Rail track asset sighting distance must be checked regularly to ensure the continued and safe operation of rolling stock. Methods currently used to check asset line-of-sight involve manual labour or laser systems. Video cameras and computer vision techniques provide one possible route for cheaper, automated systems. Three categories of computer vision method are identified for possible application: two-dimensional object recognition, two-dimensional object tracking and three-dimensional scene recovery. However, presented experimentation shows recognition and tracking methods produce less accurate asset line-of-sight results for increasing asset-camera distance. Regarding three-dimensional scene recovery, evidence is presented suggesting a relationship between image feature and recovered scene information. A novel framework which learns these relationships is proposed. Learnt relationships from recovered image features probabilistically limit the search space of future features, improving efficiency. This framework is applied to several scene recovery methods and is shown (on average) to decrease computation by two-thirds for a possible, small decrease in accuracy of recovered scenes. Asset line-of-sight results computed from recovered three-dimensional terrain data are shown to be more accurate than two-dimensional methods, not effected by increasing asset-camera distance. Finally, the analysis of terrain in terms of effect on asset line-of-sight is considered. Terrain elements, segmented using semantic information, are ranked with a metric combining a minimum line-of-sight blocking distance and the growth required to achieve this minimum distance. Since this ranking measure is relative, it is shown how an approximation of the terrain data can be applied, decreasing computation time. Further efficiency increases are found by decomposing the problem into a set of two-dimensional problems and applying binary search techniques. The combination of the research elements presented in this thesis provide efficient methods for automatically analysing asset line-of-sight and the impact of the surrounding terrain, from captured monocular video.

An adaptive autopilot design for an uninhabited surface vehicle Andy SK Annamalai The work described herein concerns the development of an innovative approach to the design of autopilot for uninhabited surface vehicles. In order to fulfil the requirements of autonomous missions, uninhabited surface vehicles must be able to operate with a minimum of external intervention. Existing strategies are limited by their dependence on a fixed model of the vessel. Thus, any change in plant dynamics has a non-trivial, deleterious effect on performance. This thesis presents an approach based on an adaptive model predictive control that is capable of retaining full functionality even in the face of sudden changes in dynamics. In the first part of this work recent developments in the field of uninhabited surface vehicles and trends in marine control are discussed. Historical developments and different strategies for model predictive control as applicable to surface vehicles are also explored. This thesis also presents innovative work done to improve the hardware on existing Springer uninhabited surface vehicle to serve as an effective test and research platform. Advanced controllers such as a model predictive controller are reliant on the accuracy of the model to accomplish the missions successfully. Hence, different techniques to obtain the model of Springer are investigated. Data obtained from experiments at Roadford Reservoir, United Kingdom are utilised to derive a generalised model of Springer by employing an innovative hybrid modelling technique that incorporates the different forward speeds and variable payload on-board the vehicle. Waypoint line of sight guidance provides the reference trajectory essential to complete missions successfully. The performances of traditional autopilots such as proportional integral and derivative controllers when applied to Springer are analysed. Autopilots based on modern controllers such as linear quadratic Gaussian and its innovative variants are integrated with the navigation and guidance systems on-board Springer. The modified linear quadratic Gaussian is obtained by combining various state estimators based on the Interval Kalman filter and the weighted Interval Kalman filter. Change in system dynamics is a challenge faced by uninhabited surface vehicles that result in erroneous autopilot behaviour. To overcome this challenge different adaptive algorithms are analysed and an innovative, adaptive autopilot based on model predictive control is designed. The acronym ‘aMPC’ is coined to refer to adaptive model predictive control that is obtained by combining the advances made to weighted least squares during this research and is used in conjunction with model predictive control. Successful experimentation is undertaken to validate the performance and autonomous mission capabilities of the adaptive autopilot despite change in system dynamics.

We consider time-of-arrival based robust geolocation in harsh line-of-sight/non-line-of-sight environments. Herein, we assume the probability density function (PDF) of the measurement error to be completely unknown and develop an iterative algorithm for robust position estimation. The iterative algorithm alternates between a PDF estimation step, which approximates the exact measurement error PDF (albeit unknown) under the current parameter estimate via adaptive kernel density estimation, and a parameter estimation step, which resolves a position estimate from the approximate log-likelihood function via a quasi-Newton method. Unless the convergence condition is satisfied, the resolved position estimate is then used to refine the PDF estimation in the next iteration. We also present the best achievable geolocation accuracy in terms of the Cramér-Rao lower bound. Various simulations have been conducted in both real-world and simulated scenarios. When the number of received range measurements is large, the new proposed position estimator attains the performance of the maximum likelihood estimator (MLE). When the number of range measurements is small, it deviates from the MLE, but still outperforms several salient robust estimators in terms of geolocation accuracy, which comes at the cost of higher computational complexity.

Nowadays, high portion of tactical missiles use gimbaled seeker. For accurate target tracking, the platform where the gimbal is mounted must be stabilized with respect to the motion of the missile body. Line of sight stabilization is critical for fast and precise tracking and alignment. Although, conventional PID framework solves many stabilization problems, it is reported that many PID feedback loops are poorly tuned. In this thesis, recently introduced robot control method, proxy based sliding mode control, is adopted for the line of sight (LOS) stabilization. Before selecting the proposed method, adaptive neural network sliding mode control and fuzzy control are also implemented for comparative purposes. Experimental and simulation results show a satisfactory response of the proxy based sliding mode controller.

Positioning is recognized as an important feature of fifth generation (\abbrFiveG) cellular networks due to the massive number of commercial use cases that would benefit from access to position information. Radio based positioning has always been a challenging task in urban canyons where buildings block and reflect the radio signal, causing multipath propagation and non-line-of-sight (NLOS) signal conditions. One approach to handle NLOS is to use data-driven methods such as machine learning algorithms on beam-based data, where a training data set with positioned measurements are used to train a model that transforms measurements to position estimates.  The work is based on position and radio measurement data from a 5G testbed. The transmission point (TP) in the testbed has an antenna that have beams in both horizontal and vertical layers. The measurements are the beam reference signal received power (BRSRP) from the beams and the direction of departure (DOD) from the set of beams with the highest received signal strength (RSS). For modelling of the relation between measurements and positions, two non-linear models has been considered, these are neural network and random forest models. These non-linear models will be referred to as machine learning algorithms.  The machine learning algorithms are able to position the user equipment (UE) in NLOS regions with a horizontal positioning error of less than 10 meters in 80 percent of the test cases. The results also show that it is essential to combine information from beams from the different vertical antenna layers to be able to perform positioning with high accuracy during NLOS conditions. Further, the tests show that the data must be separated into line-of-sight (LOS) and NLOS data before the training of the machine learning algorithms to achieve good positioning performance under both LOS and NLOS conditions. Therefore, a generalized likelihood ratio test (GLRT) to classify data originating from LOS or NLOS conditions, has been developed. The probability of detection of the algorithms is about 90\% when the probability of false alarm is only 5%.  To boost the position accuracy of from the machine learning algorithms, a Kalman filter have been developed with the output from the machine learning algorithms as input. Results show that this can improve the position accuracy in NLOS scenarios significantly. / Radiobasserad positionering av användarenheter är en viktig applikation i femte generationens (5G) radionätverk, som mycket tid och pengar läggs på för att utveckla och förbättra. Ett exempel på tillämpningsområde är positionering av nödsamtal, där ska användarenheten kunna positioneras med en noggrannhet på ett tiotal meter. Radio basserad positionering har alltid varit utmanande i stadsmiljöer där höga hus skymmer och reflekterar signalen mellan användarenheten och basstationen. En ide att positionera i dessa utmanande stadsmiljöer är att använda datadrivna modeller tränade av algoritmer baserat på positionerat testdata – så kallade maskininlärningsalgoritmer. I detta arbete har två icke-linjära modeller - neurala nätverk och random forest – bli implementerade och utvärderade för positionering av användarenheter där signalen från basstationen är skymd.% Dessa modeller refereras som maskininlärningsalgoritmer. Utvärderingen har gjorts på data insamlad av Ericsson från ett 5G-prototypnätverk lokaliserat i Kista, Stockholm. Antennen i den basstation som används har 48 lober vilka ligger i fem olika vertikala lager. Insignal och målvärdena till maskininlärningsalgoritmerna är signals styrkan för varje stråle (BRSRP), respektive givna GPS-positioner för användarenheten. Resultatet visar att med dessa maskininlärningsalgoritmer positioneras användarenheten med en osäkerhet mindre än tio meter i 80 procent av försöksfallen. För att kunna uppnå dessa resultat är viktigt att kunna detektera om signalen mellan användarenheten och basstationen är skymd eller ej. För att göra det har ett statistiskt test blivit implementerat. Detektionssannolikhet för testet är över 90 procent, samtidigt som sannolikhet att få falskt alarm endast är ett fåtal procent.\newline \newline%För att minska osäkerheten i positioneringen har undersökningar gjorts där utsignalen från maskininlärningsalgoritmerna filtreras med ett Kalman-filter. Resultat från dessa undersökningar visar att Kalman-filtret kan förbättra presitionen för positioneringen märkvärt.

Accurate indoor positioning of narrowband Internet-of-Things (IoT) sensors has drawn more attention in recent years. The introduction of Bluetooth Low Energy (BLE) technology is one of the latest developments of IoT and especially applicable for Ultra-Low Power (ULP) applications. BLE is an attractive technology for indoor positioning systems because of its low-cost deployment and reasonable accuracy. Efficient indoor positioning can be achieved by deducing the sensor position from the estimated signal Angle-of-Arrival (AoA) at multiple anchors. An anchor is a base station of known position and equipped with a narrowband multi-antenna array. However, the design and implementation of indoor positioning systems based on AoA measurements involve multiple challenges. The first part of this thesis mainly addresses the impact of hardware impairments on the accuracy of AoA measurements. In practice, the subspace-based algorithms such as Multiple Signal Classification (MUSIC) suffer from sensitivity to array calibration errors coming from hardware imperfections. A detailed experimental implementation is performed using a Software Defined Radio (SDR) platform to precisely evaluate the accuracy of AoA measurements. For this purpose, a new Over-the-Air (OTA) calibration method is proposed and the array calibration error is investigated. The experimental results are compared with the theoretical analysis. These results show that array calibration errors can cause some degrees of uncertainty in AoA estimation. Moreover, we propose iterative positioning algorithms based on AoA measurements for low capacity IoT sensors with high accuracy and fair computational complexity. Efficient positioning accuracy is obtained by iterating between the angle and position estimation steps. We first develop a Data-Aided Maximum a Posteriori (DA- MAP) estimator based on the preamble of the transmitted signal. DA-MAP estimator relies on the knowledge of the transmitted signal which makes it impractical for narrowband communications where the preamble is short. For this reason, a Non-Data- Aided Maximum a Posteriori (NDA-MAP) estimator is developed to improve the AoA accuracy. The iterative positioning algorithms are therefore classified as Data-Aided Iterative (DA-It) and Non-Data-Aided Iterative (NDA-It) depending on the knowledge of the transmitted signal that is used for estimation. Both numerical and experimental analyses are carried out to evaluate the performance of the proposed algorithms. The results show that DA-MAP and NDA-MAP estimators are more accurate than MUSIC. The results also show that DA-It comes very close to the performance of the optimal approach that directly estimates the position based on the observation of the received signal, known as Direct Position Estimation (DPE). Furthermore, the NDA-It algorithm significantly outperforms the DA-It because it can use a much higher number of samples; however, it needs more iterations to converge. In addition, we evaluate the computational savings achieved by the iterative schemes compared to DPE through a detailed complexity analysis. Finally, we investigate the performance degradation of the proposed iterative algorithms due to the impact of multipath and NLOS propagation in indoor environments. Therefore, we develop an enhanced iterative positioning algorithm with an anchor selection method in order to identify and exclude NLOS anchors. The numerical results show that applying the anchor selection strategy significantly improves the positioning accuracy in indoor environments. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

The evolution to ultra-dense next generation networks requires a massive increase in throughput and deployment flexibility. Therefore, novel wireless backhaul solutions that can support these demands are needed. In this work we present an approach for a millimeter wave line-of-sight MIMO backhaul design, targeting transmission rates in the order of 100 Gbit/s. We provide theoretical foundations for the concept showcasing its potential, which are confirmed through channel measurements. Furthermore, we provide insights into the system design with respect to antenna array setup, baseband processing, synchronization, and channel equalization. Implementation in a 60 GHz demonstrator setup proves the feasibility of the system concept for high throughput backhauling in next generation networks.

This thesis deals with testing of astronomical measurement system MAAS-1, which was developed at the Institute of Geodesy, Faculty of Civil Engineering, Brno University of Technology. Several reference measurements were done on the terrace of B building Faculty of Civil Engineering during this testing. The data obtained were processed into astronomical geographic coordinates and . These coordinates, as well as atmospheric conditions and calibration measurements, are basis for evaluation. In this test I try to detect the influence of the CCD sensor mounting of the accuracy on the final data and to assess if the construction needs to be modified.

Indoor positioning is desired in many areas for various reasons, such as positioning products in industrial environments, hospital equipment or firefighters inside a building on fire. One even tougher situation where indoor positioning can be useful is locating a specific object on a shelf in a commercial setting. This thesis aims to investigate and design different network deployment strategies in an indoor environment in order to achieve both high position estimation accuracy and availability. The investigation considers the two positioning techniques downlink time difference of arrival, DL-TDOA, and round trip time, RTT. Simulations of several deployments are performed in two standard scenarios which mimic an indoor open office and an indoor factory, respectively. Factors having an impact on the positioning accuracy and availability are found to be deployment geometry, number of base stations, line-of-sight conditions and interference, with the most important being deployment geometry. Two deployment strategies are designed with the goal of optimising the deployment geometry. In order to achieve both high positioning accuracy and availability in a simple, sparsely cluttered environment, the strategy is to deploy the base stations evenly around the edges of the deployment area. In a more problematic, densely cluttered environment the approach somewhat differs. The proposed strategy is now to identify and strategically place some base stations in the most cluttered areas but still place a majority of the base stations around the edges of the deployment area. A robust positioning algorithm is able to handle interference well and to decrease its impact on the positioning accuracy. The cost, in terms of frequency resources, of using more orthogonal signals may not be worth the small improvement in accuracy and availability.

Fog of War is a term for uncertainty in situational awareness. Fog of War is an essential part of a wargame which causes the participating units’ perception of the environment to be distorted and altered. Introducing a certain amount of uncertainty helps to better mimic the situation on the battlefield. Fog of War comes in multiple forms and levels, whereas the visual detection level is of primary interest for this thesis. Two forms of visual detection have been implemented to simulate a simple and advanced form of Fog of War. The simple level is based solely on the distance between two units, while the advanced level determines whether two units possess a clear line of sight between them, to decisively add realism to a played scenario. The two models were created based on the Entity Component System software architecture, and the maps used for the wargame were based on data from ArcGIS. Extensive testing of the two models, for different types of terrains, show good performance and computational efficiency, however with the expected caveat that flat landscape requires significantly more processing power and memory capacity than a hilly terrain. / Krigsdimma är en term för osäkerhet inom situationsmedvetenhet. Krigsdimma är en väsentlig del av ett krigsspel och medför att deltagande förbands uppfattning av miljön förvrängs och förändras samt att ett visst mått av osäkerhet introduceras för att bättre efterlikna situationen på slagfältet. Krigsdimma kommer i flera former och flera nivåer, där visuell detektering är av primärt intresse för denna avhandling. Två former av visuell detektering har implementerats för att simulera en enkel och en avancerad form av krigsdimma. Den enkla nivån bygger enbart på avståndet mellan två förband medan den avancerade nivån kan avgöra om två enheter i verkligheten har en fri siktlinje mellan sig, något som på ett avgörande sätt kan tillföra realism till ett spelat scenario. De två realiseringarna skapades baserat på en Entity Component System mjukvaruarkitektur, och kartorna som användes för krigsspelet baserade sig på data från ArcGIS. Omfattande tester av de två modellerna, för olika terrängtyper, visar på god funktion och beräkningseffektivitet, dock kräver flackt landskap betydligt mer processorkraft och minneskapacitet än kuperad terräng.