Nowadays, unmanned aerial vehicles (UAV) are brought to accomplish an ever-growing variety of missions in the presence of challenging obstacles. Interest in simple and systematic methods to handle state and input constraints is raising rapidly. Among others, this led to the development of the explicit reference governor (ERG). This master thesis’ work develops a complete and fully functional controller for small UAVs, including the ERG in the navigation part of its architecture. Besides the actual implementation, contribution of the present work is to demonstrate – for the first time – that an ERG augmented controller effectively navigates in a laboratory environment with arbitrary positional constraints. The first step develops a control layer for the UAV stabilization through a cascade scheme. An inner loop, embedded in the UAV, controls its attitude, while an outer loop, implemented on a ground-based computer, controls its position, the position and attitude being measured by a motion capture system. The controller accounts for the UAV dynamics by means of a state space model. Since motion capture solely provides position and attitude, a Kalman filter, tuned according the experimental configuration, filters position and estimates velocity. Preparatory experiments with a pendulum assess performance. In a second step, a navigation layer is supplemented to enabl obstacle avoidance through the ERG. After adequate constraints are defined, its original formulation is extended and specialized. In a third step, the architecture – combining the primary control, Kalman estimator and ERG – is implemented in a Python program. In a last step, several experiments are carried out with constraints representing multiple vertical walls. The UAV response is logged and assessed for arbitrary reference positions, violating or not the constraints. The results confirm that the ERG modified reference behaves well navigating the UAV toward the reference and preventing collision. Independent tests show the system robustness and reliability. This thesis’ work successfully implements in practice the ERG concept. The experiments demonstrate the use of ERG as a systematic tool to provide constraint handling capability to a small UAV. Besides this achievement, the products of this work includes a documented, python program which is fast and computationally-efficient, serving as a solid basis for further application. Next steps include performance improvement through Lyapunov function optimization, evaluation of alternative ERG forms, testing of other rotorcraft types and introduction of more and other forms of obstacles.
Constrained Control of a UAV in a World with Obstacles
2018-01-25
oai:zenodo.org:1170445
Hochschulschrift
Elektronische Ressource
Englisch
DDC: | 629 |
Guaranteed infinite horizon avoidance of unpredictable, dynamically constrained obstacles
British Library Online Contents | 2012
|OPTIMAL LONGITUDINAL CONTROL PLANNING WITH MOVING OBSTACLES
British Library Conference Proceedings | 2013
|