Master Thesis Presentation: Model Predictive Control of a Constrained Model Airplane

Jonas Schlagenhauf

University of Freiburg

Tuesday, January 31, 2017, 13:00 - 14:00

Room 02-012, Georges-Köhler Allee 102, Freiburg 79110, Germany


In this thesis four different linear and non-linear control schemes are designed and evaluated on a real-world embedded system, integrated into a model airplane. Objective of this work is to explore suitable approaches that allow to control the dynamic behaviour of a rotating constrained model airplane accurately and reliably in real-time. For this purpose three model types with varying degrees of physical explicitness were designed and identified: A linear black box model without explicit physical elements, a non-linear white box model as a close physical approximation of the real system and a grey box model as a middle ground, implemented in a linear and a non-linear variant. Based on the experimental evaluation and comparison of all model performances, three linear controllers were designed, including PID, LQR and MPC. In addition nonlinear MPC on the basis of the nonlinear grey box model was implemented as a fourth option. The controller performances were subsequently evaluated and compared, using in-system simulation and the experimental setup. From the evaluation of both models and controllers it could be concluded that a constrained model airplane can be sufficiently modelled by linear models to be controlled accurately and precisely. Comparisons of the controller performance showed that Nonlinear Model Predictive Control on the basis of the grey box model achieves significantly better results than linear, non-predictive controllers, while still being computationally viable. As a part of the HIGHWIND project, the results and implemented software infrastructure serve as a stepping stone for further research on rotational starts of tethered airplanes for applications in airborne wind energy generation.