Les drones ou systèmes de drones aériens jouent un rôle de plus en plus important danstous les domaines, spécialement les drones à décollage et atterrissage verticaux. L’un desplus connus est le Quadrotor et, sans doute, il est la plateforme de recherche la plus utilisée.Cette thèse traite le problème de l’estimation et de la commande d’attitude appliqué àun corps rigide se déplaçant dans l’espace 3D tel que le Quadrotor. La première contributionde cette thèse est la conception et l’implémentation d’une solution d’estimation d’attitude.Celle-ci est basée sur un ensemble de filtres complémentaires combinés avec un algorithmealgébrique tel que TRIAD, QUEST, etc. avec la possibilité de choisir deux formes différentesdes filtres: la première dénommée forme Directe, et la seconde dénommée forme Passive.Les filtres proposés ont une flexibilité dans le choix de l’ordre qui peut être pris grand afinde bien réduire l’effet du bruit de mesure et permettent d’aboutir à un estimateur qui peutprendre en compte le biais éventuel des gyromètres. L’analyse par la théorie de Lyapunovprouve que les erreurs d’estimation tendent globalement et asymptotiquement vers zéro. Unesuite logique de cette première contribution est la proposition d’une solution pour la commanded’attitude qui constitue la deuxième contribution de cette thèse. Elle se traduit par ledéveloppement d’une nouvelle loi de commande d’attitude d’un corps rigide dans l’espace3D, dans laquelle seulement les vecteurs de mesures inertiels avec les mesures des gyromètressont utilisés. Elle utilise le principe de fusion des données à travers un filtre complémentairepermettant l’élimination des bruits des mesures tout en assurant une stabilité presque globalede l’équilibre désiré. La troisième contribution est une loi de commande pour la stabilisationd’attitude sans mesure de vitesse angulaire, ni mesure d’attitude. Pour cela, un systèmelinéaire auxiliaire basé sur les mesures des vecteurs inertiels a été introduit. Ce dernier sesubstitue au manque de l’information de la vitesse angulaire. L’analyse de stabilité du contrôleurproposé est basée sur la théorie de Lyapunov couplée avec le théorème de LaSalle. Ellepermet de conclure sur la stabilité presque globale de l’équilibre désiré. Les performances dessolutions proposées ont été validées par un ensemble de tests expérimentaux / Nowadays, we see a growing popularity of the use of Unmanned Aerial Vehicles (UAV) ofespecially Vertical Take-Off and Landing (VTOL) type. One of the most known VTOL is thequadrotor or Quadcopter which is probably the most used one as a research platform. Thisthesis deal with attitude control and estimation techniques applied to a rigid body movingin 3D space such as Quadcopter VTOL. The first contribution of this thesis is the design ofa new class of complementary linear-like filters allowing the fusion of inertial vector measurementswith angular velocity measurements and combined with algebraic algorithms asTRIAD, QUEST etc. to give an efficient attitude estimation solution. This class of filtersallows several possibilities of implementation such as the order of the filters which can bechosen high in order to reduce more the measurement noise and the form of the filters thatcan be direct or passive and the ability to take into account the possible gyro bias. Lyapunovanalysis shows the global asymptotic convergence of the estimation errors to zero. The sameprinciple of data fusion is used for the proposed new attitude control law in which the complementaryfilters were included to reduce the effect of measurement noise. The obtainedcontroller ensures almost global stability of the desired equilibrium point; it represents thesecond contribution of this thesis. The third contribution takes into consideration an interestingspecial case, where instantaneous measurements of attitude and angular velocity areunavailable. A first order linear auxiliary system based directly on vector measurements isused in an observer-like system to handle the luck of angular velocity. The proposed controllerensures almost global asymptotic stability of the trajectories to the desired equilibriumpoint. Detailed sets of experiments were done to validate the obtained results

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23178号 / 工博第4822号 / 新制||工||1753(附属図書館) / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 山田 忠史, 教授 藤井 聡, 准教授 SCHMOECKER Jan-Dirk / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Ziel dieser Arbeit ist es, robuste und performante Algorithmen für die Fusion von polizeilichen Unfalldaten zur Testszenariengenerierung im Rahmen der Absicherung automatisierter Fahrfunktionen zu generieren. In dieser Arbeit werden dabei Methoden zur Datenfusion in Kombination mit generativen und Klassifikationsmodellen untersucht. Eine spezifische Variable vom Empfänger wird während des Datenfusionsverfahrens im Voraus entfernt. Ein Spender mit den gemeinsamen Variablen wird verwendet, um die Vorhersage für die fehlende spezifische Variable im Empfänger zu erhalten. Als Methode werden Ensembles aus Distance-Hot-Deck und Machine-Learning Verfahren für die Vorhersage verwendet. Nach der Vorhersage werden die Ergebnisse anhand ausgewählter Bewertungsmetriken bewertet. Darüber hinaus werden zwei generative Modelle eingeführt, um Datensätze unterschiedlicher Qualität zu synthetisieren. Ziel ist es, die Robustheit der Ensembles mit den synthetisierten „Rauschdaten“ zu testen und die Performance von Ensembles mit den synthetisierten Daten hoher Qualität zu verbessern. Schließlich können Erkenntnisse darüber gewonnen werden, welche Ensembles die besten Ergebnisse für die Datenfusion liefern.:1. Einleitung 2. Grundlagen 3. Randbedingungen 4. Vorgehensweise 5. Ergebnisse 6. Diskussion & Ausblick

This thesis presents possibilities of 3D point cloud and true colored digital video fusion that can be used in the process of robot teleoperation. Advantages of a 3D environment visualization combining more than one sensor data, tools to facilitate such data fusion, as well as two alternative practical implementations of combined data visualization are discussed. First proposed alternative estimates view frustum of the robot's camera and maps real colored video to a semi-transparent polygon placed in the view frustum. The second option is a direct coloring of the point cloud data creating a colored point cloud representing color as well as depth information about an environment.

Tato diplomová práce se zabývá tvorbou fúze pozičních dat z existující realtimové im- plementace vizuálního SLAMu a kolové odometrie. Výsledkem spojení dat je potlačení nežádoucích chyb u každé ze zmíněných metod měření, díky čemuž je možné vytvořit přesnější 3D model zkoumaného prostředí. Práce nejprve uvádí teorií potřebnou pro zvládnutí problematiky 3D SLAMu. Dále popisuje vlastnosti použitého open source SLAM projektu a jeho jednotlivé softwarové úpravy. Následně popisuje principy spo- jení pozičních informací získaných vizuálními a odometrickými snímači, dále uvádí popis diferenciálního podvozku, který byl použit pro tvorbu kolové odometrie. Na závěr práce shrnuje výsledky dosažené datovou fúzí a srovnává je s původní přesností vizuálního SLAMu.

Linked Data have emerged as a successful publication format which could mean to structured data what Web meant to documents. The strength of Linked Data is in its fitness for integration of data from multiple sources. Linked Data integration opens door to new opportunities but also poses new challenges. New algorithms and tools need to be developed to cover all steps of data integration. This thesis examines the established data integration proceses and how they can be applied to Linked Data, with focus on data fusion and conflict resolution. Novel algorithms for Linked Data fusion are proposed and the task of supporting trust with provenance information and quality assessment of fused data is addressed. The proposed algorithms are implemented as part of a Linked Data integration framework ODCleanStore.

By purposefully utilising sensors, for instance by a datafusion system, the state of some system-relevant environmentmight be adequately assessed to support decision-making. Theever increasing access to sensors o.ers great opportunities,but alsoincurs grave challenges. As a result of managingmultiple sensors one can, e.g., expect to achieve a morecomprehensive, resolved, certain and more frequently updatedassessment of the environment than would be possible otherwise.Challenges include data association, treatment of con.ictinginformation and strategies for sensor coordination. We use the term information acquisition to denote the skillof a data fusion system to actively acquire information. Theaim of this thesis is to instructively situate that skill in ageneral context, explore and classify related research, andhighlight key issues and possible future work. It is our hopethat this thesis will facilitate communication, understandingand future e.orts for information acquisition. The previously mentioned trend towards utilisation of largesets of sensors makes us especially interested in large-scaleinformation acquisition, i.e., acquisition using many andpossibly spatially distributed and heterogeneous sensors. Information acquisition is a general concept that emerges inmany di.erent .elds of research. In this thesis, we surveyliterature from, e.g., agent theory, robotics and sensormanagement. We, furthermore, suggest a taxonomy of theliterature that highlights relevant aspects of informationacquisition. We describe a function, perception management (akin tosensor management), which realizes information acquisition inthe data fusion process and pertinent properties of itsexternal stimuli, sensing resources, and systemenvironment. An example of perception management is also presented. Thetask is that of managing a set of mobile sensors that jointlytrack some mobile targets. The game theoretic algorithmsuggested for distributing the targets among the sensors proveto be more robust to sensor failure than a measurement accuracyoptimal reference algorithm. <b>Keywords:</b>information acquisition, sensor management,resource management, information fusion, data fusion,perception management, game theory, target tracking / NR 20140805

Cognitive radio networks (CRNs) have emerged as a solution for the looming spectrum crunch caused by the rapid adoption of wireless devices over the previous decade. This technology enables efficient spectrum utility by dynamically reusing existing spectral bands. A CRN achieves this by requiring its users – called secondary users (SUs) – to measure and opportunistically utilise the band of a legacy broadcaster – called a primary user (PU) – in a process called spectrum sensing. Sensing requires the distribution and fusion of measurements from all SUs, which is facilitated by a variety of architectures and topologies. CRNs possessing a central computation node are called centralised networks, while CRNs composed of multiple computation nodes are called decentralised networks. While simpler to implement, centralised networks are reliant on the central node – the entire network fails if this node is compromised. In contrast, decentralised networks require more sophisticated protocols to implement, while offering greater robustness to node failure. Relay-based networks, a subset of decentralised networks, distribute the computation over a number of specialised relay nodes – little research exists on spectrum sensing using these networks. CRNs are vulnerable to unique physical layer attacks targeted at their spectrum sensing functionality. One such attack is the Byzantine attack; these attacks occur when malicious SUs (MUs) alter their sensing reports to achieve some goal (e.g. exploitation of the CRN’s resources, reduction of the CRN’s sensing performance, etc.). Mitigation strategies for Byzantine attacks vary based on the CRN’s network architecture, requiring defence algorithms to be explored for all architectures. Because of the sparse literature regarding relay-based networks, a novel algorithm – suitable for relay-based networks – is proposed in this work. The proposed algorithm performs joint MU detection and secure sensing by large-scale probabilistic inference of a statistical model. The proposed algorithm’s development is separated into the following two parts. • The first part involves the construction of a probabilistic graphical model representing the likelihood of all possible outcomes in the sensing process of a relay-based network. This is done by discovering the conditional dependencies present between the variables of the model. Various candidate graphical models are explored, and the mathematical description of the chosen graphical model is determined. • The second part involves the extraction of information from the graphical model to provide utility for sensing. Marginal inference is used to enable this information extraction. Belief propagation is used to infer the developed graphical model efficiently. Sensing is performed by exchanging the intermediate belief propagation computations between the relays of the CRN. Through a performance evaluation, the proposed algorithm was found to be resistant to probabilistic MU attacks of all frequencies and proportions. The sensing performance was highly sensitive to the placement of the relays and honest SUs, with the performance improving when the number of relays was increased. The transient behaviour of the proposed algorithm was evaluated in terms of its dynamics and computational complexity, with the algorithm’s results deemed satisfactory in this regard. Finally, an analysis of the effectiveness of the graphical model’s components was conducted, with a few model components accounting for most of the performance, implying that further simplifications to the proposed algorithm are possible. / Dissertation (MEng)--University of Pretoria, 2020. / Electrical, Electronic and Computer Engineering / MEng / Unrestricted

Researchers now have unprecedented access to free Landsat data, enabling detailed monitoring of the Earth's land surface and vegetation.  There are gaps in the data, due in part to cloud cover. The gaps are aperiodic and localized, forcing any detailed multitemporal analysis based on Landsat data to compensate.   Harmonic regression approximates Landsat data for any point in time with minimal training images and reduced storage requirements.  In two study areas in North Carolina, USA, harmonic regression approaches were least as good at simulating missing data as STAR-FM for images from 2001.  Harmonic regression had an R^2"0.9 over three quarters of all pixels. It gave the highest R_Predicted^2 values on two thirds of the pixels.  Applying harmonic regression with the same number of harmonics to consecutive years yielded an improved fit, R^2"0.99 for most pixels.   We next demonstrate a change detection method based on exponentially weighted moving average (EWMA) charts of harmonic residuals. In the process, a data-driven cloud filter is created, enabling use of partially clouded data.  The approach is shown capable of detecting thins and subtle forest degradations in Alabama, USA, considerably finer than the Landsat spatial resolution in an on-the-fly fashion, with new images easily incorporated into the algorithm.  EWMA detection accurately showed the location, timing, and magnitude of 85% of known harvests in the study area, verified by aerial imagery.   We use harmonic regression to improve the precision of dynamic forest parameter estimates, generating a robust time series of vegetation index values.  These values are classified into strata maps in Alabama, USA, depicting regions of similar growth potential.  These maps are applied to Forest Service Forest Inventory and Analysis (FIA) plots, generating post-stratified estimates of static and dynamic forest parameters.  Improvements to efficiency for all parameters were such that a comparable random sample would require at least 20% more sampling units, with the improvement for the growth parameter requiring a 50% increase. These applications demonstrate the utility of harmonic regression for Landsat data.  They suggest further applications in environmental monitoring and improved estimation of landscape parameters, critical to improving large-scale models of ecosystems and climate effects. / Ph. D.

Cette thèse a pour but d'étudier et de quantifier la contribution de la perception multimodale assistée par le contexte pour détecter et suivre des objets en mouvement. Cette étude sera appliquée à la détection et la reconnaissance des objets pertinents dans les environnements de la circulation pour les véhicules intelligents (VI). Les résultats à obtenir devront permettre de transposer le concept proposé à un ensemble plus large de capteurs et de classes d'objets en utilisant une approche système intégrative qui implique des méthodes d'apprentissage. En particulier, ces méthodes d'apprentissage vont examiner comment l'implantation dans un système intégré, qui prévoie une multitude des sources de données différentes, peut conduire à apprendre 1) sans ou avec une supervision limitée, réduite en exploitant des corrélations 2) de façon incrémentale à la connaissance stockée au lieu de faire un entraînement complet à chaque fois qu’une nouvelle donnée arrive 3) collectivement à chaque instant d'apprentissage dans le système entraîné d'une manière qui assure approximativement une fusion optimale. Concrètement, le couplage fort entre les classifier des objets en modalités multiples aussi bien que l'extraction du contexte de la géométrie de la scène sont à étudier: d'abord en théorie, après en application du trafic routier. La nouveauté de l'approche d'intégration envisagée se pose dans le couplage fort entre les composants du système, tels que la segmentation, le suivi des objets, l'estimation de la géométrie de la scène et la catégorisation des objets basée sur la stratégie de l'inférence probabiliste. Une telle stratégie caractérise des systèmes où toutes les composants de perception émettent et reçoivent les distributions des résultats possibles avec leur score de croyance probabiliste attribué. De cette façon, chaque composant de traitement peut prendre en compte les résultats des autres composants au niveau plus bas par rapport aux combinaisons des résultats finaux. Cela diminue beaucoup le temps et les ressources pour le calcul, quand les techniques de l'application de l'inférence Bayésienne garantissent que les données d'entrée peu plausible n'apportent pas des impacts négatifs. / This thesis project will investigate and quantify the contribution of context-aided multimodal perception for detecting and tracking moving objects. This research study will be applied to the detection and recognition ofrelevant objects in road traffic environments for Intelligent Vehicles (IV). The results to be obtained will allow us to transpose the proposed concept to a wide range of state-of-the-art sensors and object classes by means of an integrative system approach involving learning methods. In particular, such learning methods will investigate how the embedding into an embodied system providing a multitude of different data sources, can be harnessed to learn 1) without, or with reduced, explicit supervision by exploiting correlations 2) incrementally, by adding to existing knowledge instead of complete retraining every time new data arrive 3) collectively, each learning instance in the system being trained in a way that ensures approximately optimal fusion. Concretely, a tight coupling between object classifiers in multiple modalities as well as geometric scene context extraction will be studied, first in theory, then in the context of road traffic. The novelty of the envisioned integration approach lies in the tight coupling between system components such as object segmentation, object tracking, scene geometry estimation and object categorization based on a probabilistic inference strategy. Such a strategy characterizes systems where all perception components broadcast and receive distributions of multiple possible results together with a probabilistic belief score. In this way, each processing component can take into account the results of other components at a much earlier stage (as compared to just combining final results), thus hugely increasing its computation power, while the application of Bayesian inference techniques will ensure that implausible inputs do not cause negative effects.