CENTRALIZED MULTIVARIABLE LQG CONTROL SYSTEM FOR LONGITUDINAL AND LATERAL SPEED HOLD AUTOPILOT FOR THE AR.DRONE 2.0 QUADCOPTER
In this work it is presented the design procedures and experimental results of a centralized multi-variable LQG control system for longitudinal and lateral speed hold autopilot for the AR.Drone 2.0 quadcopter. The main contribution is that instead of assuming that the longitudinal and lateral dynamics are completely decoupled, the quadcopter is modeled as a coupled multivariable state-space system with transport time-delay. The system identification procedure, by extended recursive least-squares estimation, is done directly in the state-space form and a detailed description of the equations derived for this project is given. The LQG design is aided by analysis on the system’s step-response tests in the time domain and is based on a non orthodox Kalman filter design, dual to the LQR. The proposed speed hold autopilot is evaluated using the system’s model and then applied to the real process. The experimental platform used was a free add-on toolbox for Matlab/Simulink, also used as a benchmark control system with a ready-to-fly example of decentralized proportional control system for the AR.Drone 2.0. Results summarized in a table of integral performance indexes and a discussion over the formalism of the LQG method and its applications to flight control systems concludes the contributions of this work.