LPV-MPC Control and Self-Tuning Feedback Gains for the Trajectory Tracking of a Quadqopter UAV
Reaching applications of artificial intelligence in control systems to autonomously drive unmanned aircraft is an increasing area of interest in engineering. The purpose of this research is to investigate and apply machine learning along with control systems algorithms to autonomously track the flight trajectory of anunmanned aerial vehicle (UAV) based on a given path.In this sense, using a combination of Feedback Input State, Model Predictive Control (MPC) and Gradient Descent, this study analyzes the capabilities to track the position and attitude of a UAV. Moreover, the use of gradient descent to generate accurate feedback gains was found to play the greatest role in providing the UAV a better performance in autonomous flight.Hence, this study definitively relates the use of fitted gains computed through machine learning and later profited by the feedback controller. Further studies are needed to stablish thresholds in the control input, required in a real prototype implementation.
Keywords - UAV, MIMO, MPC, Runge-Kutta, LPV, Feedback Input state, Machine Learning, Gradient Descent.