Pose (position and orientation) tracking in room-scaled environments is an enabling technique for many applications. Today, virtual reality (vr) and augmented reality (ar) are two examples of such applications, receiving high interest both from the public and the research community. Accurate pose tracking of the vr or ar equipment, often a camera or a headset, or of different body parts is crucial to trick the human brain and make the virtual experience realistic. Pose tracking in room-scaled environments is also needed for reference tracking and metrology. This thesis focuses on an application to metrology. In this application, photometric models of a photo studio are needed to perform realistic scene reconstruction and image synthesis. Pose tracking of a dedicated sensor enables creation of these photometric models. The demands on the tracking system used in this application is high. It must be able to provide sub-centimeter and sub-degree accuracy and at same time be easy to move and install in new photo studios. The focus of this thesis is to investigate and develop methods for a pose tracking system that satisfies the requirements of the intended metrology application. The Bayesian filtering framework is suggested because of its firm theoretical foundation in informatics and because it enables straightforward fusion of measurements from several sensors. Sensor fusion is in this thesis seen as a way to exploit complementary characteristics of different sensors to increase tracking accuracy and robustness. Four different types of measurements are considered; inertialmeasurements, images from a camera, range (time-of-flight) measurements from ultra wide band (uwb) radio signals, and range and velocity measurements from echoes of transmitted acoustic signals. A simulation study and a study of the Cramér-Rao lower filtering bound (crlb) show that an inertial-camera system has the potential to reach the required tracking accuracy. It is however assumed that known fiducial markers, that can be detected and recognized in images, are deployed in the environment. The study shows that many markers are required. This makes the solution more of a stationary solution and the mobility requirement is not fulfilled. A simultaneous localization and mapping (slam) solution, where naturally occurring features are used instead of known markers, are suggested solve this problem. Evaluation using real data shows that the provided inertial-camera slam filter suffers from drift but that support from uwb range measurements eliminates this drift. The slam solution is then only dependent on knowing the position of very few stationary uwb transmitters compared to a large number of known fiducial markers. As a last step, to increase the accuracy of the slam filter, it is investigated if and how range measurements can be complemented with velocity measurement obtained as a result of the Doppler effect. Especially, focus is put on analyzing the correlation between the range and velocity measurements and the implications this correlation has for filtering. The investigation is done in a theoretical study of reflected known signals (compare with radar and sonar) where the crlb is used as an analyzing tool. The theory is validated on real data from acoustic echoes in an indoor environment.

Un des défis majeurs pour le déploiement de systèmes autonomes dans des environnements variés, non structurés et à proximité de l'homme, est d'établir la confiance entre ces systèmes et leurs utilisateurs. En effet, les fautes internes du système, les incertitudes sur la perception, ou encore les situations non prévues, sont des menaces importantes qui pèsent sur cette confiance. Nos travaux se concentrent sur les robots autonomes qui font partis des systèmes autonomes. La validation du logiciel de navigation embarqué dans les robots est généralement centrée sur des tests sur le terrain, qui sont coûteux et potentiellement risqués pour le robot lui-même ou son environnement. De plus, il n'est possible de tester le système que dans un sous-ensemble restreint de situations. Une approche alternative consiste à effectuer des tests basés sur la simulation en immergeant le logiciel dans des mondes virtuels. L'objectif de cette thèse est d'étudier les possibilités et les limites qu'offre le test en simulation du logiciel embarqué dans les systèmes autonomes. Nos travaux traitent particulièrement du test en simulation de la couche de navigation de systèmes autonomes mobiles. Le premier chapitre présente les contextes de la sûreté de fonctionnement, des systèmes autonomes et de leur test, de la simulation et de la génération procédurale de mondes. Les problématiques liées au test des systèmes autonomes en simulation, telles que la définition et la génération des entrées ainsi que l'expression de l'oracle, sont identifiées et discutées. La génération procédurale de mondes utilisée dans les jeux vidéos est retenue comme piste pour répondre au problème de la générations des entrées de test (mondes et missions). Une première contribution est proposée dans le 2ème chapitre qui s'appuie sur la définition et l'implémentation d'une première plateforme expérimentale de test en simulation avec un robot mobile. Le logiciel de navigation utilisé est intégré dans le framework Genom et testé avec le simulateur MORSE. À travers cette expérimentation, des premières conclusions sont établies sur la pertinence de la génération procédurale de mondes, et sur l'oracle à considérer. Des mesures comme la tortuosité ou l'indéterminisme de la navigation sont définies. Ce premier travail mène également à proposer une approche permettant de définir des niveaux de difficulté de mondes. L'objectif du 3ème chapitre est d'identifier si les fautes connues et corrigées dans un logiciel de navigation auraient pu être détectées via la simulation. Près de 10 ans de commits du logiciel de navigation (dont le module P3D qui est une version académique d'un planificateur de trajectoire utilisé par la NASA) ont été ainsi analysés. Chaque faute relevée est étudiée pour déterminer si elle serait activable en simulation, et l'oracle nécessaire pour la détecter. De nombreuses recommandations sont extraites de cette étude, notamment sur les propriétés de l'oracle à mettre en place pour ce genre de système. Dans le quatrième chapitre, les enseignements tirés des deux chapitres précédents sont mis en œuvre dans une étude de cas d'un robot industriel. Le système considéré, fourni par notre partenaire industriel Naïo est celui du robot agricole bineur Oz. Les conclusions des chapitres précédents concernant la génération de monde et les oracles nécessaires sont validées par une campagne de test intensifs en simulation. / One of the major challenges for the deployment of autonomous systems in diverse, unstructured and human shared environments, is the trust that can be placed in those systems. Indeed, internal faults in those systems, uncertainties on the perception, or even unforeseen situations, threat this confidence. Our work focus in autonomous robots, which are part of autonomous systems. The validation of the navigation software embedded in robots typically involves test campaigns in the field, which are expensive and potentially risky for the robot itself or its environment. These tests are able to test the system only in a small subset of situations. An alternative is to perform simulation-based testing, by immersing the software in virtual worlds. The aim of this thesis is to study the possibilities and limits offered by simulation-based testing of embedded software in autonomous systems. Our work deals particularly with simulation-based testing of the navigation layer of autonomous mobile robots. The first chapter introduce the contexts of dependability, autonomous systems and their testing, simulation and procedural generation of worlds. We identify and discuss the issues related to autonomous systems simulation-based testing, such as the definition and generation of inputs as well as the oracle. The procedural generation of worlds used in video games is retained as a way to answer the problem of the generation of test inputs (worlds and missions). A first contribution is presented in the second chapter, which is based on the definition and implementation of a first experimental simulation-based testing framework with a mobile robot. The navigation software used is integrated into the Genom framework and tested with the MORSE simulator. Through this experiment, first conclusions are drawn on the relevance of the procedural generation of worlds, and on the oracle to be considered. Measures such as tortuousness or indeterminism of navigation are defined. This first work also leads to propose an approach to define levels of difficulty of worlds. The purpose of the third chapter is to identify whether faults known and corrected in a academic navigation software could have been detected through simulation-based testing. Nearly 10 years of commits of the navigation software (including the P3D module which is an academic version of a trajectory planner used by NASA) were thus analyzed. Each fault detected is studied to determine the oracle necessary to detect it whether it could be activated in simulation. Many recommendations are extracted from this study, especially on the properties of the oracle to set up for this type of system. In the fourth chapter, lessons learned from the previous two chapters are implemented for the case of an industrial robot. The considered system, provided by our industrial partner Naïo is the agricultural robot Oz. The conclusions of the preceding chapters regarding the world generation and the oracles are validated by an intensive test campaign in simulation.

The goal of this thesis is to use fringe-pattern phase analysis to calibrate the distortion of a camera lens. The benefit of using this method is that the distortion can be calculated using data from each individual pixel and the methodology does not need any model. The phase used to calibrate the images is calculated in two different ways, either utilizing the monogenic signal or through fringe-pattern phase analysis. The calibration approaches were also validated through different methods. Primarily by utilizing the Hough transform and calibrating simulated distortion. The thesis also introduces a validation approach utilizing the phase orientation calculated through the monogenic signal. The thesis also implements different approaches such as flat field correction to limit the impact of the image sensor noise to mitigate the phase noise. It is also investigated which fringe-pattern frequencies are best suited for calibration through comparative analysis. The comparative analysis identified problems with too high and low frequencies of the fringe-patterns when calibrating using fringe-pattern phase analysis.

This thesis explores the application of Contrastive Language-Image Pre-Training (CLIP), a vision-language model, in an automated video surveillance system for anomaly detection. The ability of CLIP to perform zero-shot learning, coupled with its robustness against minor image alterations due to its lack of reliance on pixel-level image analysis, makes it a suitable candidate for this application. The study investigates the performance of CLIP in tandem with various anomaly detection algorithms within a visual surveillance system. A custom dataset was created for video anomaly detection, encompassing two distinct views and two varying levels of anomaly difficulty. One view offers a more zoomed-in perspective, while the other provides a wider perspective. This was conducted to evaluate the capacity of CLIP to manage objects that occupy either a larger or smaller portion of the entire scene. Several different anomaly detection methods were tested with varying levels of supervision, including unsupervised, one-class classification, and weakly- supervised algorithms, which were compared against each other. To create better separation between the CLIP embeddings, a metric learning model was trained and then used to transform the CLIP embeddings to a new embedding space. The study found that CLIP performs effectively when anomalies take up a larger part of the image, such as in the zoomed-in view where some of the One- Class-Classification (OCC) and weakly supervised methods demonstrated superior performance. When anomalies take up a significantly smaller part of the image in the wider view, CLIP has difficulty distinguishing anomalies from normal scenes even using the transformed CLIP embeddings. For the wider view the results showed on better performance for the OCC and weakly supervised methods.

Liability for damages caused by an autonomous vehicle Abstract For several years now, autonomous vehicles have been one of the most interesting topics associated with the upcoming Industry 4.0 and the spread of artificial intelligence in society. This thesis therefore deals with the topic of autonomous vehicles as a subset of autonomous systems, specifically in connection with civil-law liability. It is probable that in the future there will be a massive expansion of autonomous means of transport in society, and it cannot be ruled out that a situation will arise where their setting will result in damage. The aim of this thesis is to present the models of liability, the use of which is discussed by jurisprudence in connection with autonomous vehicles, and to assess the possibility of using the current institutes of Civil Code, to find a suitable future solution. For this purpose, the thesis first deals with the concept of autonomous vehicles according to the Civil Code and other legislation and by using existing definitions of artificial intelligence deals with specific features of autonomous systems, as well as the status of specific persons associated with the operation of autonomous vehicles in light of current legislation, their typology and conditions of their operation itself. The second part of this...

Grasping novel objects is challenging because of incomplete object data and because of uncertainties inherent in real world applications. To robustly perform grasps on previously unseen objects, feedback from touch is essential. In our research, we study how information from touch sensors can be used to improve grasping novel objects. Since it is not trivial to extract relevant object properties and deduce appropriate actions from touch sensing, we employ machine learning techniques to learn suitable behaviors. We have shown that grasp stability estimation based on touch can be improved by including an approximate notion of object shape. Further we have devised a method to guide local grasp adaptations based on our stability estimation method. Grasp corrections are found by simulating tactile data for grasps in the vicinity of the current grasp. We present several experiments to demonstrate the applicability of our methods. The thesis is concluded by discussing our results and suggesting potential topics for further research. / Att greppa nya föremål är utmanande, både eftersom roboten inte har fullständig information om objekten och på grund av den inneboende osäkerheten i verkliga tillämpningar. Återkoppling från känselsensorer är viktigt för att kunna greppa föremål som inte påträffats tidigare. I vår forskning så studerar vi hur information från känselsensorer kan användas för att förbättra greppandet av nya föremål. Eftersom det är svårt att extrahera relevanta egenskaper om föremål och härleda lämpliga åtgärder, baserat på känselsensorer, så har vi använt maskininlärning för att lära roboten lämpliga beteenden. Vi har visat att uppskattningar av stabiliteten av ett grepp baserat på känselsensorer kan förbättras genom att även använda en grov approximation av föremålets form. Vi har även konstruerat en metod som vägleder lokala justeringar av grepp, baserat på vår metod som uppskattar stabiliteten av ett grepp. Dess justeringar hittas genom att simulera känselsensordata för grepp i närheten av det nuvarande greppet. Vi presenterar flera experiment som demonstrerar tillämpbarheten av våra metoder. Avhandlingen avslutas med en diskussion om våra resultat och förslag på möjliga ämnen för fortsatt forskning. / <p>QC 20170510</p>

As online retail continues to grow, large amounts of valuable data, such as transaction and search history, and, specifically for fashion retail, similarly structured images of clothing, is generated. By using unsupervised learning, it is possible to tap into this almost unlimited supply of data. This thesis set out to determine to what extent generative models – in particular, deep convolutional variational autoencoders – can be used to automatically extract representative features from images of clothing in a completely unsupervised manner. In reviewing variations of the autoencoder, both in terms of reconstruction quality and the ability to generate new realistic samples, results suggest that there exists an optimal size of the latent vector in relation to the image data complexity. Furthermore, by weighting the latent loss and generation loss in the loss function, it was possible to disentangle the learned features such that each feature captured a unique defining characteristic of clothing items (here t-shirts and tops). / I takt med att E-handeln fortsätter att växa och kunderna i ökad utsträckning rör sig online, genereras stora mängder värdefull data, exempelvis transaktions- och sökhistorik, och specifikt för klädeshandeln, välstrukturerade bilder av kläder. Genom att använda oövervakad maskininlärning (unsupervised machine learning) är det möjligt att utnyttja denna, nästan obegränsade mängd data. Detta arbete syftar till att utreda i vilken utsträckning generativa modeller, särskilt djupa självkodande neurala faltningsnätverk (deep convolutional variational autoencoders), kan användas för att automatiskt extrahera definierande drag från bilder av kläder. Genom att granska olika varianter av självkodaren framträder en optimal relation mellan storleken på den latenta vektorn och komplexiteten på den bilddata som nätverket tränades på. Vidare noterades att dragen kan fördeladas unikt på variablerna, i detta fall t-shirts och toppar, genom att vikta den latenta förlustfunktionen.

Unmanned ground vehicles (UGVs) can be used in the military to mitigate risks taken by soldiers as well as to provide solutions to physically demanding, dull or dangerous tasks. While there are benefits of using UGVs, there are also needs and limitations that come with them. This thesis explores what the end user, a light infantry battalion in the Swedish Armed Forces, wants in terms of functions for a UGV designed for military oper- ations in urban terrain. This was done through a explorative case study with focus groups, where soldiers and officers from the 31st Ranger Battalion got to use two different UGV prototypes to complete tasks. This was followed by semi structured group discussions, where needs, limitations and requirements were explored. The collected data was then analysed by a thematic analysis approach. The results from the thematic analysis found several reoccurring opinions regarding requirements from the focus groups. The requirements were grouped into four categories; (1) Speed, (2) Use cases, (3) Image generating sensors, and (4) Autonomous functions. In conclusion, this thesis identified a total of 13 requirements within the four categories. To summarise, these requirements imply that a UGV meant for military operations in urban terrain must be able to keep up with sprinting soldiers, provide visual cover, be able to interact with nearby objects, have several high quality sensors and robust autonomous functions, allowing the soldiers to focus on other things than controlling the UGV.

Many digital logic applications can take advantage of the reconfiguration capability of Field Programmable Gate Arrays (FPGAs) to dynamically patch design flaws, recover from faults, or time-multiplex between functions. Partial reconfiguration is the process by which a user modifies one or more modules residing on the FPGA device independently of the others. Partial Reconfiguration reduces the granularity of reconfiguration to be a set of columns or rectangular region of the device. Decreasing the granularity of reconfiguration results in reduced configuration filesizes and, thus, reduced configuration times. When compared to one bitstream of a non-partial reconfiguration implementation, smaller modules resulting in smaller bitstream filesizes allow an FPGA to implement many more hardware configurations with greater speed under similar storage requirements. To realize the benefits of partial reconfiguration in a wider range of applications, this thesis begins with a survey of FPGA fault-handling methods, which are compared using performance-based metrics. Performance analysis of the Genetic Algorithm (GA) Offline Recovery method is investigated and candidate solutions provided by the GA are partitioned by age to improve its efficiency. Parameters of this aging technique are optimized to increase the occurrence rate of complete repairs. Continuing the discussion of partial reconfiguration, the thesis develops a case-study application that implements one partial reconfiguration module to demonstrate the functionality and benefits of time multiplexing and reveal the improved efficiencies of the latest large-capacity FPGA architectures. The number of active partial reconfiguration modules implemented on a single FPGA device is increased from one to eight to implement a dynamic video-processing architecture for Discrete Cosine Transform and Motion Estimation functions to demonstrate a 55-fold reduction in bitstream storage requirements thus improving partial reconfiguration capability.

No description available.