Nonlinear tracking and aggressive maneuver controllers for quad-rotor robots using Learning Automata

In this paper will be presented a novel approach to design and optimization of attitude and path tracking controller for quad-rotor robots based on Learning Automata algorithm. The proposed method has superior features, including easy implementation, stable and fast convergence characteristic, and good computational efficiency while compared with others learning approaches. The quad-rotor robots have become increasingly important in recent years as platforms for both research and commercial applications. For this reason, we propose a set-up that can be used to the training and evaluation of the quad-rotor controllers in realistic conditions. First, the vehicle dynamics and mathematical model are presented. The attitude and path tracking control strategies for the robot are formulated. Next, the method used to adjust the parameters of the controllers through the Reinforcement Learning algorithm is discussed. The simulation environment is composed by 2 host computers where one host executes the control loops and the training algorithm implemented in Matlab/Simulink. The other host runs the quad-rotor model using the X-Plane Flight Simulator. The two hosts communicate using UDP (User Data Protocol) over a standard Ethernet wired network. Using this framework is possible tuning the parameters of the controllers for a nonlinear aircraft which interacts with the environment, taken into account, the aerodynamics effects present during the quad-rotor flight. This simulation environment showed to be a good set-up for researchers to investigate the application of learning algorithms to adjust the control laws for several flight maneuvers and conditions. Finally, the results obtained from the controllers adjusted by the Learning Automata algorithm are presented.