Adaptive Neural Network Control for Fixed-Wing UAV With Disturbance Observer Under Switching Disturbance

Area of Science:

• Control Engineering
• Aerospace Engineering
• Artificial Intelligence

Background:

• Fixed-wing unmanned aerial vehicles (FUAVs) face challenges from unmodeled dynamics and time-varying switching disturbances (TVSD).
• Accurate modeling and control are crucial for FUAV stability and performance, especially during flight area changes.

Purpose of the Study:

• To develop an adaptive neural network control strategy for FUAVs robust to unmodeled dynamics and TVSD.
• To improve the estimation of TVSD and unmodeled dynamics for enhanced control performance.

Main Methods:

• A switching augmented model (SAM) and parameter adaptation (PA) technique to describe and estimate TVSD.
• A disturbance observer (DO) to estimate unmodeled time-varying disturbances.
• Radial basis function neural networks (RBFNN) to approximate unknown dynamics.
• An auxiliary system in DO form to enhance estimation performance.

Main Results:

• The proposed control strategy effectively estimates and compensates for both modeled and unmodeled disturbances.
• Sufficient stability conditions for the closed-loop switched system (CLSS) were derived.
• An illustrative example demonstrated the feasibility and advantages of the adaptive control strategy.

Conclusions:

• The developed adaptive neural network control offers a robust solution for FUAVs operating under complex and uncertain conditions.
• The strategy enhances FUAV control accuracy and stability, validated through simulation on an attitude model.