La perception est un point clé pour le fonctionnement d'un véhicule autonome ou même pour un véhicule fournissant des fonctions d'assistance. Un véhicule observe le monde externe à l'aide de capteurs et construit un modèle interne de l'environnement extérieur. Il met à jour en continu ce modèle de l'environnement en utilisant les dernières données des capteurs. Dans ce cadre, la perception peut être divisée en deux étapes : la première partie, appelée SLAM (Simultaneous Localization And Mapping) s'intéresse à la construction d'une carte de l'environnement extérieur et à la localisation du véhicule hôte dans cette carte, et deuxième partie traite de la détection et du suivi des objets mobiles dans l'environnement (DATMO pour Detection And Tracking of Moving Objects). En utilisant des capteurs laser de grande précision, des résultats importants ont été obtenus par les chercheurs. Cependant, avec des capteurs laser de faible résolution et des données bruitées, le problème est toujours ouvert, en particulier le problème du DATMO. Dans cette thèse nous proposons d'utiliser la vision (mono ou stéréo) couplée à un capteur laser pour résoudre ce problème. La première contribution de cette thèse porte sur l'identification et le développement de trois niveaux de fusion. En fonction du niveau de traitement de l'information capteur avant le processus de fusion, nous les appelons "fusion bas niveau", "fusion au niveau de la détection" et "fusion au niveau du suivi". Pour la fusion bas niveau, nous avons utilisé les grilles d'occupations. Pour la fusion au niveau de la détection, les objets détectés par chaque capteur sont fusionnés pour avoir une liste d'objets fusionnés. La fusion au niveau du suivi requiert le suivi des objets pour chaque capteur et ensuite on réalise la fusion entre les listes d'objets suivis. La deuxième contribution de cette thèse est le développement d'une technique rapide pour trouver les bords de route à partir des données du laser et en utilisant cette information nous supprimons de nombreuses fausses alarmes. Nous avons en effet observé que beaucoup de fausses alarmes apparaissent sur le bord de la route. La troisième contribution de cette thèse est le développement d'une solution complète pour la perception avec un capteur laser et des caméras stéréo-vision et son intégration sur un démonstrateur du projet européen Intersafe-2. Ce projet s'intéresse à la sécurité aux intersections et vise à y réduire les blessures et les accidents mortels. Dans ce projet, nous avons travaillé en collaboration avec Volkswagen, l'Université Technique de Cluj-Napoca, en Roumanie et l'INRIA Paris pour fournir une solution complète de perception et d'évaluation des risques pour le démonstrateur de Volkswagen. / Perception is one of important steps for the functioning of an autonomous vehicle or even for a vehicle providing only driver assistance functions. Vehicle observes the external world using its sensors and builds an internal model of the outer environment configuration. It keeps on updating this internal model using latest sensor data. In this setting perception can be divided into two sub parts: first part, called SLAM(Simultaneous Localization And Mapping), is concerned with building an online map of the external environment and localizing the host vehicle in this map, and second part deals with finding moving objects in the environment and tracking them over time and is called DATMO(Detection And Tracking of Moving Objects). Using high resolution and accurate laser scanners successful efforts have been made by many researchers to solve these problems. However, with low resolution or noisy laser scanners solving these problems, especially DATMO, is still a challenge and there are either many false alarms, miss detections or both. In this thesis we propose that by using vision sensor (mono or stereo) along with laser sensor and by developing an effective fusion scheme on an appropriate level, these problems can be greatly reduced. The main contribution of this research is concerned with the identification of three fusion levels and development of fusion techniques for each level for SLAM and DATMO based perception architecture of autonomous vehicles. Depending on the amount of preprocessing required before fusion for each level, we call them low level, object detection level and track level fusion. For low level we propose to use grid based fusion technique and by giving appropriate weights (depending on the sensor properties) to each grid for each sensor a fused grid can be obtained giving better view of the external environment in some sense. For object detection level fusion, lists of objects detected for each sensor are fused to get a list of fused objects where fused objects have more information then their previous versions. We use a Bayesian fusion technique for this level. Track level fusion requires to track moving objects for each sensor separately and then do a fusion between tracks to get fused tracks. Fusion at this level helps remove false tracks. Second contribution of this research is the development of a fast technique of finding road borders from noisy laser data and then using these border information to remove false moving objects. Usually we have observed that many false moving objects appear near the road borders due to sensor noise. If they are not filtered out then they result into many false tracks close to vehicle making vehicle to apply breaks or to issue warning messages to the driver falsely. Third contribution is the development of a complete perception solution for lidar and stereo vision sensors and its intigration on a real vehicle demonstrator used for a European Union project (INTERSAFE-21). This project is concerned with the safety at intersections and aims at the reduction of injury and fatal accidents there. In this project we worked in collaboration with Volkswagen, Technical university of Cluj-Napoca Romania and INRIA Paris to provide a complete perception and risk assessment solution for this project.

Depuis des décennies, les chercheurs essaient de développer des systèmes intelligents pour les véhicules modernes, afin de rendre la conduite plus sûre et plus confortable. Ces systèmes peuvent conduire automatiquement le véhicule ou assister un conducteur en le prévenant et en l'assistant en cas de situations dangereuses. Contrairement aux conducteurs, ces systèmes n'ont pas de contraintes physiques ou psychologiques et font preuve d'une grande robustesse dans des conditions extrêmes. Un composant clé de ces systèmes est la fiabilité de la perception de l'environnement. Pour cela, les capteurs lasers sont très populaires et largement utilisés. Les capteurs laser 2D classiques ont des limites qui sont souvent compensées par l'ajout d'autres capteurs complémentaires comme des caméras ou des radars. Les avancées récentes dans le domaine des capteurs, telles que les capteurs laser 3D qui perçoivent l'environnement avec une grande résolution spatiale, ont montré qu'ils étaient une solution intéressante afin d'éviter l'utilisation de plusieurs capteurs. Bien qu'il y ait des méthodes bien connues pour la perception avec des capteurs laser 2D, les approches qui utilisent des capteurs lasers 3D sont relativement rares dans la littérature. De plus, la plupart d'entre elles utilisent plusieurs capteurs et réduisent le problème de la 3ème dimension en projetant les données 3D sur un plan et utilisent les méthodes classiques de perception 2D. Au contraire de ces approches, ce travail résout le problème en utilisant uniquement un capteur laser 3D et en utilisant les informations spatiales fournies par ce capteur. Notre première contribution est une extension des méthodes génériques de cartographie 3D fondée sur des grilles d'occupations optimisées pour résoudre le problème de cartographie et de localisation simultanée (SLAM en anglais). En utilisant des grilles d'occupations 3D, nous définissons une carte d'élévation pour la segmentation des données laser correspondant au sol. Pour corriger les erreurs de positionnement, nous utilisons une méthode incrémentale d'alignement des données laser. Le résultat forme la base pour le reste de notre travail qui constitue nos contributions les plus significatives. Dans la deuxième partie, nous nous focalisons sur la détection et le suivi des objets mobiles (DATMO en anglais). La deuxième contribution de ce travail est une méthode pour distinguer les objets dynamiques des objets statiques. L'approche proposée utilise une détection fondée sur le mouvement et sur des techniques de regroupement pour identifier les objets mobiles à partir de la grille d'occupations 3D. La méthode n'utilise pas de modèles spécifiques d'objets et permet donc la détection de tout type d'objets mobiles. Enfin, la troisième contribution est une méthode nouvelle pour classer les objets mobiles fondée sur une technique d'apprentissage supervisée. La contribution finale est une méthode pour suivre les objets mobiles en utilisant l'algorithme de Viterbi pour associer les nouvelles observations avec les objets présents dans l'environnement, Dans la troisième partie, l'approche propose est testée sur des jeux de données acquis à partir d'un capteur laser 3D monté sur le toit d'un véhicule qui se déplace dans différents types d'environnement incluant des environnements urbains, des autoroutes et des zones piétonnes. Les résultats obtenus montrent l'intérêt du système intelligent proposé pour la cartographie et la localisation simultanée ainsi que la détection et le suivi d'objets mobiles en environnement extérieur et dynamique en utilisant un capteur laser 3D. / With an anticipation to make driving experience safer and more convenient, over the decades, researchers have tried to develop intelligent systems for modern vehicles. The intended systems can either drive automatically or monitor a human driver and assist him in navigation by warning in case of a developing dangerous situation. Contrary to the human drivers, these systems are not constrained by many physical and psychological limitations and therefore prove more robust in extreme conditions. A key component of an intelligent vehicle system is the reliable perception of the environment. Laser range finders have been popular sensors which are widely used in this context. The classical 2D laser scanners have some limitations which are often compensated by the addition of other complementary sensors including cameras and radars. The recent advent of new sensors, such as 3D laser scanners which perceive the environment at a high spatial resolution, has proven to be an interesting addition to the arena. Although there are well-known methods for perception using 2D laser scanners, approaches using a 3D range scanner are relatively rare in literature. Most of those which exist either address the problem partially or augment the system with many other sensors. Surprisingly, many of those rely on reducing the dimensionality of the problem by projecting 3D data to 2D and using the well-established methods for 2D perception. In contrast to these approaches, this work addresses the problem of vehicle perception using a single 3D laser scanner. First contribution of this research is made by the extension of a generic 3D mapping framework based on an optimized occupancy grid representation to solve the problem of simultaneous localization and mapping (SLAM). Using the 3D occupancy grid, we introduce a variance-based elevation map for the segmentation of range measurements corresponding to the ground. To correct the vehicle location from odometry, we use a grid-based incremental scan matching method. The resulting SLAM framework forms a basis for rest of the contributions which constitute the major achievement of this work. After obtaining a good vehicle localization and a reliable map with ground segmentation, we focus on the detection and tracking of moving objects (DATMO). The second contribution of this thesis is the method for discriminating between the dynamic objects and the static environment. The presented approach uses motion-based detection and density-based clustering for segmenting the moving objects from 3D occupancy grid. It does not use object specific models but enables detecting arbitrary traffic participants. Third contribution is an innovative method for layered classification of the detected objects based on supervised learning technique which makes it easier to estimate their position with time. Final contribution is a method for tracking the detected objects by using Viterbi algorithm to associate the new observations with the existing objects in the environment. The proposed framework is verified with the datasets acquired from a laser scanner mounted on top of a vehicle moving in different environments including urban, highway and pedestrian-zone scenarios. The promising results thus obtained show the applicability of the proposed system for simultaneous localization and mapping with detection, classification and tracking of moving objects in dynamic outdoor environments using a single 3D laser scanner.

Made available in DSpace on 2014-12-17T14:54:56Z (GMT). No. of bitstreams: 1 AndreMS_TESE_1-100.pdf: 5113772 bytes, checksum: 19386f80f787c926c4fb29b85bac4ecf (MD5) Previous issue date: 2011-02-11 / The goal of this work is to propose a SLAM (Simultaneous Localization and Mapping) solution based on Extended Kalman Filter (EKF) in order to make possible a robot navigates along the environment using information from odometry and pre-existing lines on the floor. Initially, a segmentation step is necessary to classify parts of the image in floor or non floor . Then the image processing identifies floor lines and the parameters of these lines are mapped to world using a homography matrix. Finally, the identified lines are used in SLAM as landmarks in order to build a feature map. In parallel, using the corrected robot pose, the uncertainty about the pose and also the part non floor of the image, it is possible to build an occupancy grid map and generate a metric map with the obstacle s description. A greater autonomy for the robot is attained by using the two types of obtained map (the metric map and the features map). Thus, it is possible to run path planning tasks in parallel with localization and mapping. Practical results are presented to validate the proposal / O objetivo desta tese ? apresentar uma t?cnica de SLAM (Localiza??o e Mapeamento Simult?neos) adequada para ambientes planos com linhas presentes no ch?o, de modo a permitir que o rob? navegue no ambiente fundindo informa??es de odometria e de vis?o monocular. Inicialmente, ? feita uma etapa de segmenta??o para classificar as partes da imagem em ch?o e n?o-ch?o . Em seguida, o processadomento de imagem identifica linhas na parte ch?o e os par?metros dessas linhas s?o mapeados para o mundo, usando uma matriz de homografia. Finalmente, as linhas identificadas s?o usadas como marcos no SLAM, para construir um mapa de caracter?sticas. Em paralelo, a pose corrigida do rob?, a incerteza em rela??o ? pose e a parte n?och?o da imagem s?o usadas para construir uma grade de ocupa??o, gerando um mapa m?trico com descri??o dos obst?culos. A utiliza??o simult?nea dos dois tipos de mapa obtidos (m?trico em grade e de caracter?sticas) d? maior autonomia ao rob?, permitindo acrescentar tarefas de planejamento em simult?neo com a localiza??o e mapeamento. Resultados pr?ticos s?o apresentados para validar a proposta

Made available in DSpace on 2014-12-17T14:55:52Z (GMT). No. of bitstreams: 1 VitorMUA_DISSERT.pdf: 1704175 bytes, checksum: aa664311278faf5559b37a1627d7e89e (MD5) Previous issue date: 2011-07-29 / In Simultaneous Localization and Mapping (SLAM - Simultaneous Localization and Mapping), a robot placed in an unknown location in any environment must be able to create a perspective of this environment (a map) and is situated in the same simultaneously, using only information captured by the robot s sensors and control signals known. Recently, driven by the advance of computing power, work in this area have proposed to use video camera as a sensor and it came so Visual SLAM. This has several approaches and the vast majority of them work basically extracting features of the environment, calculating the necessary correspondence and through these estimate the required parameters. This work presented a monocular visual SLAM system that uses direct image registration to calculate the image reprojection error and optimization methods that minimize this error and thus obtain the parameters for the robot pose and map of the environment directly from the pixels of the images. Thus the steps of extracting and matching features are not needed, enabling our system works well in environments where traditional approaches have difficulty. Moreover, when addressing the problem of SLAM as proposed in this work we avoid a very common problem in traditional approaches, known as error propagation. Worrying about the high computational cost of this approach have been tested several types of optimization methods in order to find a good balance between good estimates and processing time. The results presented in this work show the success of this system in different environments / No SLAM (Simultaneous Localization and Mapping), um rob? posicionado em uma localiza??o desconhecida de um ambiente qualquer deve ser capaz de construir uma perspectiva deste ambiente (um mapa) e se localizar no mesmo simultaneamente, utilizando apenas informa??es captadas pelos sensores do rob? e muitas vezes sinais de controle conhecidos. Recentemente, impulsionados pelo avan?o computacional, trabalhos nessa ?rea propuseram usar c?mera de v?deo como sensor e surgiu assim o SLAM Visual. Este possui v?rias abordagens e a grande maioria delas funcionam, basicamente, extraindo caracter?sticas do ambiente, calculando as devidas correspond?ncias e atrav?s destas, e de filtros estat?sticos, estimam os par?metros necess?rios. Neste trabalho ? apresentado um sistema de SLAM Visual Monocular que utiliza registro direto de imagem para calcular o erro de reproje??o entre imagens e m?todos de otimiza??o que minimizam esse erro e assim obter os par?metros relativos ? pose do rob? e o mapa do ambiente diretamente dos pixels das imagens. Dessa forma as etapas de extra??o e correspond?ncia de caracter?sticas s?o dispensadas, possibilitando que nosso sistema funcione bem em ambientes onde as abordagens tradicionais teriam dificuldades. Al?m disso, ao se abordar o problema do SLAM da forma proposta nesse trabalho evitase um problema muito comum nas abordagens tradicionais, conhecido como acumulo do erro. Preocupando-se com o elevado custo computacional desta abordagem foram testados v?rios tipos de m?todos de otimiza??o afim de achar um bom equil?brio entre boas estimativas e tempo de processamento. Os resultados apresentados neste trabalho comprovam o funcionamento desse sistema em diferentes ambientes

Made available in DSpace on 2014-12-17T14:56:19Z (GMT). No. of bitstreams: 1 LuizHRS_DISSERT.pdf: 5685489 bytes, checksum: 6c8bd08a9e57fe6ca72df5d1593b834d (MD5) Previous issue date: 2014-01-24 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior / This work intends to show a new and few explored SLAM approach inside the simultaneous localization and mapping problem (SLAM). The purpose is to put a mobile robot to work in an indoor environment. The robot should map the environment and localize itself in the map. The robot used in the tests has an upward camera and encoders on the wheels. The landmarks in this built map are light splotches on the images of the camera caused by luminaries on the ceil. This work develops a solution based on Extended Kalman Filter to the SLAM problem using a developed observation model. Several developed tests and softwares to accomplish the SLAM experiments are shown in details / Este trabalho visa mostrar uma abordagem pouco explorada do problema de mapeamento e localiza??o simult?neos (SLAM). Comfimde trabalhar emumambiente fechado, uma plataforma rob?tica m?vel deve construir um mapa do ambiente e se localizar dentro deste mapa. A plataforma rob?tica utilizada possui uma c?mera voltada para o teto (ascendente) e odometria para as rodas. As marcas que comp?em o mapa s?o manchas luminosas na imagem capturada pela c?mera causadas pelas lumin?rias no teto. Este trabalho desenvolve uma solu??o baseada no Filtro de Kalman Estendido para o SLAM com um modelo de observa??o desenvolvido. Diversos testes e programas desenvolvidos para realiza??o do SLAM s?o apresentados em detalhes

Este trabalho apresenta uma arquitetura de hardware, baseada em FPGA (Field-Programmable Gate Array) e com multi-câmeras, para o problema de localização e mapeamento simultâneos - SLAM (Simultaneous Localization And Mapping) aplicada a sistemas robóticos embarcados. A arquitetura é composta por módulos de hardware altamente especializados para a localização do robô e para geração do mapa do ambiente de navegação em tempo real com features extraídas de imagens obtidas diretamente de câmeras CMOS a uma velocidade de 30 frames por segundo. O sistema é totalmente embarcado em FPGA e apresenta desempenho superior em, pelo menos, uma ordem de magnitude em relaçãoo às implementações em software processadas por computadores pessoais de última geração. Esse desempenho deve-se à exploração do paralelismo em hardware junto com o processamento em pipeline e às otimizações realizadas nos algoritmos. As principais contribuições deste trabalho são as arquiteturas para o filtro de Kalman estendido - EKF (Extended Kalman Filter) e para a detecção de features baseada no algoritmo SIFT (Scale Invariant Feature Transform). A complexidade para a implementaçãoo deste trabalho pode ser considerada alta, uma vez que envolve uma grande quantidade de operações aritméticas e trigonométricas em ponto utuante e ponto fixo, um intenso processamento de imagens para extração de features e verificação de sua estabilidade e o desenvolvimento de um sistema de aquisição de imagens para quatro câmeras CMOS em tempo real. Adicionalmente, foram criadas interfaces de comunicação para o software e o hardware embarcados no FPGA e para o controle e leitura dos sensores do robô móvel. Além dos detalhes e resultados da implementação, neste trabalho são apresentados os conceitos básicos de mapeamento e o estado da arte dos algoritmos SLAM com visão monocular e estéreo / This work presents a hardware architecture for the Simultaneous Localization And Mapping (SLAM) problem applied to embedded robots. This architecture, which is based on FPGA and multi-cameras, is composed by highly specialized blocks for robot localization and feature-based map building in real time from images read directly from CMOS cameras at 30 frames per second. The system is completely embedded on an FPGA and its performance is at least one order of magnitude better than a high end PC-based implementation. This result is achieved by investigating the impact of several hardwareorientated optimizations on performance and by exploiting hardware parallelism along with pipeline processing. The main contributions of this work are the architectures for the Extended Kalman Filter (EKF) and for the feature detection system based on the SIFT (Scale Invariant Feature Transform). The complexity to implement this work can be considered high, as it involves a significant number of arithmetic and trigonometric operations in oating and fixed-point format, an intensive image processing for feature detection and stability checking, and the development of an image acquisition system from four CMOS cameras in real time. In addition, communication interfaces were created to integrate software and hardware embedded on FPGA and to control the mobile robot base and to read its sensors. Finally, besides the implementation details and the results, this work also presents basic concepts about mapping and state-of-the-art algorithms for SLAM with monocular and stereo vision.

Cette thèse s’inscrit dans le cadre d’un projet visant à développer un robot mobile autonome capable de réaliser des tâches spécifiques dans une zone préalablement définie par l’utilisateur. Afin de faciliter la mise en œuvre du système, des balises radiofréquences fournissant des mesures de distance par rapport au robot sont disposées au préalable autour du terrain. Le déploiement du robot s’effectue en deux phases, une première d’apprentissage du terrain, puis une seconde, où le robot effectue ses tâches de façon autonome. Ces deux étapes nécessitent de résoudre les problèmes de localisation et de localisation et cartographie simultanées pour lesquels différentes solutions sont proposées et testées en simulation et sur des jeux de données réelles. De plus, afin de faciliter l’installation et d’améliorer les performances du système, un algorithme de placement des balises est présenté puis testé en simulation afin de valider notamment l’amélioration des performances de localisation. / This thesis is part of a project which aims at developing an autonomous mobile robot able to perform specific tasks in a preset area. To ease the setup of the system, radio-frequency beacons providing range measurements with respect to the robot are set up beforehand on the borders of the robot’s workspace. The system deployment consists in two steps, one for learning the environment, then a second, where the robot executes its tasks autonomously. These two steps require to solve the localization and simultaneous localization and mapping problems for which several solutions are proposed and tested in simulation and on real datasets. Moreover, to ease the setup and improve the system performances, a beacon placement algorithm is presented and tested in simulation in order to validate in particular the improvement of the localization performances.

Pour assurer la navigation autonome d’un robot mobile, les traitements effectués pour sa localisation doivent être faits en ligne et doivent garantir une précision suffisante pour permettre au robot d’effectuer des tâches de haut niveau pour la navigation et l’évitement d’obstacles. Les auteurs de travaux basés sur le SLAM visuel (Simultaneous Localization And Mapping) tentent depuis quelques années de garantir le meilleur compromis rapidité/précision. La majorité des solutions SLAM visuel existantes sont basées sur une représentation éparse de l’environnement. En suivant des primitives visuelles sur plusieurs images, il est possible d’estimer la position 3D de ces primitives ainsi que les poses de la caméra. La communauté du SLAM visuel a concentré ses efforts sur l’augmentation du nombre de primitives visuelles suivies et sur l’ajustement de la carte 3D, afin d’améliorer l’estimation de la trajectoire de la caméra et les positions 3D des primitives. Cependant, la localisation par SLAM visuel présente souvent des dérives dues au cumul d’erreurs, et dans le cas du SLAM visuel monoculaire, la position de la caméra n’est connue qu’à un facteur d’échelle près. Ce dernier peut être fixé initialement mais dérive au cours du temps. Pour faire face à ces limitations, nous avons centré nos travaux de thèse sur la problématique suivante : intégrer des informations supplémentaires dans un algorithme de SLAM visuel monoculaire afin de mieux contraindre la trajectoire de la caméra et la reconstruction 3D. Ces contraintes ne doivent pas détériorer les performances calculatoires de l’algorithme initial et leur absence ne doit pas mettre l’algorithme en échec. C’est pour cela que nous avons choisi d’intégrer l’information de profondeur fournie par un capteur 3D (e.g. Microsoft Kinect) et des informations géométriques sur la structure de la scène. La première contribution de cette thèse est de modifier l’algorithme SLAM visuel monoculaire proposé par Mouragnon et al. (2006b) pour prendre en compte la mesure de profondeur fournie par un capteur 3D, en proposant particulièrement un ajustement de faisceaux qui combine, d’une manière simple, des informations visuelles et des informations de profondeur. La deuxième contribution est de proposer une nouvelle fonction de coût du même ajustement de faisceaux qui intègre, en plus des contraintes sur les profondeurs des points, des contraintes géométriques d’appartenance aux plans de la scène. Les solutions proposées ont été validées sur des séquences de synthèse et sur des séquences réelles, représentant des environnements variés. Ces solutions ont été comparées aux récentes méthodes de l’état de l’art. Les résultats obtenus montrent que les différentes contraintes développées permettent d’améliorer significativement la précision de la localisation du SLAM. De plus les solutions proposées sont faciles à déployer et peu couteuses en temps de calcul. / To guarantee autonomous and safely navigation for a mobile robot, the processing achieved for its localization must be fast and accurate enough to enable the robot to perform high-level tasks for navigation and obstacle avoidance. The authors of Simultaneous Localization And Mapping (SLAM) based works, are trying since year, to ensure the speed/accuracy trade-off. Most existing works in the field of monocular (SLAM) has largely centered around sparse feature-based representations of the environment. By tracking salient image points across many frames of video, both the positions of the features and the motion of the camera can be inferred live. Within the visual SLAM community, there has been a focus on both increasing the number of features that can be tracked across an image and efficiently managing and adjusting this map of features in order to improve camera trajectory and feature location accuracy. However, visual SLAM suffers from some limitations. Indeed, with a single camera and without any assumptions or prior knowledge about the camera environment, rotation can be retrieved, but the translation is up to scale. Furthermore, visual monocular SLAM is an incremental process prone to small drifts in both pose measurement and scale, which when integrated over time, become increasingly significant over large distances. To cope with these limitations, we have centered our work around the following issues : integrate additional information into an existing monocular visual SLAM system, in order to constrain the camera localization and the mapping points. Provided that the high speed of the initial SLAM process is kept and the lack of these added constraints should not give rise to the failure of the process. For these last reasons, we have chosen to integrate the depth information provided by a 3D sensor (e.g. Microsoft Kinect) and geometric information about scene structure. The primary contribution of this work consists of modifying the SLAM algorithm proposed by Mouragnon et al. (2006b) to take into account the depth measurement provided by a 3D sensor. This consists of several rather straightforward changes, but also on a way to combine the depth and visual data in the bundle adjustment process. The second contribution is to propose a solution that uses, in addition to the depth and visual data, the constraints lying on points belonging to the plans of the scene. The proposed solutions have been validated on a synthetic sequences as well as on a real sequences, which depict various environments. These solutions have been compared to the state of art methods. The performances obtained with the previous solutions demonstrate that the additional constraints developed, improves significantly the accuracy and the robustness of the SLAM localization. Furthermore, these solutions are easy to roll out and not much time consuming.

La localisation et cartographie simultanées connue, communément, sous le nom de SLAM (Simultaneous Localization And Mapping) est un processus qui permet à un robot explorant un environnement inconnu de reconstruire une carte de celui-ci tout en se localisant, en même temps, sur cette carte. Dans ce travail de thèse, nous nous intéressons au SLAM par optimisation de graphe. Celui-ci utilise un graphe pour représenter et résoudre le problème de SLAM. Une optimisation de graphe consiste à trouver une configuration de graphe (trajectoire et carte) qui correspond le mieux aux contraintes introduites par les mesures capteurs. L'optimisation de graphe présente une forte complexité algorithmique et requiert des ressources de calcul et de mémoire importantes, particulièrement si l'on veut explorer de larges zones. Cela limite l'utilisation de cette méthode dans des systèmes embarqués temps-réel. Les travaux de cette thèse contribuent à l'atténuation de la complexité de calcul du SLAM par optimisation de graphe. Notre approche s’appuie sur deux axes complémentaires : la représentation mémoire des données et l’implantation sur architectures hétérogènes embarquées. Dans le premier axe, nous proposons une structure de données incrémentale pour représenter puis optimiser efficacement le graphe. Dans le second axe, nous explorons l'utilisation des architectures hétérogènes récentes pour accélérer le SLAM par optimisation de graphe. Nous proposons, donc, un modèle d’implantation adéquat aux applications embarquées en mettant en évidence les avantages et les inconvénients des architectures évaluées, à savoir SoCs basés GPU et FPGA. / Simultaneous Localization And Mapping is the process that allows a robot to build a map of an unknown environment while at the same time it determines the robot position on this map.In this work, we are interested in graph-based SLAM method. This method uses a graph to represent and solve the SLAM problem. A graph optimization consists in finding a graph configuration (trajectory and map) that better matches the constraints introduced by the sensors measurements. Graph optimization is characterized by a high computational complexity that requires high computational and memory resources, particularly to explore large areas. This limits the use of graph-based SLAM in real-time embedded systems. This thesis contributes to the reduction of the graph-based computational complexity. Our approach is based on two complementary axes: data representation in memory and implementation on embedded heterogeneous architectures. In the first axis, we propose an incremental data structure to efficiently represent and then optimize the graph. In the second axis, we explore the use of the recent heterogeneous architectures to speed up graph-based SLAM. We propose an efficient implementation model for embedded applications. We highlight the advantages and disadvantages of the evaluated architectures, namely GPU-based and FPGA-based System-On-Chips.

The focus of this thesis is to theoretically design, describe, implement and verify thefunctionality of the selected concept for race track mapping. The theoretical part ofthe thesis describes the ORB-SLAM2 algorithm for vehicle localization. It then furtherdescribes the format of the map - occupancy grid and the method of its creation. Suchmap should be in a suitable format for use by other trajectory planning systems. Severalcameras, as well as computer units, are described in this part, and based on parametersand tests, the most suitable ones are selected. The thesis also proposes the architectureof the mapping system, it describes the individual units that make up the system, aswell as what is exchanged between the units, and in what format the system output issent. The individual parts of the system are first tested separately and subsequently thesystem is tested as a whole. Finally, the achieved results are evaluated as well as thepossibilities for further expansion.