Autonomous lawn mowers have been available to consumers for more than 20 years. During this period, advancements in embedded device computations and sensor performance have led to improvements in the reliability of these robots. Despite recent improvements, the opportunity for further innovation of such systems remains significant. Currently, many autonomous robots rely on electric wires installed underground to delimit the boundaries of the lawn. Such a configuration is simple, but more effective autonomous solutions are available. This thesis focuses on the analysis and related implementation of both localisation and mapping features for autonomous lawn mowers. Heterogeneous sensors and their different configurations are investigated and an Adaptive Extended Kalman Filter is proposed to fuse their measurements. This technique improves the pose estimation of the autonomous lawn mower, which is then exploited by the mapping module. Based on Bayesian’s inference, the mapping module updates the knowledge of the map based on direct interactions with the environment. The final results highlight the importance of precise localisation as the bottleneck for the development of new features. The improved pose estimation enables the employment of a virtual boundary, but it is not accurate enough to precisely map the presence of objects in the environment. Advanced features which could be developed from the proposed configuration are related to deterministic coverage algorithms and the interaction with lawn objects. / Autonoma gräsklippare har varit tillgängliga för konsumenter i mer än 20 år. Under denna period har framsteg inom beräkningar av inbyggda enheter och sensorprestanda lett till förbättringar av tillförlitligheten hos dessa robotar. Trots de senaste förbättringarna är möjligheten till innovation av sådana system fortfarande betydande. Autonoma robotar har fortfarande begränsade funktioner. De förlitar sig på elektriska ledningar installerade under jord för att avgränsa gräsmattans gränser, som de reagerar på utan resonemang. En sådan konfiguration anses nu vara föråldrad och mer effektiva autonoma lösningar finns tillgängliga. Den här avhandlingen fokuserar på att använda för närvarande tillgängliga tekniker för att designa de kärnmoduler som behövs för att förbättra kapaciteten hos dessa system. Analysen och relaterad implementering av både lokaliserings och kartläggningsfunktioner för autonoma gräsklippare presenteras. Heterogena sensorer och deras olika konfigurationer undersöks och ett Adaptive Extended Kalman Filter föreslås för att smälta samman deras mätningar. Denna teknik förbättrar poseuppskattningen av den autonoma gräsklipparen, som sedan utnyttjas av kartläggningsmodulen. Det valda tillvägagångssättet för den senare, baserat på Bayesians slutledning, lyckas uppdatera kunskapen om kartan baserat på direkta interaktioner med omgivningen. De slutliga resultaten belyser vikten av exakt lokalisering som den verkliga flaskhalsen för utvecklingen av nya funktioner. Den förbättrade positionsuppskattningen gör det möjligt att definiera en virtuell gräns. Definitionen inte tillräckligt korrekt för att korrekt kartlägga förekomsten av objekt i miljön Exempel på avancerade funktioner från den föreslagna konfigurationen är implementeringen av deterministiska täckningsalgoritmer och interaktionen med gräsmattaobjekt. / I tosaerba autonomi sono disponibili per i consumatori da oltre 20 anni. Durante questo periodo, i progressi nei calcoli dei dispositivi integrati e nelle prestazioni dei sensori hanno portato a miglioramenti nell’affidabilità di questi robot. Nonostante i recenti miglioramenti, l’opportunità di innovazione di tali sistemi rimane significativa. I robot autonomi hanno ancora funzionalità limitate. Si affidano a fili elettrici installati sottoterra per delimitare i confini del prato, a cui reagiscono senza ragionamento. Tale configurazione è ormai considerata obsoleta e sono disponibili soluzioni autonome più efficaci. Questa tesi si concentra sull’utilizzo delle tecniche attualmente disponibili per progettare i moduli principali necessari per far avanzare le capacità di questi sistemi. Vengono presentate l’analisi e la relativa implementazione delle funzionalità di localizzazione e mappatura per i tosaerba autonomi. Vengono studiati sensori eterogenei e le loro diverse configurazioni e viene proposto un filtro di Kalman adattivo esteso per fondere le loro misurazioni. Questa tecnica migliora la stima della posa del rasaerba autonomo, che viene poi sfruttata dal modulo di mappatura. L’approccio scelto per quest’ultimo, basato sull’inferenza bayesiana, riesce ad aggiornare la conoscenza della mappa basata su interazioni dirette con l’ambiente. I risultati finali evidenziano l’importanza di una localizzazione precisa come vero collo di bottiglia per lo sviluppo di nuove funzionalità. La stima della posa migliorata consente la definizione di un confine virtuale. La definizione non è sufficientemente precisa per mappare correttamente la presenza di oggetti nell’ambiente Esempi di funzionalità avanzate a partire dalla configurazione proposta sono l’implementazione di algoritmi di copertura deterministici e l’interazione con gli oggetti del prato.

The importance of accurate position estimation is becoming more necessary as industries and society increasingly rely on autonomous or wireless devices while the capabilities of existing positioning solutions fail to meet the demanding requirements. This situation has provided opportunities for wireless positioning techniques with the rollout of 5G which has led to many enhancements to the network protocol related to positioning. This report investigates the feasibility of meeting these requirements with the use of existing GNSS positioning solutions and yet-to-be implemented 5G positioning methods. We evaluate the performance using different measurements separately as well as a hybridization between them to examine the optimal result. The report also demonstrates the potential of using only a single BS to achieve accurate positioning, which is not possible with e.g. LTE.  The method in this report is based on partly well-proven theory for positioning together with recent developed concepts for radio network localization. By using an advanced simulator that generates realistic signals and measurements in virtual deployments of base-stations and users, our method can be well evaluated, which makes the results interesting for both academia and industry. The results show good potential for both 5G stand-alone positioning as well as hybrid 5G and GNSS positioning. This report demonstrates that a single BS can locate a UE in line-of-sight of 200 meters within 1.5 meters for 80% of the cases without using any GNSS system.

Reinforcement learning has proven itself very useful in certain areas, such as games. However, the approach has been seen as quite limited. Reinforcement-based learning has for instance not been commonly used for classification tasks as it is receiving feedback on how well it did for an action performed on a specific input. This slows the performance convergence rate as compared to other classification approaches which has the input and the corresponding output to train on. Nevertheless, this thesis aims to investigate whether reinforcement-based learning could successfully be employed on a classification task. Moreover, as sensor fusion is an expanding field which can for instance assist autonomous vehicles in understanding its surroundings, it is also interesting to see how sensor fusion, i.e., fusion between lidar and RGB images, could increase the performance in a classification task. In this thesis, a reinforcement-based learning approach is compared to a semi-classical approach. As an example of a reinforcement learning model, a deep Q-learning network was chosen, and a support vector machine classifier built on top of a deep neural network, was chosen as an example of a semi-classical model. In this work, these frameworks are compared with and without sensor fusion to see whether fusion improves their performance. Experiments show that the evaluated reinforcement-based learning approach underperforms in terms of metrics but mainly due to its slow learning process, in comparison to the semi-classical approach. However, on the other hand using reinforcement-based learning to carry out a classification task could still in some cases be advantageous, as it still performs fairly well in terms of the metrics presented in this work, e.g. F1-score, or for instance imbalanced datasets. As for the impact of sensor fusion, a notable improvement can be seen, e.g. when training the deep Q-learning model for 50 episodes, the F1-score increased with 0.1329; especially, when taking into account that the most of the lidar data used in the fusion is lost since this work projects the 3D lidar data onto the same 2D plane as the RGB images.

Modern day intelligent transportation system (ITS) relies on reliable and accurate estimated traffic parameters. Travel speed, traffic flow, and traffic state classification are the main traffic parameters of interest. These parameters can be estimated through efficient vision-based algorithms and appropriate camera sensor technology. With the advances in camera technologies and increasing computing power, use of monocular vision, stereo vision, and camera sensor fusion technologies have been an active research area in the field of ITS. In this thesis, we investigated stationary monocular and stereo camera technology for traffic parameters estimation. Stationary camera sensors provide large spatial-temporal information of the road section with relatively low installation costs. Two novel scientific contributions for vehicle detection and recognition are proposed. The first one is the use stationary stereo camera technology, and the second contribution is the fusion of monocular and stereo camera technologies. A vision-based ITS consists of several hardware and software components. The overall performance of such a system does not only depend on these single modules but also on their interaction. Therefore, a systematic approach considering all essential modules was chosen instead of focusing on one element of the complete system chain. This leads to detailed investigations of several core algorithms, e.g. background subtraction, histogram based fingerprints, and data fusion methods. From experimental results on standard datasets, we concluded that proposed fusion-based approach, consisting of monocular and stereo camera technologies performs better than each particular technology for vehicle detection and vehicle recognition. Moreover, this research work has a potential to provide a low-cost vision-based solution for online traffic monitoring systems in urban and rural environments.

This dissertation presents a novel approach to develop a system for real-time tracking of the position and orientation of the human hand in three-dimensional space, using MEMS inertial measurement units (IMUs) and infrared cameras. This research focuses on the study and implementation of an algorithm to correct the gyroscope drift, which is a major problem in orientation tracking using commercial-grade IMUs. An algorithm to improve the orientation estimation is proposed. It consists of: 1.) Prediction of the bias offset error while the sensor is static, 2.) Estimation of a quaternion orientation from the unbiased angular velocity, 3.) Correction of the orientation quaternion utilizing the gravity vector and the magnetic North vector, and 4.) Adaptive quaternion interpolation, which determines the final quaternion estimate based upon the current conditions of the sensor. The results verified that the implementation of the orientation correction algorithm using the gravity vector and the magnetic North vector is able to reduce the amount of drift in orientation tracking and is compatible with position tracking using infrared cameras for real-time human hand motion tracking. Thirty human subjects participated in an experiment to validate the performance of the hand motion tracking system. The statistical analysis shows that the error of position tracking is, on average, 1.7 cm in the x-axis, 1.0 cm in the y-axis, and 3.5 cm in the z-axis. The Kruskal-Wallis tests show that the orientation correction algorithm using gravity vector and magnetic North vector can significantly reduce the errors in orientation tracking in comparison to fixed offset compensation. Statistical analyses show that the orientation correction algorithm using gravity vector and magnetic North vector and the on-board Kalman-based orientation filtering produced orientation errors that were not significantly different in the Euler angles, Phi, Theta and Psi, with the p-values of 0.632, 0.262 and 0.728, respectively. The proposed orientation correction algorithm represents a contribution to the emerging approaches to obtain reliable orientation estimates from MEMS IMUs. The development of a hand motion tracking system using IMUs and infrared cameras in this dissertation enables future improvements in natural human-computer interactions within a 3D virtual environment.

No description available.

<p>Object detection is an increasingly popular tool used in many fields, especially in the<br> development of autonomous vehicles. The task of object detections involves the localization<br> of objects in an image, constructing a bounding box to determine the presence and loca-<br> tion of the object, and classifying each object into its appropriate class. Object detection<br> applications are commonly implemented using convolutional neural networks along with the<br> construction of feature pyramid networks to extract data.<br> Another commonly used technique in the automotive industry is sensor fusion. Each<br> automotive sensor – camera, radar, and lidar – have their own advantages and disadvantages.<br> Fusing two or more sensors together and using the combined information is a popular method<br> of balancing the strengths and weakness of each independent sensor. Together, using sensor<br> fusion within an object detection network has been found to be an effective method of<br> obtaining accurate models. Accurate detections and classifications of images is a vital step<br> in the development of autonomous vehicles or self-driving cars.<br> Many studies have proposed methods to improve neural networks or object detection<br> networks. Some of these techniques involve data augmentation and hyperparameter opti-<br> mization. This thesis achieves the goal of improving a camera and radar fusion network by<br> implementing various techniques within these areas. Additionally, a novel idea of integrating<br> a third sensor, the lidar, into an existing camera and radar fusion network is explored in this<br> research work.<br> The models were trained on the Nuscenes dataset, one of the biggest automotive datasets<br> available today. Using the concepts of augmentation, hyperparameter optimization, sensor<br> fusion, and annotation filters, the CRF-Net was trained to achieve an accuracy score that<br> was 69.13% higher than the baseline</p>

Kalman filtering is a well-established method for fusing sensor data in order to accuratelyestimate unknown variables. Recently, the unscented Kalman filter (UKF) has beenused due to its ability to propagate the first and second moments of the probability distribution of an estimated state through a non-linear transformation. The design of ageneric algorithm which implements this filter occupies the first part of this thesis. The generality and functionality of the filter were tested on a toy example and the results are within machine accuracy when compared to those of an equivalent C++ implementation.Application of this filter to the attitude navigation problem becomes non-trivial when coupled to quaternions. Challenges present include the non-commutation of rotations and the dimensionality difference between quaternions and the degrees of freedom of the motion. The second part of this thesis deals with the formulation of the UKF in the quaternion space. This was achieved by implementing an error-state formulation of the process model, bounding estimation in the infinitesimal space and thus de-coupling rotations from non-commutation and bridging the dimensionality discrepancy of quaternions and their respective covariances.The attitude navigation algorithm was then tested using an IMU and a magnetometer.Results show a bounded estimation error which settles to around 1 degree. A detailed look of the filter mechanization process was also presented showing expected behavior for estimation of the initial attitude with error tolerance of 1 mdeg. The structure and design of the proposed formulation allows for trivially incorporating other sensors inthe estimation process and more intricate modelling of the stochastic processes present,potentially leading to greater estimation accuracy. / Kalman filtrering är en vältablerad metod for att sammanväga sensordata för att erhålla noggranna estimat av okända variabler. Nyligen har den typ av kalman filter som kallas unscented Kalman filter (UKF) ökat i populäritet pa grund av dess förmåga att propagera de första och andra momenten för sannolikhetsfördelningen för ett estimera tillstånd genom en ickelinjär transformation. Designen av en generisk algoritm som implementerar denna typ av filter upptar den första delen av denna avhandling. Generaliteten och funktionaliteten för detta filter testades på ett minimalt exempel och resultaten var identiska med de för en ekvivalent C++-implementation till den noggrannhet som tillåts av den nita maskinprecisionen. Användandet av detta filter för attitydnavigering blir icke-trivialt när det anvands forkvaternioner. De utmaningar som uppstar inkluderar att rotationer inte kommuterar och att de finns en skillnad i dimensionalitet mellan kvaternioner och antalet frihetsgrader i rörelsen. Den andra delen av denna avhandling behandlar formuleringen av ett UKF för ett tillstånd som inkluderar en kvaternion. Detta gjordes genom att implementera en så kallad error state-formulering av processmodellen, vilken begränsar estimeringen till ett innitesimalt tillstånd och därigenom undviker problemen med att kvaternionmultiplikation inte kommuterar och överbryggar skillnaden i dimensionalitet hos kvaternioner och deras motsvarande vinkelosäkerheter.Attitydnavigeringen testades sedan med hjälp av en IMU och en magnetometer.Resultaten visade ett begränsat estimeringsfel som ställer in sig kring 1 grad. Strukturen och designen av den föreslagna formuleringen möjliggör på ett rattframt satt tillägg av andra sensorer i estimeringsprocessen och mer detaljerad modellering av de stokastiska processerna, vilket potentiellt leder till högre estimering noggrannhet.

Today, everyday objects are becoming more and more intelligent and some-times even have self-learning capabilities. These self-learning capacities in particular also act as catalysts for new developments in the steel industry.Technical developments that enhance the sustainability and productivity of steel production are very much in demand in the long-term. The methods of Industry 4.0 can support the steel production process in a way that enables steel to be produced in a more cost-effective and environmentally friendly manner. This thesis describes the development of an opto-acoustic system for the early detection of slag slopping in the BOF (Basic Oxygen Furnace) converter process. The prototype has been installed in Salzgitter Stahlwerks, a German steel plant for initial testing. It consists of an image monitoring camera at the converter mouth, a sound measurement system and an oscillation measurement device installed at the blowing lance. The camera signals are processed by a special image processing software. These signals are used to rate the amount of spilled slag and for a better interpretation of both the sound data and the oscillation data. A certain aspect of the opto-acoustic system for slopping detection is that all signals, i.e. optic, acoustic and vibratory, are affected by process-related parameters which are not always relevant for the slopping event. These uncertainties affect the prediction of the slopping phenomena and ultimately the reliability of the entire slopping system. Machine Learning algorithms have been been applied to predict the Slopping phenomenon based on the data from the sensors as well as the other process parameters. / Idag blir vardagliga föremål mer och mer intelligenta och ibland har de självlärande möjligheter. Dessa självlärande förmågor fungerar också som katalysatorer för den nya utvecklingen inom stålindustrin. Teknisk utveckling som stärker hållbarheten och produktiviteten i stålproduktionen är mycket efterfrågad på lång sikt. Metoderna för Industry 4.0 kan stödja stålproduktionsprocessen på ett sätt som gör att stål kan produceras på ett mer kostnadseffektivt och miljövänligt sätt. Denna avhandling beskriver utvecklingen av ett opto-akustiskt system för tidig detektering av slaggsslipning i konverteringsprocessen BOF (Basic Oxygen Furnace). Prototypen har installerats i Salzgitter Stahlwerks, en tysk stålverk för första provning. Den består av en bildövervakningskamera på omvandlarens mun, ett ljudmätningssystem och en oscillationsmätningsenhet som installeras vid blåsans. Kamerans signaler behandlas av en speciell bildbehandlingsprogram. Dessa signaler används för att bestämma mängden spilld slagg och för bättre tolkning av både ljuddata och oscillationsdata. En viss aspekt av det optoakustiska systemet för släckningsdetektering är att alla signaler, dvs optiska, akustiska och vibrerande, påverkas av processrelaterade parametrar som inte alltid är relevanta för slöjningsevenemanget. Dessa osäkerheter påverkar förutsägelsen av slopfenomenerna och i slutändan tillförlitligheten för hela slöjningssystemet. Maskininlärningsalgoritmer har tillämpats för att förutsäga Slopping-fenomenet baserat på data från sensorerna liksom de andra processparametrarna.

Location-based services (LBSs) have become pervasive, and the demand for these systems and services is rising. Indoor Positioning Systems (IPSs) are key to extend location-based services indoors where the Global Positioning System (GPS) is not reliable due to low signal strength and complicated signal propagation environment. Most existing IPSs either require the installation of special hardware devices or build a fingerprint map, which is expensive, time-consuming, and labor-intensive. Developments in microelectromechanical systems (MEMS) have resulted in significant advancements in the low-cost compact MARG inertial sensors, making it possible to achieve low-cost and high-accuracy IPSs. This research considers the indoor positioning problem and aims to design and develop an infrastructure-free self-contained indoor positioning and tracking system based on Pedestrian Dead Reckoning (PDR) using MEMS MARG inertial sensors. PDR-based systems rely on MARG inertial sensor measurements to estimate the current position of the object by using a previously determined position without external references. Many issues still exist in developing such systems, such as cumulative errors, high-frequency sensor noises, the gyro drift issue, magnetic distortions, etc. As the MARG sensors are inherently error-prone, the most significant challenge is how to design sensor fusion models and algorithms to accurately extract useful location-based information from individual motion and magnetic sensors. The objective of this thesis is to solve these issues and mitigate the challenges. The proposed positioning system is designed with four main modules at the system level and a dual-mode feature. Specifically, the four main modules are mode detection, step detection and moving distance estimation, heading and orientation estimation, and position estimation. To address the cumulative error issue of using low-cost inertial sensors, signal processing and sensor fusion techniques are utilized for algorithm design. Experimental evaluations show that the proposed position estimation algorithm is able to achieve high positioning accuracy at low costs for the indoor environment. / Thesis / Master of Applied Science (MASc) / With the maturity of microelectromechanical systems (MEMS) technology in recent years, Magnetic, Angular Rate, and Gravity (MARG) sensors are embedded in most smart devices. This research considers the indoor positioning problem and aims to design and develop an infrastructure-free self-contained MEMS MARG inertial sensor-based indoor positioning and tracking system with high precision. The proposed positioning system uses the Pedestrian Dead Reckoning (PDR) approach and includes four main modules at the system level with a dual-mode feature. Specifically, the four main modules are mode detection, step detection and moving distance estimation, heading and orientation estimation, and position estimation. The two modes are static mode and dynamic mode. To address the cumulative error issue of using low-cost inertial sensors, signal processing and sensor fusion techniques are utilized for algorithm design. The detection and estimation algorithms of each module are presented in the system design chapter. Experimental evaluations including trajectory results under five scenarios show that the proposed position estimation algorithm achieves a higher position accuracy than that of conventional estimation methods.