When reconstructing the Earth in 3D, the imagery can come from various mediums, including satellites, planes, and drones. One significant benefit of utilizing drones in combination with a Visual Simultaneous Localization and Mapping (V-SLAM) system is that specific areas of the world can be accurately mapped in real-time at a low cost. Drones can essentially be equipped with any camera sensor, but most commercially available drones use a monocular rolling shutter camera sensor. Therefore, on behalf of Maxar Technologies, multiple monocular V-SLAM systems were studied during this thesis, and ORB-SLAM3 and LDSO were determined to be evaluated further. In order to provide an accurate and reproducible result, the methods were benchmarked on the public datasets EuRoC MAV and TUM monoVO, which includes drone imagery and outdoor sequences, respectively. A third dataset was collected with a DJI Mavic 2 Enterprise Dual drone to evaluate how the methods would perform with a consumer-friendly drone. The datasets were used to evaluate the two V-SLAM systems regarding the generated 3D map (point cloud) and estimated camera trajectory. The results showed that ORB-SLAM3 is less impacted by the artifacts caused by a rolling shutter camera sensor than LDSO. However, ORB-SLAM3 generates a sparse point cloud where depth perception can be challenging since it abstracts the images using feature descriptors. In comparison, LDSO produces a semi-dense 3D map where each point includes the pixel intensity, which improves the depth perception. Furthermore, LDSO is more suitable for dark environments and low-texture surfaces. Depending on the use case, either method can be used as long as the required prerequisites are provided. In conclusion, monocular V-SLAM systems are highly dependent on the type of sensor being used. The differences in the accuracy and robustness of the systems using a global shutter and a rolling shutter are significant, as the geometric artifacts caused by a rolling shutter are devastating for a pure visual pipeline. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>

Doctor of Philosophy / Department of Electrical and Computer Engineering / Balasubramaniam Natarajan / Dale E. Schinstock / In this dissertation, two applications related to vision-based localization and mapping are considered: (1) improving navigation system based satellite location estimates by using on-board camera images, and (2) deriving position information from video stream and using it to aid an auto-pilot of an unmanned aerial vehicle (UAV). In the first part of this dissertation, a method for analyzing a minimization process called bundle adjustment (BA) used in stereo imagery based 3D terrain reconstruction to refine estimates of camera poses (positions and orientations) is presented. In particular, imagery obtained with pushbroom cameras is of interest. This work proposes a method to identify cases in which BA does not work as intended, i.e., the cases in which the pose estimates returned by the BA are not more accurate than estimates provided by a satellite navigation systems due to the existence of degrees of freedom (DOF) in BA. Use of inaccurate pose estimates causes warping and scaling effects in the reconstructed terrain and prevents the terrain from being used in scientific analysis. Main contributions of this part of work include: 1) formulation of a method for detecting DOF in the BA; and 2) identifying that two camera geometries commonly used to obtain stereo imagery have DOF. Also, this part presents results demonstrating that avoidance of the DOF can give significant accuracy gains in aerial imagery. The second part of this dissertation proposes a vision based system for UAV navigation. This is a monocular vision based simultaneous localization and mapping (SLAM) system, which measures the position and orientation of the camera and builds a map of the environment using a video-stream from a single camera. This is different from common SLAM solutions that use sensors that measure depth, like LIDAR, stereoscopic cameras or depth cameras. The SLAM solution was built by significantly modifying and extending a recent open-source SLAM solution that is fundamentally different from a traditional approach to solving SLAM problem. The modifications made are those needed to provide the position measurements necessary for the navigation solution on a UAV while simultaneously building the map, all while maintaining control of the UAV. The main contributions of this part include: 1) extension of the map building algorithm to enable it to be used realistically while controlling a UAV and simultaneously building the map; 2) improved performance of the SLAM algorithm for lower camera frame rates; and 3) the first known demonstration of a monocular SLAM algorithm successfully controlling a UAV while simultaneously building the map. This work demonstrates that a fully autonomous UAV that uses monocular vision for navigation is feasible, and can be effective in Global Positioning System denied environments.

Building maps within an unknown environment while keeping track of the current position is a major step to accomplish safe and autonomous robot navigation. Within the last 20 years, Simultaneous Localization And Mapping (SLAM) became a topic of great interest in robotics. The basic idea of this technique is to combine proprioceptive robot motion information with external environmental information to minimize global positioning errors. Because the robot is moving in its environment, exteroceptive data comes from different points of view and must be expressed in the same coordinate system to be combined. The latter process is called registration. Iterative Closest Point (ICP) is a registration algorithm with very good performances in several 3D model reconstruction applications, and was recently applied to SLAM. However, SLAM has specific needs in terms of real-time and robustness comparatively to 3D model reconstructions, leaving room for specialized robotic mapping optimizations in relation to robot mapping. After reviewing existing SLAM approaches, this thesis introduces a new registration variant called Kd-ICP. This referencing technique iteratively decreases the error between misaligned point clouds without extracting specific environmental features. Results demonstrate that the new rejection technique used to achieve mapping registration is more robust to large initial positioning errors. Experiments with simulated and real environments suggest that Kd-ICP is more robust compared to other ICP variants. Moreover, the Kd-ICP is fast enough for real-time applications and is able to deal with sensor occlusions and partially overlapping maps. Realizing fast and robust local map registrations opens the door to new opportunities in SLAM. It becomes feasible to minimize the cumulation of robot positioning errors, to fuse local environmental information, to reduce memory usage when the robot is revisiting the same location. It is also possible to evaluate network constrains needed to minimize global mapping errors.

Many mobility impaired users are unable to operate a powered wheelchair safely, without causing harm to themselves, others, and the environment. Smart wheelchairs that assist or replace user control have been developed to cater for these users, utilising systems and algorithms from autonomous robots. Despite a sustained period of research and development of robotic wheelchairs, there are very few available commercially. This thesis describes work towards developing a navigation system that is aimed at being retro-fitted to powered wheelchairs. The navigation system developed takes a systems engineering approach, integrating many existing open-source software projects to deliver a system that would otherwise not be possible in the time frame of a master's thesis. The navigation system introduced in this thesis is aimed at operating in an unstructured indoor environment, and requires no a priori information about the environment. The key components in the system are: obstacle avoidance, map building, localisation, path planning, and autonomously travelling towards a goal. The test electric wheelchair was instrumented with the following: a laptop, a laser scanner, wheel encoders, camera, and a variety of user input methods. The user interfaces that have been implemented and tested include a touch screen friendly graphical user interface, keyboard and joystick.

This thesis aims to provide a close analysis of poetry slam in the United Kingdom and United States, using the tools of ethnography and discourse analysis to produce an in-depth account, which is sensitive to the discursively constructed, situated meanings of slam participants. The aim is to explore how slam is understood by its participants, producing a partial ethnography, rather than a definitive history, defence or critique of slam. The thesis is based predominantly on research conducted in four key sites (Bristol and London in the U.K. and Chicago and New York in the U.S.), and considers how slam has been reconstructed in different geographical and social contexts. In addition, this study seeks to highlight issues around: the ways in which artists understand art worlds and their positions within them; the multiple and complex power relations with which art world participants engage; the transient, enduring and virtual communities which art world participants form; the local, translocal and transnational networks which connect these communities and individuals; and the interactions between new/avant-garde and established/dominant art worlds. It is hoped that this analysis will enrich substantially the existing meagre body of research into poetry slam, providing valuable theoretical contributions to the study of art worlds and the social construction of self and relationships. Beyond this, the thesis aims to elucidate a social scientific paradigm which links micro level analyses with macro level social structures and processes, by allying work from multiple theoretical perspectives including those of interactionism, Antonio Gramsci, Pierre Bourdieu and discourse analysis. This paradigm is mobilised to illuminate how slam participants actively construct their identities and negotiate the complex power relations which structure their everyday interactions. In line with the poetic focus of this research, each analytic chapter of this thesis concludes with a haiku. I begin with this thought: Power relations/Are complex navigations/Through interaction.

Il y a présentement de la demande dans plusieurs milieux cherchant à utiliser des robots afin d'accomplir des tâches complexes, par exemple l'industrie de la construction désire des travailleurs pouvant travailler 24/7 ou encore effectuer des operation de sauvetage dans des zones compromises et dangereuses pour l'humain. Dans ces situations, il devient très important de pouvoir transporter des charges dans des environnements encombrés. Bien que ces dernières années il y a eu quelques études destinées à la navigation de robots dans ce type d'environnements, seulement quelques-unes d'entre elles ont abordé le problème de robots pouvant naviguer en déplaçant un objet volumineux ou lourd. Ceci est particulièrement utile pour transporter des charges ayant de poids et de formes variables, sans avoir à modifier physiquement le robot. Un robot humanoïde est une des plateformes disponibles afin d'effectuer efficacement ce type de transport. Celui-ci a, entre autres, l'avantage d'avoir des bras et ils peuvent donc les utiliser afin de manipuler précisément les objets à transporter. Dans ce mémoire de maîtrise, deux différentes techniques sont présentées. Dans la première partie, nous présentons un système inspiré par l'utilisation répandue de chariots de fortune par les humains. Celle-ci répond au problème d'un robot humanoïde naviguant dans un environnement encombré tout en déplaçant une charge lourde qui se trouve sur un chariot de fortune. Nous présentons un système de navigation complet, de la construction incrémentale d'une carte de l'environnement et du calcul des trajectoires sans collision à la commande pour exécuter ces trajectoires. Les principaux points présentés sont : 1) le contrôle de tout le corps permettant au robot humanoïde d'utiliser ses mains et ses bras pour contrôler les mouvements du système à chariot (par exemple, lors de virages serrés) ; 2) une approche sans capteur pour automatiquement sélectionner le jeu approprié de primitives en fonction du poids de la charge ; 3) un algorithme de planification de mouvement qui génère une trajectoire sans collisions en utilisant le jeu de primitive approprié et la carte construite de l'environnement ; 4) une technique de filtrage efficace permettant d'ignorer le chariot et le poids situés dans le champ de vue du robot tout en améliorant les performances générales des algorithmes de SLAM (Simultaneous Localization and Mapping) défini ; et 5) un processus continu et cohérent d'odométrie formés en fusionnant les informations visuelles et celles de l'odométrie du robot. Finalement, nous présentons des expériences menées sur un robot Nao, équipé d'un capteur RGB-D monté sur sa tête, poussant un chariot avec différentes masses. Nos expériences montrent que la charge utile peut être significativement augmentée sans changer physiquement le robot, et donc qu'il est possible d'augmenter la capacité du robot humanoïde dans des situations réelles. Dans la seconde partie, nous abordons le problème de faire naviguer deux robots humanoïdes dans un environnement encombré tout en transportant un très grand objet qui ne peut tout simplement pas être déplacé par un seul robot. Dans cette partie, plusieurs algorithmes et concepts présentés dans la partie précédente sont réutilisés et modifiés afin de convenir à un système comportant deux robot humanoides. Entre autres, nous avons un algorithme de planification de mouvement multi-robots utilisant un espace d'états à faible dimension afin de trouver une trajectoire sans obstacle en utilisant la carte construite de l'environnement, ainsi qu'un contrôle en temps réel efficace de tout le corps pour contrôler les mouvements du système robot-objet-robot en boucle fermée. Aussi, plusieurs systèmes ont été ajoutés, tels que la synchronisation utilisant le décalage relatif des robots, la projection des robots sur la base de leur position des mains ainsi que l'erreur de rétroaction visuelle calculée à partir de la caméra frontale du robot. Encore une fois, nous présentons des expériences faites sur des robots Nao équipés de capteurs RGB-D montés sur leurs têtes, se déplaçant avec un objet tout en contournant d'obstacles. Nos expériences montrent qu'un objet de taille non négligeable peut être transporté sans changer physiquement le robot.

Flight and drop tests of the Martin 270 (M270) seaplane were conducted in 1955. Theoretical and empirical pressures were determined by use of Wagner’s theory and also by the Code of Federal Regulations (CFR). The pressure results from the experimental tests on the hull were compared with pressures calculated from Wagner’s theory to determine how well the theory correlated with the measured pressures. The experimental pressure data was also compared with the CFR results to determine how the current industry standard of estimating impact loads compares with actual pressures a seaplane is subjected to. Using the structural design and geometry of the M270 the seaplane hull was modeled in Maestro with a coarse mesh finite element model. The pressures from Wagner’s theory and the CFR were applied to the model of the M270 hull. The structural reactions of the drop test section were compared with the reactions determined from Maestro.

Simultaneous localization and mapping (SLAM) is a challenging and vital problem in robotics. It is important in tasks such as disaster response, deep-sea and cave exploration, in which robots must construct a map of an unknown terrain, and at the same time localize themselves within the map. The issue with single- robot SLAM is the relatively high rate of failure in a realistic application, as well as the time and energy cost. In this work, we propose a new approach to decentralized multi-robot SLAM which uses a robot swarm to map the environment. This system is capable of mapping an environment without human assistance and without the need for any additional infrastructure. We assume that 1) no robot possesses sufficient memory to store the entire map of the environment, 2) the communication range of the robots is limited, and 3)there is no infrastructure present in the environment to assist the robot in communicating with others. To cope with these limitations, the swarm system is designed to work as an independent entity. The swarm can deploy new robots towards the region that is yet to be explored, coordinate the communication between the robots by using itself as the communication network and replace any malfunctioning robots. The proposed method proves to be a reliable and robust exploration algorithm. It is shown to be a self-growing mapping network that is able to coordinate among numerous robots and replace any broken robots hence reducing the chance of system failure.

Simultaneous localization and mapping (SLAM) is a challenging and vital problem in robotics. It is important in tasks such as disaster response, deep-sea and cave exploration, in which robots must construct a map of an unknown terrain, and at the same time localize themselves within the map. The issue with single- robot SLAM is the relatively high rate of failure in a realistic application, as well as the time and energy cost. In this work, we propose a new approach to decentralized multi-robot SLAM which uses a robot swarm to map the environment. This system is capable of mapping an environment without human assistance and without the need for any additional infrastructure. We assume that 1) no robot possesses sufficient memory to store the entire map of the environment, 2) the communication range of the robots is limited, and 3)there is no infrastructure present in the environment to assist the robot in communicating with others. To cope with these limitations, the swarm system is designed to work as an independent entity. The swarm can deploy new robots towards the region that is yet to be explored, coordinate the communication between the robots by using itself as the communication network and replace any malfunctioning robots. The proposed method proves to be a reliable and robust exploration algorithm. It is shown to be a self-growing mapping network that is able to coordinate among numerous robots and replace any broken robots hence reducing the chance of system failure.

Kelp beds support a vast and diverse ecosystem including marine mammals, fish, invertebrates, other algae and epibiota, yet these kelp beds can be highly ephemeral. Mapping the density and distribution of kelp beds, and assessing change over yearly cycles, are important objectives for coastal oceanography. However, nearshore habitat mapping is challenging, affected by dynamic currents, tides, shallow depths, frequent non-uniform obstacles and often turbid water. Noisy and often incomplete sensor data compound a lack of landmarks available for navigation. The intelligent, position-aware holonomic ROV (iROV) SeaBiscuit was designed specifically for this nearshore habitat mapping application and represents a novel synthesis of techniques and innovative solutions to nearshore habitat mapping. The concept of an iROV combines the benefits of autonomous underwater navigation and mapping while maintaining the flexibility and security of remote high-level control and supervision required for operation in hostile, complex underwater environments. An onboard battery provides an energy buffer for high-powered thrust and security of energy supply. Onboard low-level autonomy provides robust autopilot features, including station-keeping or course-holding in a flow, allowing the operator to direct the survey and supervise mapping data in realtime during acquisition. With the aim of providing high-usability maps on a budget feasible for small-scale field research groups, SeaBiscuit fuses the data from an orthogonal arrangement of a forward-facing multibeam sonar and a complementary 360° scanning sonar with a full navigation suite to explore and map the nearshore environment. Sensor fusion, coupled with the holonomic propulsion system, also allows optimal use of the information available from the limited budget sensor suite. Robust and reliable localisation is achieved even with noisy and incomplete sensor data using a relatively basic Inertial Navigation System and sonar-aided SLAM in the absence of an expensive Doppler velocity log or baseline navigation system. Holonomic motion in the horizontal plane and an axisymmetric hull provide the manoeuvrability required to operate in this complex environment, while allowing 3D maps to be generated in-transit. The navigation algorithms were tested mapping a piling dock and the habitat mapping sensors calibrated using an ‘artificial’ kelp bed of manually dimensioned kelp stipes transplanted to a sheltered but open-water real-world environment. Sea trials demonstrated mapping open ocean kelp beds, identifying clusters of stipes, converting this into a useful measure of biomass and generating a density surface across the kelp bed. This research provides field-proven techniques to improve the nearshore habitat mapping capabilities of underwater vehicles. Future work includes the transition to full-scale kelp bed mapping, and further development of the vehicle and sensor fusion algorithms to improve nearshore navigation.