Neural network-based observer combined with nonsingular terminal sliding mode control for QUAV tracking: Experimental...

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Quadrotor trajectory tracking is challenged by state uncertainty, dynamic disturbances, and unknown system dynamics.
• Existing control methods may struggle with the complex, real-world conditions encountered by quadrotors.

Purpose of the Study:

• To develop an advanced control framework for rapid and accurate quadrotor trajectory tracking.
• To enhance quadrotor robustness against uncertainties and external disturbances.
• To validate the proposed control strategy through simulations and experiments.

Main Methods:

• Modeling quadrotor dynamics with inherent uncertainties and disturbances.
• Designing a state and disturbance observer for estimating unknown states and rejecting disturbances, proven via Lyapunov theory.
• Implementing a nonsingular terminal sliding mode control (TSMC) for error stabilization, with exponential convergence.
• Utilizing an adaptive neural network (ANN) to approximate unknown system components.

Main Results:

• The state observer demonstrated exponential convergence of estimation errors.
• The TSMC framework ensured exponential convergence of sliding surfaces to zero.
• Simulations and experiments confirmed the proposed adaptive neural network-based TSMC strategy's effectiveness.
• The control strategy showed significant robustness and rapid trajectory tracking capabilities.

Conclusions:

• The integrated adaptive neural network, state observer, and nonsingular terminal sliding mode control framework effectively addresses quadrotor trajectory tracking challenges.
• The proposed method provides a robust solution for quadrotor control in uncertain and disturbed environments.
• Experimental validation confirms the practical applicability and high performance of the developed control strategy.