Design of Predictive Control Structures to Track Trajectories for Multi-rotor Unnamed Aerial Vehicle

Abstract

This paper presents the synthesis of four model predictive controllers for a multi-rotor unmanned aerial vehicle to track trajectories. Three control structures are used to control the state variables: x, y, z positions and the yaw angle. For the first structure, a centralized control loop is used and two control strategies are proposed. The first strategy uses the outputs described above, while the second one increases the outputs with the state variables: roll and pitch angles. The second and third control structures use a non-centralized control loop. The second one splits the control loop into a cascade structure with a master and a slave loops. The last structure combines a master model predictive control strategy with a slave PD-P control combination. In both cases, the slave loop controls the new reference state variables roll and pitch angles, given by the master loop. The development of each controller is accomplished by changing the set points and is later tested by tracking a square trajectory. For a 1[m] step change in x or y position, the system response has a setting time of around 1.44[s], 1.45[s], and 2.74[s], with an overshoot of approximately 1.4%, 0% and 2.2% for each structure, respectively. For a 1[m] change in z position, the setting time is 1.94[s] and the overshoot is 2.88% for the first two structures. While for the last one, the setting time is 3[s] without overshooting. For a half turn change of the yaw angle, the setting time is 1.74[s] for the first two structures and 4.16[s] for the last one, all of them without overshoot. Finally, disturbances are included to test the robustness of the control strategies tracking a square trajectory. Based on these results, the conclusion is that the first and second structures have the best performance.