures.
• Nonlinear systems with uncertainties pose significant challenges for traditional control strategies.
• Actuator redundancy offers a pathway to enhance system resilience.

Purpose of the Study:

• To develop a novel reinforcement learning-based adaptive FTC scheme for nonlinear strict-feedback systems.
• To address challenges posed by nonlinear dynamics, uncertainties, and actuator failures.
• To ensure robust tracking control and system stability.

Main Methods:

• A learning-based switching function technique was developed to manage actuator groups based on a performance index.
• The optimal tracking control problem (OTCP) was transformed into an optimal regulation problem using adaptive feedforward controllers.
• A reinforcement learning algorithm minimized objective functions related to Hamilton-Jacobi-Bellman (HJB) estimate errors from neural network (NN) approximations.

Main Results:

• The proposed FTC scheme effectively mitigates the impact of faulty actuators through automated actuator group steering.
• The reinforcement learning algorithm successfully minimized estimation errors without requiring value or policy iterations.
• Tracking objectives were achieved, and all closed-loop system signals were proven to be bounded.

Conclusions:

• The developed reinforcement learning-based adaptive FTC scheme provides a robust solution for nonlinear systems with actuator redundancy.
• The method guarantees system stability and performance, even in the presence of nonlinear dynamics and uncertainties.
• Simulation results validated the theoretical findings, demonstrating the effectiveness of the proposed control strategy.