Zustandsschätzung und prädiktive Trajektorienfolgeregelung für Multikopter in Inspektionsanwendungen
- State estimation and predictive trajectory tracking control for multirotors in inspection tasks
Konrad, Thomas; Abel, Dirk (Thesis advisor); Horn, Martin (Thesis advisor)
Dissertation / PhD Thesis
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020
The subject of this thesis is the investigation of the main control engineering problems for enabling automated flights of unmanned multicopters in inspection applications, namely state estimation, and trajectory tracking control. Furthermore, the development of a suitable concept for a communication infrastructure, functional safety, and hardware design of an experimental multicopter system is a further focus. The objective is to develop different methods in the fields of navigation and trajectory control and to evaluate them regarding their use in outdoor inspection applications. Three different integration concepts of Inertial Navigation (INS) and Global Navigation Satellite Systems (GNSS) are investigated concerning accuracy, availability, and bandwidth requirements. In addition to the use of a precalculated GNSS positioning solution in a loose coupling, the tight coupling based on the raw measurement data of individual satellites is of particular focus. With explicit correction of measurement delays, INS and GNSS are fused using multi-rate extended and unscented Kalman filters. Extensive studies with experimental data show that the tight coupling provides a requirement-compliant state estimation with low gradients even with short-term GNSS outages. For precise and efficient trajectory tracking control of the nonlinear multicopter system, the combination of flatness-based two-degree of freedom control and predictive optimization for trajectory generation is investigated. The superimposed optimization based on a constrained quadratic program generates feasible and four times continuously differentiable trajectories at runtime, which are executed by nonlinear feedforward control. The exact state linearization based on flatness is presented in a variant reduced by the rotational dynamics in addition to the full one. By means of simulative and experimental investigations it is shown that the reduced linearization shows better robustness properties and similar performance in the dynamics relevant for inspection applications. For validation of the overall system, an industrial radio mast is flown with the experimental quadcopter system. Despite the limited computing capacity, the selected methods allow high accuracies to be achieved in automatic flight, thus meeting the application-specific requirements.
- DOI: 10.18154/RWTH-2020-10107
- RWTH PUBLICATIONS: RWTH-2020-10107