Robust Fuzzy Q-Learning-Based Strictly Negative Imaginary Tracking Controllers for the Uncertain Quadrotor Systems.

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Quadrotors (unmanned aerial vehicles) offer versatility but face challenges in stable trajectory control due to tuning difficulties, disturbances, and parameter uncertainties.
• Robust adaptive control is crucial for maintaining stable flight performance in quadrotors under varying conditions.

Purpose of the Study:

• To introduce a novel robust adaptive control synthesis methodology for quadrotor attitude and altitude stabilization.
• To address the limitations of traditional control methods in handling disturbances and uncertainties in quadrotor systems.

Main Methods:

• Utilizing feedback linearization (FL) to transform the nonlinear quadrotor system into a negative-imaginary (NI) linear model.
• Designing an adaptive control scheme that adjusts strictly negative imaginary (SNI) controller gains online using fuzzy Q-learning.
• Employing the negative imaginary (NI) theorem for stability analysis of the proposed controller and adaptive laws.

Main Results:

• The proposed fuzzy reinforcement learning-based SNI controller demonstrates superior performance compared to fixed-gain SNI, fuzzy-SNI, and PID controllers in numerical simulations.
• The adaptive control strategy effectively stabilizes quadrotor attitude and altitude, maintaining trajectory accuracy under disturbances.
• Stability proofs for the controller and adaptive laws confirm the robustness of the proposed methodology.

Conclusions:

• The novel robust adaptive control methodology effectively enhances quadrotor flight stability and trajectory tracking.
• Fuzzy reinforcement learning combined with SNI properties offers a promising approach for advanced UAV control systems.
• This method provides a stable and adaptive solution for quadrotor control challenges, outperforming conventional techniques.