Adaptive support vector regression for UAV flight control.

Area of Science:

• Robotics and Control Systems
• Machine Learning Applications
• Aerospace Engineering

Background:

• Unmanned Aerial Vehicles (UAVs) require robust adaptive control strategies to manage complex dynamics and uncertainties.
• Traditional control methods may struggle with the nonlinear dynamics inherent in UAVs.
• Support Vector Regression (SVR) offers advantages over neural networks due to its global solution properties derived from quadratic programming.

Purpose of the Study:

• To develop and validate an adaptive control algorithm for UAVs utilizing Support Vector Regression (SVR).
• To address inversion errors and unexpected uncertainties in UAV models through an online adaptation mechanism.
• To ensure the stability of the control system using nonlinear system theory.

Main Methods:

• Offline identification of the UAV's inverse dynamic model using SVR (I-SVR).
• Offline identification of the inversion error compensation term using SVR (C-SVR).
• Development of an online adaptation algorithm for the C-SVR to handle real-time uncertainties.
• Stability analysis of the closed-loop system using the uniformly ultimately bounded (UUB) property.

Main Results:

• The proposed adaptive controller, based on SVR, effectively compensates for inversion errors and uncertainties.
• Numerical simulations demonstrate the controller's ability to maintain stability and achieve accurate control of the UAV model.
• The I-SVR and C-SVR approach provides a robust framework for adaptive UAV control.

Conclusions:

• The SVR-based adaptive control strategy is effective for UAVs, offering improved performance over traditional methods.
• The online adaptation of the compensation term ensures robustness against unmodeled dynamics and external disturbances.
• This research contributes a novel and stable adaptive control solution for unmanned aerial vehicles.