Ce mémoire présente les concepts théoriques et pratiques qui permettront à un groupe de véhicules sans-pilote hétérogène d'estimer la position et l'orientation relative des véhicules voisins en temps réel. Le système de localisation relative leur permettra donc de se coordonner localement afin d'effectuer des missions d'équipe telles que le transport d'une charge et le vol en formation. La localisation relative est effectuée avec une matrice de capteurs ultrasoniques et un algorithme de différence de temps d'arrivée des signaux. Le système présenté peut être monté à bord de véhicules sans-pilote et ne requiert pas d'architecture externe.

Behavioural responses can reveal important fitness trade-offs and ecological traps in evolutionarily novel contexts created by anthropogenic stimuli, and are of increasing conservation concern due to possible links to population-level impacts. This thesis illustrates the use of proxies for energy acquisition and expenditure within multivariate and state-based modelling approaches to quantify the relative time and energetic costs of behavioural disturbance for a deep-diving marine mammal (Physeter macrocephalus) in foraging grounds in Kaikoura Canyon (New Zealand) and near Lofoten Islands (Norway). A conceptual framework is first developed to identify and explore links between individual motivation, condition and external constraints to behavioural disturbance [Chapter 1]. The following chapters then use data from behavioural response studies (BRS) to: 1) derive biologically relevant metrics of behaviour [all chapters], 2) investigate effects of boat-based focal follows and tagging procedures [Chapters 2-3], and 3) relate responses to specific disturbance stimuli (distance, approach, noise) from whale-watching [Chapter 2], naval sonar and playback of presumed natural predator (killer whale Orcinus orca) sounds [Chapter 4]. A novel hidden state model was developed to estimate behavioural budgets of tagged sperm whales from multiple streams of biologging (DTAG) data [Chapter 3]. Sperm whales traded off time spent at foraging depths in a non-foraging and non-resting state in response to both tag boat presence, 1-2 kHz naval sonar (SPL 131-165 rms re 1μPa) and mammal-eating killer whale sound playbacks, indicating that parallel non-lethal costs were incurred in both anthropogenic disturbance and presumed antipredatory contexts. While behavioural responses were highly variable by individual, biologically informed state-based models appeared effective to control for variability in energy proxies across different functional contexts. These results and Chapter 5 “linking buzzes to prey” demonstrate that behavioural context is a signal that can aid understanding of how individual non-lethal disturbance responses can impact fitness.

[pt] Esta dissertação apresenta o projeto e desenvolvimento de um sistema para detecção e desvio de obstáculos para veículos aéreos não tripulados (VANTs), implementado por um controlador neuro-fuzzy. Neste contexto, este trabalho apresenta uma revisão teórica sobre veículos aéreos não tripuláveis, legislação brasileira aplicável, métodos de detecção de obstáculos, lógica nebulosa e redes neurais. O controlador desenvolvido foi implementado de forma a imitar as ações realizadas por um operador humano, visando desviar de obstáculos encontrados no caminho de navegação do VANT. Regras de inferência são estabelecidas baseadas na consultoria de especialistas da área e os pesos ajustados pela rede neural. O processo de tomada de decisão ocorre levando em consideração as informações coletadas por um Lidar multicanal e sensores ultrassônicos embarcados no VANT. Por sua vez, o algoritmo desenvolvido foi incorporado em um controlador de vôo comercial. O sistema completo do quadricóptero é detalhado, destacando as principais características de todos os sensores e do controlador de vôo. Os resultados das simulações computacionais e testes experimentais são apresentados, discutidos e comparados, a fim de avaliar o desempenho do sistema desenvolvido. / [en] This dissertation presents the design and development of an obstacle detection and avoidance system for unmanned aerial vehicles, implemented by a neuro-fuzzy controller. In this context, this work presents a theoretical review of unmanned aerial vehicles, the applicable Brazilian legislation, obstacle detection methods, fuzzy logic and neural networks. The developed controller was implemented in order to mimic the actions taken by a human operator, aiming at avoiding obstacles found in the navigation path of the UAV. Inference rules were established based on consultation with specialists in the field and the weights adjusted by neural networks. The decisionmaking process takes into account information collected by a multichannel Lidar and ultrasonic sensors embedded in the UAV. In turn, the developed algorithm was embedded in a commercial flight controller. The complete quadricopter system is detailed, highlighting the key features of all sensors and the flight controller. The results of computational simulations and experimental tests are presented, discussed and compared, in order to evaluate the performance of the developed system.

La recherche d'un compromis entre cadence et qualité de l'image est un enjeu majeur dans la définition de nouvelles méthodes d'imagerie cohérente. L'objectif de cette thèse est de proposer des modes d'imagerie innovants exploitant au mieux les performances des plateformes matérielles. Les applications étant l'échographie ultrasonore et l'acoustique sous-marine. Le manuscrit propose d'abord un cadre formel au problème du compromis cadence/contraste et démontre l'optimalité des méthodes de synthèse d'émission. L'équivalence, en termes de contraste et de résolution, entre la synthèse canonique et l'imagerie focalisée est ensuite établie. Le problème du rapport signal à bruit est adressé par l'utilisation de matrice d'émission maximisant l'énergie d'insonification (matrice de Hadamard). On introduit ensuite une méthode originale, dite d' " allègement de synthèse ", permettant une amélioration significative de la cadence. Pour cela, on recherche un optimum matriciel en terme de maximisation de l'énergie des faisceaux d'émission dans la zone d'insonification. Cet optimum est une matrice composée de séquences sphéroïdales aplaties. On introduit alors la " synthèse sphéroïdale " qui exploite les propriétés de cette matrice et permet de déplacer le compromis cadence/contraste en un compromis cadence/taille de l'image. La dernière partie de ce manuscrit est consacrée à la mesure et la correction des mouvements propres de la plateforme par des méthodes de micronavigation. Les résultats de ce travail de thèse font l'objet de validations expérimentales sur un échographe ultrasonore à 128 voies programmables et sur un sonar à antenne synthétique.

Au cours des dernières décennies, les avancées en termes de technologies robotiques sous-marines ont permis de réaliser des levés sur les fonds marins à l'aide de véhicules sous-marins autonomes (AUV). Ainsi, équiper un AUV avec un sonar latéral permet de scanner une vaste zone de manière rapide. Naturellement, les forces armées se sont intéressées à de tels dispositifs pour effectuer des missions de chasses aux mines rapides et sécurisées pour le facteur humain. Néanmoins, analyser des images sonar par un ordinateur plutôt que par un opérateur reste très complexe. En effet, les chaînes de reconnaissance automatique de cibles (ATR) doivent faire face à la variabilité de l'environnement marin et il a été démontré qu'une forte relation existe entre la texture d'une image et la difficulté d'y détecter des mines. Effectivement, sur des fonds fortement texturés, voire encombrés, les performances d'une chaîne ATR peuvent être très dégradées. Ainsi, intégrer des informations environnementales dans le processus apparaît comme une piste crédible pour améliorer ses performances. Ces travaux de thèse proposent d'étudier la manière de décrire cet environnement marin et comment l'intégrer dans un processus ATR. Pour répondre à ces défis, nous proposons tout d'abord une nouvelle représentation des images sonar basée sur l'utilisation du signal monogène. Ce dernier permet d'extraire des informations énergétiques, géométriques et structurelles sur la texture locale d'une image. La nature multi-échelle de cet outil permet de tenir compte de la variabilité en taille des structures sous-marines. Ensuite, le concept de dimension intrinsèque est introduit pour décrire une image sonar en termes d'homogénéité, d'anisotropie et de complexité. Ces trois descripteurs sont directement reliés à la difficulté de détection des mines sous-marines dans un fond texturé et permettent de réaliser une classification très précise des images sonar en fonds homogènes, anisotropes et complexes. De notre point de vue, la chasse aux mines sous-marines ne peut pas être réalisée de la même manière sur ces trois types de fond. En effet, leurs natures et caractéristiques propres mènent à des challenges variés pour le processus ATR. Pour le démontrer, nous proposons de réaliser un premier algorithme de détection spécifique, appliqué aux zones anisotropes, qui prend en considération les caractéristiques environnementales de ces régions. / In the last decades, advances in marine robot technology allowed to perform accurate seafloor surveys by means of autonomous underwater vehicles (AUVs). Thanks to a sidescan sonar carried by an AUV, a wide area can be scanned quickly. Navies are really interested in using such vehicles for underwater mine countermeasures (MCM) purposes, in order to perform mine hunting missions rapidly and safely for human operators. Nevertheless, on-board intelligence, which intends to replace human operator for sonar image analysis, remains challenging. Current automatic target recognition (ATR) processes have to cope with the variability of the seafloor. Indeed, there is a strong relationship between the seafloor appearance on sidescan sonar images and the underwater target detection rates. Thus, embed some environmental information in the ATR process seems to be a way for achieving more effective automatic target recognition. In this thesis, we address the problem of improving the ATR process by taking into account the local environment. To this end, a new representation of sonar images is considered by use of the theory of monogenic signal. It provides a pixelwise energetic, geometric and structural information into a multi-scale framework. Then a seafloor characterization is carried out by estimating the intrinsic dimensionality of the underwater structures so as to describe sonar images in terms of homogeneity, anisotropy and complexity. These three features are directly linked to the difficulty of detecting underwater mines and enable an accurate classification of sonar images into benign, rippled or complex areas. From our point of view, underwater mine hunting cannot be performed in the same way on these three seafloor types with various challenges from an ATR point of view. To proceed with this idea, we propose to design a first specific detection algorithm for sand rippled areas. This algorithm takes into consideration an environmental description of ripples which allow to outperform classic approaches in this type of seafloor.

Effective management of the potential environmental impacts of naval sonar requires quantitative data on the behaviour and hearing physiology of cetaceans. Here, novel experimental and analytical methods were used to obtain such information and to test the effectiveness of an operational mitigation method for naval sonar. A Bayesian method was developed to estimate whale locations through time, integrating visual observations with measurements from on-animal inertial, acoustic, depth, and Fastloc-GPS sensors. The track reconstruction method was applied to 13 humpback whale (Megaptera novaeangliae) data sets collected during a multi-disciplinary behavioural response study in Norwegian waters. Thirty-one controlled exposure experiments with and without active transmissions of 1.3-2 kHz sounds were conducted using a moving vessel that towed a sonar source. Dose-response functions, representing the relationships between measured sonar dose and behavioural responses identified from the reconstructed tracks, predicted that 50% of the humpbacks would initiate avoidance at a relatively high received sound pressure level of 166 dB re 1 µPa. Very similar dose-response functions were obtained for cessation of feeding. In a laboratory study, behavioural reaction times of a harbour porpoise (Phocoena phocoena) to sonar-like sounds were measured using operant conditioning and a psychoacoustic method. Auditory weighting functions, which can be used to improve dose-response functions, were obtained for the porpoise based on the assumption that sounds of equal loudness elicit equal reaction time. Additional analyses of the humpback whale data set provided evidence that ramp-up of naval sonar mitigates harmful sound levels in responsive cetaceans located directly in the path of the source, and suggested that a subset of the humpback whale population, such as mother-calf pairs, and more responsive species would benefit from the use of sonar ramp-up. The findings in this thesis are intended to inform sound exposure criteria and mitigation guidelines for anthropogenic noise exposure to cetaceans.

Den svenska vattenkraften står för nästan 50 % av landets elproduktion vilket gör dess konstruktioner till en viktig del av samhällets infrastruktur. För att ett vattenkraftverk ska fungera effektivt och säkert ställs höga krav på komponenternas funktion och kondition. I dagsläget genomförs därför rutinmässiga kontroller på, i och runt dessa anläggningar. För de aggregatnära inre vattenvägarna genomförs detta ca vart åttonde år i samband med reparationer av turbinerna, vilka då måste tömmas på vatten. För till- och utloppstunnlarna, som ofta är drivna i berg, finns det dock inte några sådana rutiner. Detta i kombination med att majoriteten av Sveriges vattenkraftverk har uppnått en ålder på över 50 år gör att branschen har börjat bli intresserade av att få kännedom om skicket på tunnlarna. Att genom-föra inspektioner genom att torrlägga dessa är dock en tidsödande process som kräver att kraftverket är avstängt under en lång tid och därmed ger en stor ekonomisk förlust för ägaren. Detta arbete har därför utförts med syfte att utvärdera till vilken grad inspektioner med undervattensfarkoster, ROV, kan användas som alternativ till inspektioner av inre vatten-vägar, både aggregatnära och i till- och utloppstunnlarna. En litteraturstudie har genomförts, om vattenkraftverk, vanliga skador, dagens inspektionsmetoder samt hur en ROV är uppbyggd och fungerar. Arbetet innefattar även tre praktiska moment bestående av deltagande på en inspektion av torrlagda aggregatnära inre vattenvägar och en inspektion med ROV i en 2,5 km lång utloppstunnel samt egna försök med inhyrd ROV för att undersöka vilka typer av betongskador som är möjliga att detektera. Resultatet från samtliga delar i arbetet pekar på att besiktningar med ROV kan övervägas för flera olika typer av inspektioner. För till- och utloppstunnlar är denna typ av inspektion många gånger den enda rimliga. För de aggregatnära inre vattenvägarna ger dagens inspektionsmetoder med torrläggning ett fullgott resultat, men ROV skulle kunna vara ett alternativ vid svårtillgängliga delar, t.ex. på grund av utformning eller stora läckage. / Swedish hydro power accounts for nearly 50 % of the country's electricity production which makes the constructions within this production system an important part of the societal infrastructure. To ensure safe and efficient operation of these hydroelectric power plants high demands are put on the components function and condition. Therefore, routine inspections are carried out inside and around these facilities. For the waterways directly connected to the power station, such as penstocks, wicket gates and draft tubes, these checkups are made about every eight years simultaneously as planned repairs of the turbine, when the waterways have to be emptied. For the head and trail race tunnels, which often are driven in bedrock, there are no such routines. This, in combination with the fact that the majority of the Swedish hydroelectric plants have attained the age of over 50 years, has led to an increased interest within the industry in ways to increase their knowledge about the condition of these water passageways. To conduct inspections by draining the tunnels is however a very time consuming process that requires the power plant to be turned off for a long time causing a major financial loss for the owner. The purpose of this master thesis is to assess the extent to which inspections with remotely operated vehicles, ROV, can be used as an alternative method for inspections of covered waterways, both those close to the power house and the head and tail race tunnels. A literature study of hydro power plants, common damages, current inspection methods and the construction and function of an ROV, has been conducted. The method also includes three practical activities, participation at inspections in drained waterways and at an ROV inspection of a 2.5 km tail race tunnel. Furthermore, tests with a rented ROV were carried out to investigate what type of concrete damages that are possible to detect using this method. The result from all parts in this thesis indicates that examinations by ROV can be considered for several types of inspections. For head and tail race tunnels, this type of inspection is often the only reasonable method. For the waterways directly connected to the power station, current inspection methods by draining show good result. But the ROV could here be an alternative in inaccessible areas, for instance due to waterway design or heavy leakage for instance.

Sidescan sonars are tools used in seabed inspection and imagery. As a smaller and cheaper compared to the alternatives tool, it has attracted attention and many studies have been developed to extract information about the seabed altitude from the produced images. The main issue is that sidescan sonars do not provide elevation angle information, therefore a 3D map of the seabed cannot be inferred directly. One of the most recent techniques to tackle this problem is called neural rendering [1], in which the sea surface bathymetry is implicitly represented using a neural network. The purpose of this thesis is (1) to find the minimum altitude change that can be detected using this technique, (2) to check whether the position of the sonar ensonification has any effect on these results, and (3) to check from how many sides is it sufficient to ensonify the region with altitude change in order to detect it confidently. To conduct this research, simulations of missions conducted by an autonomous underwater vehicle with sidescan sonar heads on both sides are done on a map, where different objects from various sizes and shapes are put. Then, neural rendering is used to reconstruct the bathymetry of the maps before and after the object insertion from the sidescan sonar. The reconstructed seabed elevations are then compared and the objects with the smallest size or altitude that were detected (meaning that the predicted height from the model trained on the map with the objects is significantly larger than that of the model trained on the initial map) would be the answer to the first question. Then, those smallest objects are again put on the same map, and now smaller autonomous underwater vehicle missions are used to check how many sides are need so that the objects are still detectable. The conducted experiments suggest that objects with bathymetry elevation in the range of centimeters can be detected, and in some cases ensonification from 2 sides is sufficient to detect an object with confidence. / Sidenskannings-sonarer spelar en avgörande roll i inspektionen av havsbotten och erbjuder kostnadseffektiva alternativ till traditionella verktyg. Bristen på information om elevationsvinklar utgör dock en utmaning för att direkt härleda en 3D-karta över havsbotten. Denna avhandling undersöker tillämpningen av neural rendering [1], en nyligen utvecklad teknik som implicit representerar havsytsbathymetri med neurala nätverk, för att adressera denna begränsning. Målen med denna forskning är tre: (1) att bestämma den minsta detekterbara höjdändringen med hjälp av neural rendering, (2) att bedöma effekten av sonarens ensonifieringsposition på detektionsresultaten och (3) att undersöka det minsta antalet sidor som krävs för pålitlig objektdetektion i områden med höjdändringar. Metoden innefattar simuleringar av autonoma undervattensfordonsuppdrag utrustade med sidenskannings-sonarer på båda sidor. Olika objekt av varierande storlekar och former introduceras på en karta, och neural rendering används för att återskapa bathymetrier före och efter objektets insättning. Analysen fokuserar på att jämföra de återskapade havsbottenhöjderna och identifiera de minsta objekten eller höjdändringarna som är möjliga att detektera med modellen. Därefter återintroduceras dessa minimala objekt på kartan, och mindre uppdrag med autonoma undervattensfordon genomförs för att fastställa det minsta antalet sidor som krävs för pålitlig detektion. Forskningsresultaten indikerar att objekt med höjdändringar i centimeterskalan kan detekteras pålitligt. Dessutom tyder experimenten på att i vissa fall är ensonifiering från endast två sidor tillräckligt för pålitlig objektdetektion. Denna forskning bidrar med värdefulla insikter för att optimera sidenskanningssonarapplikationer för havsbotteninspektion, vilket erbjuder potentiella förbättringar av effektivitet och kostnadseffektivitet för undervattensutforskning och kartläggning.

Common Machine-Learning (ML) approaches for scene classification require a large amountof training data. However, for classification of depth sensor data, in contrast to image data, relativelyfew databases are publicly available and manual generation of semantically labeled 3D point clouds isan even more time-consuming task. To simplify the training data generation process for a wide rangeof domains, we have developed theBLAINDERadd-on package for the open-source 3D modelingsoftware Blender, which enables a largely automated generation of semantically annotated point-cloud data in virtual 3D environments. In this paper, we focus on classical depth-sensing techniquesLight Detection and Ranging (LiDAR) and Sound Navigation and Ranging (Sonar). Within theBLAINDERadd-on, different depth sensors can be loaded from presets, customized sensors can beimplemented and different environmental conditions (e.g., influence of rain, dust) can be simulated.The semantically labeled data can be exported to various 2D and 3D formats and are thus optimizedfor different ML applications and visualizations. In addition, semantically labeled images can beexported using the rendering functionalities of Blender.

Substrate composition plays a critical role in determining the spawning success of Atlantic sturgeon. A benthic analysis of the tidal freshwater portion of the James River, Virginia, was performed to locate and protect remaining sturgeon spawning habitat within the James River system. I modeled structural habitat, substrate distribution, and river bathymetry from Richmond, Virginia to the Appomattox River confluence. A classification model was developed to describe the dominant substrate type (mud/silt, sand, gravel, bedrock) using side scan sonar data collected from August 2011-Febuary 2012. River depth, bottom imagery, substrate density (hardness), and ground truth substrate samples were interpolated into a GIS model to spatially describe and quantify essential sturgeon spawning habitat. Finally, I attempted a change analysis of historical substrate composition throughout the study area. Gravel, cobble, and bedrock, swept clean of silt or mud, was deemed a hard bottom substrate suitable for spawning success. Mud and silt dominated the vast majority of river substrate, representing approximately 67 % of river bottom surveyed. Sand comprised 17 % of river bottom, gravel represented 11 % and bedrock represented 5 %. Sixteen percent of the reach was hard bottom habitat consisting of a bed substrate dominated by gravel, cobble, or bedrock. Regions of hard bottom habitat found at depths ≥ 10 m were selected to model essential sturgeon spawning habitat. The river bottom within the reach contained approximately 8 % essential spawning habitat. The majority of hard bottom habitat was located in major bends of the river where scouring occurs. The historical comparison of available hard bottom habitat identified a 28 % loss of hard bottom since 1853. The greatest losses in hard bottom occurred in the upper portions of the study area (55 % loss in hard bottom habitat). Hard bottom habitat lost in the lower portion of the study area was partially offset by the creation of new hard bottom habitat within the narrow channel cuts bypassing Jones Neck and Turkey Island. Historical comparison of the Hatcher Island, Turkey Island, and Jones Neck oxbows identified heavy siltation and reduced depths likely due to anthropogenic alterations in the meander bends linked to shipping channel creation. The altered flow regime has resulted in increased sedimentation and has drastically reduced available hard bottom substrate within the natural channel of Jones Neck and Turkey Island. The increased availability of hard bottom habitat within the confines of the shipping channel has indicated that the alteration of the river bottom, through flow modification and dredging practices, may have replaced a portion of lost historical spawning habitat. Fisheries managers could use the data from the substrate analysis to better understand and protect essential areas necessary for Atlantic sturgeon spawning success.