Control of an UAV using LPV techniques

Small quadcopters have demonstrated to be the perfect testing bench for addressing the autonomous flying problem. Unmanned Air Vehicles (UAVs) represent a very challenging research topic that requires from different disciplines such as electronics, computer vision, geolocalization, control and planning. The dynamics of a quadcopter presents six degrees of freedom: three translations and three angular movements, but only four control commands (the overall thrust force and three moments), leading to an underactuated control problem. Thus, this project tackles the position and heading, yaw angle, control problem for a quadcopter, making use of a cascade control structure. This cascade controller is composed by an outer position controller which returns the needed thrust force and roll-pitch angles to achieve a desired position, and an inner attitude controller that tracks the previous roll-pitch angles plus an external yaw angle, inducing three possible moments in the quadrotor structure. After a model derivation stage, the inner attitude controller was designed by means of a Linear Parameter-Varying (LPV) approach, such that minimizes the Linear Quadratic Regulator (LQR) problem, defined as a set of Linear Matrix Inequalities (LMIs) for its vertex systems. Furthermore, using the same technique an Unknown Input Observer (UIO) was designed for decoupling possible disturbances affecting the attitude subsystem. For the position control, a feedback linearization structure was used, analytically deriving the control actions. Also, an outer velocities controller was designed by means of feedback linearization. All the proposed control variations were validated in simulation, including the response to disturbances of the vulnerable attitude subsystem. Besides, the system response with a velocities path planner was tested. Finally, the inner attitude controller was implemented in a real platform demonstrating its capabilities in flight conditions.Peer reviewe