Solutions de navigation pour lanceurs de satellites : observabilité et nouvelles approches

Authors: Beaudoin, Yanick
Advisor: Desbiens, AndréGagnon, Éric; Landry, René Jr.
Abstract: This research evaluates new approaches to improve the performance of navigation solutions for satellite launchers. The first is the use of data from the reference trajectory. The idea of this approach is to weigh, throughout the mission, the confidence that the launcher follows the predicted trajectory in order to exploit the attitude data of the reference trajectory in the navigation solution. The second approach is to add a stochastic model of the launcher to the navigation solution. For this model, the acceleration and the angular velocity are represented by a random walk whose variance is adjusted according to the knowledge of the forces acting on the launcher. Finally, the use of several low-end inertial measurement units is compared to the use of a single higher quality inertial measurement unit. The use of several inertial measurement units makes it possible to distribute these on the structure of the launcher. The effect of the positioning of these measurement units is evaluated by comparing navigation solutions of which all the sensors are placed in the head of the launcher to a solution where the sensors are distributed in the head of each stage of the launcher. Some approaches for merging the inertial measurement units data are tested; fusion of all the inertial measurement units in a single inertial navigation system, fusion of several inertial navigation systems and fusion of several inertial navigation systems with geometric constraints. For many of the proposed solutions as well as for several others already in existence, an exhaustive study of the observability is carried out. The observability is verified using the observability matrix, the study of the propagation of the covariance matrix and the analysis of the sensitivity to measurement outliers. Finally, a simplified method for evaluating the performance of inertial navigation solutions on a fixed trajectory is presented. The results show that the use of the reference trajectory data makes it possible to improve the roll estimate. However, adding these data does not improve the observability of the base model on which they are added. The use of a stochastic model of the launcher dynamics provides marginal gains on attitude, velocity and position estimates. On the other hand, this approach substantially improves the acceleration and angular velocity estimates. The distribution of inertial sensors on the launcher structure does not improve navigation accuracy. The latter is even degraded when sensors are lost due to stage jettisoning. Merging several inertial measurement units with a single inertial navigation system provides performance equivalent to that achieved with a single inertial measurement unit. On the other hand, the fusion of several inertial navigation systems makes it possible to reduce the estimation error. Moreover, additional gains can be obtained when geometric constraints on the relative attitude, velocity and position between inertial navigation systems are added to this latter approach. Observability tests have shown that modelling sensor biases by a Markov process rather than a random walk has little impact on the observability of the model. In addition, on a short-term mission, the choice of the model to represent the sensor bias has only a negligible effect on the precision of the estimates. Observability analyzes have also shown that the use of a single GPS receiver is not sufficient to ensure the observability of the roll and that the position measurement bias of the GPS receiver is not observable.
Document Type: Thèse de doctorat
Issue Date: 2019
Open Access Date: 24 April 2019
Permalink: http://hdl.handle.net/20.500.11794/34536
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
34982.pdf6.09 MBAdobe PDFThumbnail
View/Open
34982_Annexe.zipAnnexe3.49 MBArchive ZIPView/Open
34982_Annexe_Contenu.txt934 BTextView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.