A striking feature of swarm robotics is its ability to solve complex tasks through simple local interactions between robots. Those interactions require a good infrastructure in communication and localization. However, in underwater environments, the severe attenuation of radio waves complicates communication and localization of different vehicles. Existing literature on underwater swarms use centralized network topology which require physical vicinity to the central node to ensure reliability. We are interested in building a decentralized underwater swarm with a decentralized network topology that only requires neighbour communication and self-localization. We develop a simple leader-follower interaction rule where the follower estimates the leader's position and acts upon that estimation. The leader shines a 450 nm diffracted blue laser that the follower uses to continuously align its light sensors to the light source. Furthermore, the leader's laser can be modulated for explicit communication purposes. The proposed leader-follower approach produces satisfactory results in surge and sway axes, however, it is not robust against illumination changes in the environment. We then proceed to solve the self-localization problem, by fusing Inertial Measurement Unit (IMU) values with the thrust to estimate a robot’s position. In an Ardusub Simulation in the loop (SITL), the particle filter showed a slightly better performance than the Extended Kalman Filter (EKF) in the surge axis. However, both filters are prone to drifting after a while. We have observed that IMU values need to be filtered properly or another reliable sensor must be used alternatively.

Cette thèse porte sur la localisation de sources acoustiques sous-marines avec application à une expérience en mer. Nous proposons une nouvelle méthode d'appariement basée sur une métrique appelée distance de Hausdorff (HD) en tant que fonction de coût à minimiser, afin d'effectuer l'inversion de localisation. La localisation 2D, en distance et en profondeur, est réalisée en faisant correspondre les schémas de différence de temps d'arrivée (TDOA) en utilisant un seul hydrophone à la réception et en faisant correspondre le TDOA et l'Angle d'arrivée (AOA) lors de l'utilisation d'un tableau des hydrophones à la réception, entre des séquences respectivement mesurées et modélisées. Le TDOA modélisé a été obtenu sur la base du modèle de propagation acoustique Ray-path. Les ensembles de données analysés ici ont été collectés dans un contexte de localisation passive en considérant une cible immobile et dans deux expériences : la cuve acoustique de GIPSA-LAB utilisant des systèmes coopératifs et non coopératifs vérifiés par des simulations du rapport signal sur bruit et sur la campagne ALMA 2015, collectée par la Direction générale de l'armement (DGA) en utilisant un système coopératif qui s'est déroulé dans un environnement en eaux peu profondes de la côte sud de la France. Au cours de l’expérience ALMA, les données acoustiques ont été mesurées sur un réseau linéaire vertical (VLA) de 10 m de haut, composé de 64 hydrophones, ce qui permet non seulement d’adapter le TDOA mais également l’angle d’arrivée (AOA). Plusieurs variantes de la distance de Hausdorff sont appliquées dans deux processus différents: premièrement, séparément dans chaque hydrophone, puis combinées pour améliorer la précision de la localisation (diversité spatiale), et la seconde où les informations des différents hydrophones sont combinées (formation de faisceaux), pour trouver l'emplacement cible. Les résultats des deux processus sont comparés et prouvés pour réduire l'ambiguïté soit la profondeur et la portée, améliorant ainsi la précision finale. Le Cramer Rao Bound montrant la variance minimale effectuée sur la base d’équations déterministes est présenté avec le meilleur résultat de chaque processus. Une performance et une précision très satisfaisantes sont obtenues. Les conclusions et les perspectives de ce travail sont discutées à la fin. / This thesis addresses an acoustic underwater source localization with application to an at-sea experiment. We propose a new matching method based on a fit-metric called as Hausdorff distance (HD) as a cost-function to be minimized, in order to perform the localization inversion. The 2-D localization, in range and depth, is performed by matching the patterns of time difference of arrival (TDOA) when using only one hydrophone at the reception and by matching the TDOA and the Angle of Arrival (AOA) when using an array of hydrophones at the reception, between respectively measured and modeled sequences. The modelled TDOA was obtained based on the Ray-path acoustic propagation model. The data sets analyzed here were collected during two experiments in a context of passive localization considering a motionless target: The tank of GIPSA-LAB using cooperative and non-cooperative systems which were verified by simulations with respect to the signal-to-noise ratio and the ALMA 2015, collected by the Direction générale de l’armement (DGA) using a cooperative system which took place in a shallow water environment of the southern coast of France. During the ALMA experiment the acoustic data were measured over a 10m-high vertical linear array (VLA), composed of 64 hydrophones, allowing not only matching the TDOA but also the Angle of Arrival (AOA). Several variants of the Hausdorff Distance are applied in two different processes: First, separately in each single hydrophone, and then combined in order to improve the localization accuracy (spatial diversity), and the second, the information from the different hydrophones are combined (beamforming) and the HD variants are applied to find the target location. The results of both processes are compared and proved to reduce the ambiguity either is depth and in range, thus improving the final accuracy. The Cramer Rao Bound showing the minimal variance performed based on deterministic equations is presented with the best result of each process. Very satisfactory performance and accuracy are obtained. The conclusions and perspectives of this work are discussed at the end.

Autonomous vehicle navigation is a complex and challenging task. Land and aerial vehicles often use highly accurate GPS sensors to localize themselves in their environments. These sensors are ineffective in underwater environments due to signal attenuation. Autonomous underwater vehicles utilize one or more of the following approaches for successful localization and navigation: inertial/dead-reckoning, acoustic signals, and geophysical data. This thesis examines autonomous localization in a simulated environment for an OpenROV Underwater Drone using a Kalman Filter. This filter performs state estimation for a dead reckoning system exhibiting an additive error in location measurements. We evaluate the accuracy of this Kalman Filter by analyzing the effect each parameter has on accuracy, then choosing the best combination of parameter values to assess the overall accuracy of the Kalman Filter. We find that the two parameters with the greatest effects on the system are the constant acceleration and the measurement uncertainty of the system. We find the filter employing the best combination of parameters can greatly reduce measurement error and improve accuracy under typical operating conditions.

Le but de cette thèse est l'étude et la mise en oeuvre d'un système de localisation sous-marine compact et simple à mettre en place pour une utilisation en zones portuaires, côtières et environnements peu profonds. Nous proposons un système SBL (Small Distance Baseline) avec un nombre réduit de transducteurs (une balise d'émission et deux hydrophones). La configuration géométrique du système étant contraignante (hydrophones proches) la précision du positionnement obtenue par les méthodes classiques est faible. Nous proposons une nouvelle méthode de localisation améliorant la précision. La balise à localiser émet un signal à étalement de spectre. La différence de distance entre les trajets des signaux des hydrophones est mesurée par corrélation. Nous proposons deux boucles de poursuites pour l'estimation conjointe de la fréquence Doppler et du délai entre les signaux reçus. Ces techniques de poursuite basées sur un filtre de Kalman sont implémentées en boucle fermée et ouverte. Les observations TDOA (Time Difference Of Arrival) conduisent à utiliser une technique de localisation hyperbolique. Nous proposons une représentation statistique qui exploite la géométrie de notre système de mesure pour déterminer une zone de localisation probable autour de chaque hyperbole. En utilisant des positions de bateau successives, on construit une densité de probabilité dont le maximum définit la position de la balise. On montre sur données synthétiquesque pour un bruit de mesure réaliste, il est possible de déterminer la position de la balise avec une précision submétrique. Les expérimentations réelles confirment la faisabilité du système et la précision obtenue est dans ce cas métrique. / The purpose of that thesis work is the research and implementation of a compact and simple underwater localization system that aim to be used in ports, coastal areas and other low-depth environments. Our system is SBL (Small Distance Baseline), with a small number of transceivers (only one emitter and two hydrophones). Due to the system's geometric configuration not being optimal (both hydrophones are close to one another), the precision obtained using classical approaches is poor. We therefore propose a new localization approach that will improve it. The emitter we wish to localize emit a spread spectrum signal. The time difference of arrival (TDOA) between the two hydrophones is then determined using correlations methods. We propose in our work two tracking loops that will estimate both the delay and the doppler frequency between the signals. Using a Kalman filter , those methods are implemented respectively in open and close loop. From each TDOA measurement, we can calculatea hyperbolic area of possible emitter location. We thus use a statistical model which takes into account the local geometry of our measurements system in order to create a probable localization area around each hyperbole. By using the measurements at several different boat positions, we create a probability density whose maximum will be centered around the emitter's position. We show that, on simulated data, it is possible to localize the beacon with a precision beneath a meter with a realistic noise level. Experimental work and real data collection confirm that the method can in that context achieve the same result with a precision of a few meters.

Robust state estimation solutions must deal with faulty measurements, called outliers, and unknown data associations, which lead to multiple feasible hypotheses. Take, for instance, the scenario of tracking two indistinguishable targets based on position measurements, where each measurement could refer to either of the targets or even be a faulty reading. Common estimation methods model the state as having a unimodal distribution, so they are called unimodal methods. Likewise, multimodal methods model the state as a multimodal distribution. Difficult problems, such as autonomous underwater vehicle (AUV) navigation relying on acoustic localization, frequently involve recurring outliers. In these situations, the correct hypothesis only emerges as the most likely one when a substantial number of measurements are considered. Robust solutions for these problems need to consider multiple hypotheses simultaneously, which, in turn, calls for the representation of multimodal distributions. In this work, a novel approximate inference method is presented, called the Gaussian mixture sum-product algorithm (GM-SPA), as it implements the sum-product algorithm (SPA) for Gaussian mixtures. The GM-SPA can exactly represent under-constrained linear measurements and approximate important non-linear models, such as range measurements and 2D pose kinematics. The outlier robustness of the GM-SPA is tested and compared against the particle filter (PF) and multimodal incremental smoothing and mapping (MMiSAM), both of which are non-parametric methods. Robustness, accuracy, and run-time are improved in simulation tests. The test problems include 1D localization with unknown data association, 3D linear target tracking with correlated outliers, and 2D range-only pose estimation with Gaussian mixture noise. / Robusta lösningar för tillståndsuppskattning måste kunna hantera felaktiga mätningar, så kallade outliers, och okända dataassociationer, vilket leder till flera möjliga hypoteser. Ta till exempel scenariot att spåra två likadana mål baserat på positionsmätningar, där varje mätning kan tillhöra något av målen eller till och med vara en felaktig avläsning. Vanliga skattningsmetoder modellerar tillståndet som en unimodal fördelning, och kallas därför unimodala metoder. På samma sätt modellerar multimodala metoder tillståndet som en multimodal fördelning. Svåra problem, som navigering av autonoma undervattensfarkoster (AUV) med hjälp av akustisk lokalisering, involverar ofta upprepade outliers. I dessa situationer framstår den korrekta hypotesen som den mest sannolika först när ett stort antal mätningar beaktas. Robusta lösningar för dessa problem måste ta hänsyn till flera hypoteser samtidigt, vilket i sin tur kräver representation av multimodala fördelningar. I detta arbete presenteras en ny approximativ inferensmetod, kallad Gaussian mixture sum-product algorithm (GM-SPA), eftersom den implementerar sum-product algorithm (SPA) för gaussiska blandningar. GM-SPA kan representera underbegränsade linjära mätningar exakt och approximera viktiga icke-linjära modeller, till exempel avståndsmätningar eller 2D-posekinematik. GM-SPA:s robusthet mot outliers testas och jämförs med partikelfiltret (PF) och multimodal incremental smoothing and mapping (MM-iSAM), som båda är icke-parametriska metoder. Robusthet, noggrannhet och körtid förbättras i simuleringstester. Simulerade tester inkluderar 1D-lokalisering med okänd dataassociation, 3D linjär målföljning med korrelerade outliers och 2D-ställningsuppskattning av endast räckvidd med Gaussiskt blandningsljud. / Soluções robustas para estimação de estado devem lidar com medidas defeituosas, chamadas de outliers, e com associações de dados desconhecidas, que levam a múltiplas hipóteses possíveis. Considere-se, por exemplo, o cenário de rastreamento de dois alvos indistinguíveis com base em medidas de posição, em que cada medida pode-se referir a qualquer um dos alvos ou até mesmo ser uma leitura defeituosa. Métodos de estimação comuns modelam o estado como tendo uma distribuição unimodal, sendo assim chamados de métodos unimodais. Da mesma forma, métodos multimodais modelam o estado como uma distribuição multimodal. Problemas difíceis, como a navegação de veículos subaquáticos autónomos (AUVs) baseada em localização acústica, frequentemente envolvem outliers recorrentes. Nestas situações, a hipótese correta apenas surge como a mais provável quando um número substancial de medidas é considerado. Soluções robustas para estes problemas precisam de considerar múltiplas hipóteses simultaneamente, o que, por sua vez, exige a representação de distribuições multimodais. Neste trabalho, é apresentado um novo método de inferência aproximada, chamado Gaussian mixture sum-product algorithm (GM-SPA), pois implementa o sum-product algorithm (SPA) para misturas Gaussianas. O GM-SPA pode representar exatamente medidas lineares sub-determinadas e aproximar modelos não lineares importantes, como medidas de distância e cinemática de pose 2D. A robustez a outliers do GM-SPA é testada e comparada com o filtro de partículas (PF) e com multimodal incremental smoothing and mapping (MM- -iSAM), ambos métodos não-paramétricos. A robustez, a exatidão e o tempo de execução em testes de simulação são melhorados. Os problemas de teste incluem localização 1D com associação de dados desconhecida, rastreamento linear de alvos em 3D com outliers correlacionados e estimação de pose 2D com base em medidas de distância com ruído de mistura Gaussiana.