The use of drones is increasing not just for private use, but also for companieswithin different sectors including critical infrastructures, such as the energysector. This due to the many benefits that drones can bring in terms of loweredcosts, increased safety for people, better accessibility, and more, whenconducting varies activities such as inspections and looking for errors.Although the use of drones may be comprised of benefits, drones may also beprone to security risks related to privacy, safety, as well as informationsecurity and the confidentiality, integrity, and availability of the informationthat is collected. This Master´s thesis focused on the critical infrastructuresector following an energy company in Sweden and their use of drones forconducting various operations. The thesis also investigated how security risksfocusing on information security are viewed by the company in terms oflikelihood and consequence and how the company handles these risks. Risksrelated to information security may be caused by threats such as adversaries,technical malfunctions, weather, and climate. The seriousness of the risksmuch depend on the sensitivity of the information and, how much theinformation covers, for example of the power grid. With regards to likelihood,a factor in play is whether the information would be possible for another partyto collect as well as if information would be possible to collect at anothertime. This would impact whether an adversary would gain anything fromcompromising the information collected using a drone or, how technicalmalfunctions and natural occurrences affecting the drone, and the informationcollected would cause problems if the flight is possible to conduct at anothertime. The thesis was conducted using a qualitative approach in the form of asingle-case study. / <p>The presentation was conducted digitally through zoom. </p>

<p>This dissertation offers an in-depth exploration of how temporality—and the imperative to control the unfolding of time in particular—is embedded in the practices, processes, and dynamics of contemporary world politics. While most International Relations scholarship remains conspicuously uninterested in questions relating to time, this study sees such temporal blindness as inhibiting the development of adequately nuanced and critically oriented understandings of key theoretical and practical issues in the global political realm. It thus attempts to demonstrate how time can be “brought in” to the study of world politics, and to highlight the analytical utility and critical potential of doing so. In this respect, Part I considers the importance of temporality to perhaps the most fundamental global political concept—state sovereignty—and then moves on to discuss how shifts in the contemporary political imagination have (re-)inscribed temporal contingency as a pressing problem that requires a political response. Part II then attempts to critically think through what is at stake in the resulting proliferation of anticipatory governance strategies premised upon controlling the unfolding of the future through pre-emptive intervention in the present. It is argued that by prioritizing imagination and conjecture in the context of political decision-making, such temporally-inflected strategies serve to radically reconfigure the way political power is organized and exercised, such that a paradigm of political authority best described as "exceptionalism” is enacted. This line of argument is developed through a comprehensive conceptual engagement with one particularly prominent manifestation of this ongoing “temporalization” of the political—namely, the “pre-emptive security” strategies that have emerged as central to the conduct of the global War on Terror. It is concluded that the adoption of anticipatory political rationalities is particularly problematic for the liberal democratic states that have most enthusiastically done so—both in the security realm and beyond.</p> / Doctor of Philosophy (PhD)

The drone industry is growing and the need for increased autonomy will be required if large fleetof drones will be able to fly without a single pilot per drone. A useful part of automating the flighten-route can be achieved with the upcoming standard of Direct Remote Id (DRI), which signalspositional data for drones and can be used as the perceptive part in a collision avoidance systembetween drones with the advantage of limited weight penalties and minimal financial cost.Simulations were carried out to understand different kinds of evasive maneuvers and develop asimple yet effective algorithm for avoiding obstacles and continue towards the next waypoint ona mission. Positional data can be retrieved with an ESP-32 board from a flight computer withMavlink protocol, which can then be broadcasted and received to an ESP-32 board using DirectRemote Id. The distances between the nearest drones can be computed, along with the shortest al-lowable distance and closest positions of the drones, if they were to continue on a straight course. Ifthe closest passing distance turned out closer than a set safety distance, an evasive maneuver is cal-culated and executed, with preliminary work focusing on evasion maneuvers on an horizontal plane.Flight tests showed that an evasive position could be calculated, and the drone successfully di-verted to it, while continuing with the mission after the evasion was completed. These resultsshowed the potential of using Direct Remote Id as a simple close proximity detection for use withcollision avoidance / Drönarindustrin växer allt snabbare och det kommer att krävas en större grad av autonomitet för att kunna få drönare att flyga av sig själva utan att ha en pilot per drönare. En användbar del av att kunna autonomisera flygrutten vid flykt kan vara den nya standarden Direct Remote Id (DRI), som sänder ut positionsdata för den individuella drönaren och kan användas för att kunna upptäcka och bli upptäckt av andra drönare med minimal vikt- och priskostnad.Simuleringar gjordes för att undersöka samt förstå olika typer av undanmanövrar och för att utveckla en simpel och effektiv algoritm för att kunna undvika objekt och fortsätta med en planerad rutt. Positionsdata kan skickas till ett ESP-32 kretskort fån en flygdator med hjälp av Mavlink protokoll, denna data kan sedan sändas och bli mottaget av ett annat ESP-32 kretskort med Direct Remote Id standarden. Avståndet till den närmsta drönaren kan beräknas samt den minsta passagedistansen mellan drönarna om de skulle fortsätta i rak riktning. Om den minsta passagedistansen mellan drönarna är mindre än ett satt säkerhetsavstånd, beräknas en undanmanöver samt utförs. Flygtester visade att en undanmanöver kunde beräknas samt att drönaren omdirigerades till sidan om objektet och därefter fortsatte på sin planerade flygrutt. Dessa resultat visade potentialen i att använda Direct Remote Id som ett enkelt sätt att upptäcka andra drönare för att användas i ett kollisionsundvikande system.

Thèse ou mémoire avec insertion d'articles / Les systèmes aériens autonomes (UAV/UAS), communément appelés drones, sont un sujet de plus en plus important dans les inspections par essais non-destructifs (END). Avec les avancées technologiques significatives des caméras thermiques, les méthodes d'inspection visuelle acquièrent continuellement de l'attention dans les inspections END. Les inspections dans les zones difficiles d'accès sont coûteuses, parfois impossibles en raison de la nature de la zone ou des dangers possibles pour les ressources humaines. L'inspection de spécimens complexes et de grande taille, notamment les des structures courbes, nécessite des relevés approfondis sous différents aspects, ce qui est presque impossible ou très coûteux avec des véhicules terrestres ou des ressources humaines. Ainsi, en raison de leur grande manœuvrabilité, les industries investissent davantage dans les drones pour surmonter les problèmes mentionnés et aider les inspecteurs à examiner les composants de manière approfondie. De plus, grâce à des développements récents, les UAVs peuvent également accéder à des zones éloignées ou difficiles d'accès et transporter de nombreuses charges utiles. Malgré les énormes avantages de l'utilisation des drones pour l'inspection, certains défis doivent être relevés. Ces dernières années, de nombreuses études se sont concentrées sur l'utilisation d'images thermiques/visibles pour inspecter différentes structures. Cependant, l'utilisation de données d'inspection multimodales par drone, y compris les données d'imagerie visible, thermique et de profondeur, pour fournir une compréhension approfondie de l'échantillon et de son environnement afin de produire une analyse plus précise, doit être étudiée en détail. Tout d'abord, cette étude aborde les défis communs des inspections par drone. La détection de l'effet de la réflexion thermique dans une inspection thermographique est le premier défi abordé dans cette étude. Ensuite, l'effet des mouvements constants et soudains d'un drone sur l'analyse des séquences d'images thermiques est étudié de manière approfondie. En outre, les résultats sont évalués à l'aide d'un scénario d'utilisation où le drone surveille un endroit fixe tout en restant en vol stationnaire. Par la suite, cette étude vise à développer une plateforme multi-sensorielle comprenant une structure de montage, des capteurs d'imagerie et un ordinateur embarqué. La solution logicielle intégrée à cette plate-forme fournit les fonctions requises d'acquisition, de transmission, de stockage et de traitement des données. De plus, cette étude se concentre sur le traitement de modalités multiples ou individuelles. Notamment, une méthode de segmentation par auto-apprentissage est proposée dans le contexte de la détection de défauts dans les images thermiques. Aussi, un algorithme de détection de fissures par drone est présenté pour analyser l'inspection visuelle des chaussées et des structures en béton. Ensuite, cette étude s'est concentrée sur le traitement des données multi-modales acquises par la plateforme multi-sensorielle présentée. En effet, l'utilisation d'images thermiques et visibles couplées pour améliorer la détection des anomalies est étudiée de manière approfondie. Plusieurs scénarios d'utilisation sont introduits présentant différentes approches pour améliorer l'efficacité de la détection. Ces derniers fournissent un aperçu de l'applicabilité des sous-études introduites. Pour chacun d'entre eux, de multiples expériences sont menées pour démontrer les applications des méthodes proposées dans des scénarios de cas réels. / Unmanned Aerial Vehicles/Systems (UAVs/UAS), commonly known as drones, is a rising topic in Non-Destructive Testing (NDT) inspections. With significant technological advancements in thermal cameras, visual inspection methods continuously gain much attention in non-destructive inspections. Inspections in remote or hard-to-access areas are costly and sometimes impossible due to the area's nature or the possible dangers facing human resources. Inspection of complex and large specimens, especially with curvaceous structures, requires extensive surveys from different aspects, which is nearly impossible or very costly using ground vehicles or human resources. Thus, industries are investing more in drones to overcome mentioned problems as they have high flexibility of maneuver, which can assist inspectors in examining the components thoroughly. They can also access remote or hard-to-access areas and carry many payloads thanks to recent developments. Despite the enormous benefits of using drones for inspection, some challenges need to be addressed. In recent years, many studies focused on using thermal/visible images to inspect different structures. However, using multi-modal data, including visible, thermal, and depth imagery data, provides an extensive understanding of the specimen and surrounding environment in case of drone-enabled inspections and produces a more accurate analysis that needs to be thoroughly studied. Firstly, this study addresses the common challenges in drone-based inspections in the scope of this research. Detecting the effect of thermal reflection in a thermographic inspection is the first challenge addressed in this study. Later, the effect of a drone's constant and sudden motions on analyzing thermal image sequences is investigated comprehensively. Also, the results are evaluated using a use-case scenario where the drone monitors a fixed location while hovering. Also, the next part of this study aims to develop a multi-sensory platform, including a mounting structure, imagery sensors, and an onboard computer. The software solution embedded in this platform provides the required data acquisition, transmission, storage, and processing features. Later, this study focuses on the processing of multiple or individual modalities. Firstly, a self-training segmentation method is proposed in the context of defect detection in thermal images. Also, a drone-enabled crack detection algorithm is presented for analyzing the visual inspection of pavement and concrete structures. Next, this study focused on processing multi-modal data acquired by the presented multi-sensory platform. Firstly, using coupled thermal and visible images to enhance abnormality detection is investigated thoroughly. Several use-case scenarios are introduced, presenting different approaches to enhance the detection's efficiency. In order to provide insight into the applicability of the introduced sub-studies. For each of them, multiple experiments are conducted demonstrating the applications of the proposed methods in real-case scenarios.

The NASA Performance Adaptive Aeroelastic Wing (PAAW) project goals include significant reductions in fuel burn, emissions, and noise via efficient aeroelastic design and improvements in propulsion systems. As modern transport airplane designs become increasingly lightweight and incorporate high aspect-ratio wings, aeroservoelastic effects gain prominence in modeling and design considerations. As a result, the influence of the flight dynamics and controls on the optimal structural and aerodynamic design needs to be captured in the design process. There is an increasing interest in more integrated aircraft multidisciplinary design optimization (MDAO) processes that can bring flight control design into the early stage of an aircraft design cycle. So, in this thesis different flight dynamics modeling methodologies are presented that can be integrated within the MDAO framework. MDAO studies are conducted to maximize the controllability and observability of a UAV type aircraft using curvilinear SpaRibs and straight spars and ribs as the internal structural layout. The impulse residues and controllability Gramians are used as surrogates for the control objectives in the MDAO to maximize the controllability and observability of the aircraft. The optimal control designs are compared with those obtained using weight minimization as the design objective. It is found that using the aforementioned control objectives, the resulting aircraft design is more controllable and can be used to expand the flight envelope by up to 50% as compared to the weight minimized design. / Doctor of Philosophy / Over the last two decades, several attempts have been made towards multidisciplinary design analysis and optimization (MDAO) of flexible wings by integrating flight control laws in the wing design so that the aircraft will have sufficient control authority across different flying conditions. However, most of the studies have been restricted to the wing design only using a predefined control architecture approach, which would be very difficult to implement at the conceptual design stage. There is a need for an approach that would be faster and more practical. Including control surface and control law designs at the conceptual design stage is becoming increasingly important, due to the complexity of both the aircraft control laws and that of the actuation and sensing, and the enhanced wing flexibility of future transport aircraft. A key question that arises is, can one design an aircraft that is more controllable and observable? So, in this thesis, a more fundamental approach, in which the internal structural layout of the aircraft is optimized to design an aircraft that is more controllable, is presented and implemented. The approach uses the fundamentals of linear systems theory for maximizing the controllability and observability of the aircraft using an MDAO framework.

The main purpose of this thesis is to develop a beam element in order to model the wing of a drone, made of composite materials. The proposed model consists of the framework for the structural design and analysis of long slender beam like structures, e.g., wings, wind turbine blades, and helicopter rotor blades, etc. The main feature consists of the addition of the coupling between axial and bending with torsional effects that may arise when using composite materials and the coupling stemming from the inhomogeneity in cross-sections of any arbitrary geometry. This type of modeling approach allows for an accurate yet computationally inexpensive representation of a general class of beam-like structures. The framework for beam analysis consists of main two parts, cross-sectional analysis of the beam sections and then using this section analysis to build up the finite element model. The cross-sectional analysis is performed in order to predict the structural properties for composite sections, which are used for the beam model. The thesis consists of the model to validate the convergence of the element size required for the cross-sectional analysis. This follows by the validation of the shell models of constant cross-section to assess the performance of the beam elements, including coupling terms. This framework also has the capability of calculating the strains and displacements at various points of the cross-section. Natural frequencies and mode shapes are compared for different cases of increasing complexity with those available in the papers. Then, the framework is used to analyze the wing of a drone and compare the results to a model developed in NASTRAN. / Master of Science

Generalized collaborative drones are a technology that has many potential benefits. General purpose drones that can handle exploration, navigation, manipulation, and more without having to be reprogrammed would be an immense breakthrough for usability and adoption of the technology. The ability to develop these multi-task, multi-agent drone systems is limited by the lack of available training environments, as well as deficiencies of multi-task learning due to a phenomenon known as catastrophic forgetting. In this thesis, we present a set of simulation environments for exploring the abilities of multi-task drone systems and provide a platform for testing agents in incremental single-agent and multi-agent learning scenarios. The multi-task platform is an extension of an existing drone simulation environment written in Python using the PyBullet Physics Simulation Engine, with these environments incorporated. Using this platform, we present an analysis of Incremental Learning and detail the beneficial impacts of using the technique for multi-task learning, with respect to multi-task learning speed and catastrophic forgetting. Finally, we introduce a novel algorithm, Incremental Learning with Second-Order Approximation Regularization (IL-SOAR), to mitigate some of the effects of catastrophic forgetting in multi-task learning. We show the impact of this method and contrast the performance relative to a multi-agent multi-task approach using a centralized policy sharing algorithm. / Master of Science / Machine Learning techniques allow drones to be trained to achieve tasks which are otherwise time-consuming or difficult. The goal of this thesis is to facilitate the work of creating these complex drone machine learning systems by exploring Reinforcement Learning (RL), a field of machine learning which involves learning the correct actions to take through experience. Currently, RL methods are effective in the design of drones which are able to solve one particular task. The next step in this technology is to develop RL systems which are able to handle generalization and perform well across multiple tasks. In this thesis, simulation environments for drones to learn complex tasks are created, and algorithms which are able to train drones in multiple hard tasks are developed and tested. We explore the benefits of using a specific multi-task training technique known as Incremental Learning. Additionally, we consider one of the prohibitive factors of multi-task machine learning-based solutions, the degradation problem of agent performance on previously learned tasks, known as catastrophic forgetting. We create an algorithm that aims to prevent the impact of forgetting when training drones sequentially on new tasks. We contrast this approach with a multi-agent solution, where multiple drones learn simultaneously across the tasks.

Cette thèse propose une stratégie de contrôle avancée pour guider une flotte d'aéronefs sans pilote (UAV) dans un environnement à la fois stochastique et dynamique. Pour ce faire, un simulateur de vol 3D a été développé avec MATLAB® pour tester les algorithmes de la stratégie de guidage en fonctions de différents scénarios. L'objectif des missions simulées est de s'assurer que chaque UAV intercepte une cible ellipsoïdale mobile tout en évitant une panoplie d'obstacles ellipsoïdaux mobiles détectés en route. Les UAVs situés à l'intérieur des limites de communication peuvent coopérer afin d'améliorer leurs performances au cours de la mission. Le simulateur a été conçu de façon à ce que les UAV soient dotés de capteurs et d'appareils de communication de portée limitée. De plus, chaque UAV possède un pilote automatique qui stabilise l'aéronef en vol et un planificateur de trajectoires qui génère les commandes à envoyer au pilote automatique. Au coeur du planificateur de trajectoires se trouve un contrôleur prédictif à horizon fuyant qui détermine les commandes à envoyer à l'UAV. Ces commandes optimisent un critère de performance assujetti à des contraintes. Le critère de performance est conçu de sorte que les UAV atteignent les objectifs de la mission, alors que les contraintes assurent que les commandes générées adhèrent aux limites de manoeuvrabilité de l'aéronef. La planification de trajectoires pour UAV opérant dans un environnement dynamique et stochastique dépend fortement des déplacements anticipés des objets (obstacle, cible). Un filtre de Kalman étendu est donc utilisé pour prédire les trajectoires les plus probables des objets à partir de leurs états estimés. Des stratégies de poursuite et d'évitement ont aussi été développées en fonction des trajectoires prédites des objets détectés. Pour des raisons de sécurité, la conception de stratégies d'évitement de collision à la fois efficaces et robustes est primordiale au guidage d'UAV. Une nouvelle stratégie d'évitement d'obstacles par approche probabiliste a donc été développée. La méthode cherche à minimiser la probabilité de collision entre l'UAV et tous ses obstacles détectés sur l'horizon de prédiction, tout en s'assurant que, à chaque pas de prédiction, la probabilité de collision entre l'UAV et chacun de ses obstacles détectés ne surpasse pas un seuil prescrit. Des simulations sont présentées au cours de cette thèse pour démontrer l'efficacité des algorithmes proposés.

This study is an offensive realist comparative analysis of the drone policies used by the Bush administration and Obama administration during the Global War on Terror. The emergence of violent non-state actors have led to states having to develop new strategies for countering them. Drones were used by the United States in order to combat al-Qaeda, using new technologies in warfare to achieve this goal. This research addresses a gap by focusing on drone policies rather than the legality and morality of drone attacks or the effectiveness of drone strikes. This paper uses three offensive realist concepts; power maximisation, security maximisation and preventive warfare. The method used is a comparative analysis of the Bush administration, with the operationalisation of each concept. The findings of this research show that the drone policies used by each administration can be explained by the three offensive realist concepts. The Obama and Bush administrations prioritised power and security maximisation, while the Obama administration employed preventive drone strikes to a higher degree than the Bush administration. These findings are significant as offensive realism was able to explain the drone policies made by the United States during an asymmetric conflict.

This thesis attempts to illustrate how texts engage in a struggle to define social reality. The academic and political debate on the drone provides a case study illustrating pro-drone, anti-drone and drone-deconstructivist text´s attempt to define social reality around the drone. Through this exploration the thesis contests the positivist notion that reality is fixed and can be discovered. The thesis utilizes IR poststructuralism as a theoretical guideline and engages with a range of poststructuralist concepts and sub-theories identified in previous research. Methodologically, the thesis conducts a qualitative content analysis on three journal articles, two speeches and two chapters of a report. Moreover, the thesis conducts a discourse analysis to contextualize findings and provide in-depth analysis of examples. The thesis presents and analyzes findings in accordance with the categories identified in coding, highlights an example through discourse analysis and links findings to the research question concerned with texts struggle to define social reality. The thesis concludes by arguing that the drone-debate is a “battlespace” for defining social reality, thus contesting the positivist notion that social reality is fixed.