State estimation in unmanned aerial vehicles: theoretical approaches for increasing accuracy.

Minnaar, Nicholas Johan (2017-12)

Thesis (MEng)--Stellenbosch University, 2017.

Thesis

ENGLISH ABSTRACT: The utility of unmanned aerial vehicles (UAVs) are enhanced by their ability to navigate accurately. Unfortunately, it is not possible to capture the states which describe a UAV's position and orientation in a measurement system without some degree of uncertainty. The work presented here resolves this issue by developing strategies for generating accurate navigational outputs from the measurements taken on board an UAV. The three strategies that are ultimately developed address fundamental misnomers that are related to the processing of navigational measurement data. The first strategy relates to the choice of state variable when exogenous signals are available in the UAV's measurement matrix. It is suggested that these signals be used to update error estimates related to other sensors instead of acting as observations of internal system states. The second strategy addresses a common theoretical conflict which arises in UAV applications where accelerometers are used for orientation observations. It is suggested that redundant accelerometers be added to the measurement system and a strategy for distilling orientation rate information from this redundant information is also provided. The final strategy which is provided addresses the numerical errors which slowly disintegrate the orientation estimates of a UAV. It is suggested that a singular value decomposition be leveraged to restore the welfare of the orientation estimates. All the proposals which are made improve the navigational performance of the UAV and are validated through simulation.

AFRIKAANSE OPSOMMING: Die nut, aanwending en gebruik van Hommeltuie word verbeter met hul direkte vermoë om meer akkuraat te kan navigeer. Ongelukkig is dit nie moontlik om die staat van die veranderlikes in terme van posisie en orientasie te bepaal sonder 'n mate van metingsonsekerheid nie. Hierdie voorgestelde werk bied 'n oplossing deur drie verskillende strategie te ontwikkel om meer akkurate navigasie data te genereer uit die Hommeltuig se bestaande, aanboord metingsstelsel. Die drie strategieë behandel die verwerkings metodes van die navigasie metingsdata wat die foute veroorsaak gedurende die prosesering, vooruitskatting en bepaling van staatskattings. Die eerste strategie hou verband met die voorkeur van veranderlike state sodra eksogene seine ook teenwoordig is in die meetingsmatriks van die Hommeltuig. Daar word voorgestel dat hierdie seine aangewend word om foutberamings (wat verband hou met die insete van ander sensors) eerder voordurend op te dateer in plaas van om die interne staat veranderlikes op te dateer. Die tweede strategie spreek 'n algemene teoretiese teenstrydigheid aan by Hommeltuig aanwendings wat onstaan sodra versnelingsmeters gebruik word vir orienteringswaarnemings. Dit word voorgestel dat addisionele versnellingsmeters by die meetstelsel gevoeg word en 'n strategie om die oortollige inligting te gebruik om orienteringsberamings te bereken. Die finale gegewe strategie spreek die numeriese foute aan wat die orientasie beramings van die Hommeltuig stelselmatig laat disintegreer. Daar word voorgestel dat 'n enkelvoudige afgeleide waarde ("Singular Value Decomposition") aangewend word om die welstand van die orientasieberamings te herstel. Elk van hierdie drie unieke voorstelle verbeter die akkuraatheid van navigasie vermoe van 'n Hommeltuig en is deur middel van simmulasie bewys.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/102867
This item appears in the following collections: