Efficient topology estimation for large scale optical mapping

Large scale image mosaicing methods are in great demand among scientists who study different aspects of the seabed, and have been fostered by impressive advances in the capabilities of underwater robots in gathering optical data from the seafloor. Cost and weight constraints mean that lowcost Remotely operated vehicles (ROVs) usually have a very limited number of sensors. When a low-cost robot carries out a seafloor survey using a down-looking camera, it usually follows a predetermined trajectory that provides several non time-consecutive overlapping image pairs. Finding these pairs (a process known as topology estimation) is indispensable to obtaining globally consistent mosaics and accurate trajectory estimates, which are necessary for a global view of the surveyed area, especially when optical sensors are the only data source. This thesis presents a set of consistent methods aimed at creating large area image mosaics from optical data obtained during surveys with low-cost underwater vehicles. First, a global alignment method developed within a Feature-based image mosaicing (FIM) framework, where nonlinear minimisation is substituted by two linear steps, is discussed. Then, a simple four-point mosaic rectifying method is proposed to reduce distortions that might occur due to lens distortions, error accumulation and the difficulties of optical imaging in an underwater medium. The topology estimation problem is addressed by means of an augmented state and extended Kalman filter combined framework, aimed at minimising the total number of matching attempts and simultaneously obtaining the best possible trajectory. Potential image pairs are predicted by taking into account the uncertainty in the trajectory. The contribution of matching an image pair is investigated using information theory principles. Lastly, a different solution to the topology estimation problem is proposed in a bundle adjustment framework. Innovative aspects include the use of fast image similarity criterion combined with a Minimum spanning tree (MST) solution, to obtain a tentative topology. This topology is improved by attempting image matching with the pairs for which there is the most overlap evidence. Unlike previous approaches for large-area mosaicing, our framework is able to deal naturally with cases where time-consecutive images cannot be matched successfully, such as completely unordered sets. Finally, the efficiency of the proposed methods is discussed and a comparison made with other state-of-the-art approaches, using a series of challenging datasets in underwater scenarios

Topology-based denoising of chaos

In this article, we propose a denoising algorithm to denoise a time series y(i) = x(i) + e(i), where {x(i)} is a time series obtained from a time- T map of a uniformly hyperbolic or Anosov flow, and {e(i)} a uniformly bounded sequence of independent and identically distributed (i.i.d.) random variables. Making use of observations up to time n, we create an estimate of x(i) for i<n. We show under typical limiting behaviours of the orbit and the recurrence properties of x(i), the estimation error converges to zero as n tends to infinity with probability 1.

Pipelined ring algorithm for matrix multiplication on transputer networks: performance analysis and estimation

A parallel matrix multiplication algorithm is presented, and studies of its performance and estimation are discussed. The algorithm is implemented on a network of transputers connected in a ring topology. An efficient scheme for partitioning the input matrices is introduced which enables overlapping computation with communication. This makes the algorithm achieve near-ideal speed-up for reasonably large matrices. Analytical expressions for the execution time of the algorithm have been derived by analysing its computation and communication characteristics. These expressions are validated by comparing the theoretical results of the performance with the experimental values obtained on a four-transputer network for both square and irregular matrices. The analytical model is also used to estimate the performance of the algorithm for a varying number of transputers and varying problem sizes. Although the algorithm is implemented on transputers, the methodology and the partitioning scheme presented in this paper are quite general and can be implemented on other processors which have the capability of overlapping computation with communication. The equations for performance prediction can also be extended to other multiprocessor systems.

Real-time temperature estimation for power MOSFETs considering thermal aging effects

This paper presents a novel real-time power-device temperature estimation method that monitors the power MOSFET's junction temperature shift arising from thermal aging effects and incorporates the updated electrothermal models of power modules into digital controllers. Currently, the real-time estimator is emerging as an important tool for active control of device junction temperature as well as online health monitoring for power electronic systems, but its thermal model fails to address the device's ongoing degradation. Because of a mismatch of coefficients of thermal expansion between layers of power devices, repetitive thermal cycling will cause cracks, voids, and even delamination within the device components, particularly in the solder and thermal grease layers. Consequently, the thermal resistance of power devices will increase, making it possible to use thermal resistance (and junction temperature) as key indicators for condition monitoring and control purposes. In this paper, the predicted device temperature via threshold voltage measurements is compared with the real-time estimated ones, and the difference is attributed to the aging of the device. The thermal models in digital controllers are frequently updated to correct the shift caused by thermal aging effects. Experimental results on three power MOSFETs confirm that the proposed methodologies are effective to incorporate the thermal aging effects in the power-device temperature estimator with good accuracy. The developed adaptive technologies can be applied to other power devices such as IGBTs and SiC MOSFETs, and have significant economic implications. 

Estimation des longueurs de branche et artefact sur la datation moléculaire

La phylogénie moléculaire fournit un outil complémentaire aux études paléontologiques et géologiques en permettant la construction des relations phylogénétiques entre espèces ainsi que l’estimation du temps de leur divergence. Cependant lorsqu’un arbre phylogénétique est inféré, les chercheurs se focalisent surtout sur la topologie, c'est-à-dire l’ordre de branchement relatif des différents nœuds. Les longueurs des branches de cette phylogénie sont souvent considérées comme des sous-produits, des paramètres de nuisances apportant peu d’information. Elles constituent cependant l’information primaire pour réaliser des datations moléculaires. Or la saturation, la présence de substitutions multiples à une même position, est un artefact qui conduit à une sous-estimation systématique des longueurs de branche. Nous avons décidé d’estimer l‘influence de la saturation et son impact sur l’estimation de l’âge de divergence. Nous avons choisi d’étudier le génome mitochondrial des mammifères qui est supposé avoir un niveau élevé de saturation et qui est disponible pour de nombreuses espèces. De plus, les relations phylogénétiques des mammifères sont connues, ce qui nous a permis de fixer la topologie, contrôlant ainsi un des paramètres influant la longueur des branches. Nous avons utilisé principalement deux méthodes pour améliorer la détection des substitutions multiples : (i) l’augmentation du nombre d’espèces afin de briser les plus longues branches de l’arbre et (ii) des modèles d’évolution des séquences plus ou moins réalistes. Les résultats montrèrent que la sous-estimation des longueurs de branche était très importante (jusqu'à un facteur de 3) et que l’utilisation d'un grand nombre d’espèces est un facteur qui influence beaucoup plus la détection de substitutions multiples que l’amélioration des modèles d’évolutions de séquences. Cela suggère que même les modèles d’évolution les plus complexes disponibles actuellement, (exemple: modèle CAT+Covarion, qui prend en compte l’hétérogénéité des processus de substitution entre positions et des vitesses d’évolution au cours du temps) sont encore loin de capter toute la complexité des processus biologiques. Malgré l’importance de la sous-estimation des longueurs de branche, l’impact sur les datations est apparu être relativement faible, car la sous-estimation est plus ou moins homothétique. Cela est particulièrement vrai pour les modèles d’évolution. Cependant, comme les substitutions multiples sont le plus efficacement détectées en brisant les branches en fragments les plus courts possibles via l’ajout d’espèces, se pose le problème du biais dans l’échantillonnage taxonomique, biais dû à l‘extinction pendant l’histoire de la vie sur terre. Comme ce biais entraine une sous-estimation non-homothétique, nous considérons qu’il est indispensable d’améliorer les modèles d’évolution des séquences et proposons que le protocole élaboré dans ce travail permettra d’évaluer leur efficacité vis-à-vis de la saturation.

Identifying multiple interacting bad data in power system state estimation

This paper presents an intelligent search strategy for the conforming bad data errors identification in the generalized power system state estimation, by using the tabu search meta heuristic. The main objective is to detect critical errors involving both analog and topology errors. These errors are represented by conforming errors, whose nature affects measurements that actually do not present bad data and also the conventional bad data identification strategies based on the normalized residual methods. ©2005 IEEE.

Wind tunnel analysis of the flow topology over complex topographies

El principal objetivo de este trabajo es aportar conocimiento para contestar la pregunta: ¿hasta que punto los ensayos en túnel aerodinámico pueden contribuir a determinar las características que afectan la respuesta dinámica de los aerogeneradores operando en terreno complejo?. Esta pregunta no es nueva, de hecho, el debate en la comunidad científica comenzó en el primer tercio del siglo pasado y aún está intensamente vivo. El método generalmente aceptado para enfrentar el mencionado problema consiste en analizar un caso de estudio determinado en el cual se aplican tanto ensayos a escala real como análisis computacionales y ensayos en túnel aerodinámico. Esto no es ni fácil ni barato. Esta es la razón por la cual desde el experimento de Askervein en 1988, los modelizadores del flujo atmosférico tuvieron que esperar hasta 2007 a que el experimento de Bolund fuese puesto en marcha con un despliegue de medios técnicos equivalentes (teniendo en cuenta la evolución de las tecnologías de sensores y computación). El problema contempla tantos aspectos que ambas experiencias fueron restringidas a condiciones de atmósfera neutra con efectos de Coriolis despreciables con objeto de reducir la complejidad. Este es el contexto en el que se ha desarrollado la presente tesis doctoral. La topología del flujo sobre la isla de Bolund ha sido estudiada mediante la reproducción del experimento de Bolund en los túneles aerodinámicos A9 y ACLA16 del IDR. Dos modelos de la isla de Bolund fueron fabricados a dos escalas, 1:230 y 1:115. El flujo de entrada en el túnel aerodinámico simulando la capa límite sin perturbar correspondía a régimen de transición (transitionally rough regime) y fue usado como situación de referencia. El modelo a escala 1:230 fue ensayado en el túnel A9 para determinar la presión sobre su superficie. La distribución del coeficiente de presión sobre la isla proporcionó una visualización y estimación de una región de desprendimiento sobre el pequeño acantilado situado al frente de la misma. Las medidas de presión instantánea con suficiente grado de resolución temporal pusieron de manifiesto la no estacionariedad en la región de desprendimiento. El modelo a escala 1:115 fue ensayado utilizando hilo caliente de tres componentes y un sistema de velocimetría por imágenes de partículas de dos componentes. El flujo fue caracterizado por el ratio de aceleración, el incremento normalizado de energía cinética turbulenta y los ángulos de inclinación y desviación horizontal. Los resultados a lo largo de la dirección 270°y alturas de 2 m y 5 m presentaron una gran similitud con los resultados a escala real del experimento de Bolund. Los perfiles verticales en las localizaciones de las torres meteorológicas mostraron un acuerdo significativo con los resultados a escala real. El análisis de los esfuerzos de Reynolds y el análisis espectral en las localizaciones de los mástiles meteorológicos presentaron niveles de acuerdo variados en ciertas posiciones, mientras que en otras presentaron claras diferencias. El mapeo horizontal del flujo, para una dirección de viento de 270°, permitió caracterizar el comportamiento de la burbuja intermitente de recirculación sobre el pequeño acantilado existente al frente de la isla así como de la región de relajación y de la capa de cortadura en la región corriente abajo de Bolund. Se realizaron medidas de velocidad con alta resolución espacial en planos perpendiculares a la dirección del flujo sin perturbar. Estas medidas permitieron detectar y caracterizar una estructura de flujo similar a un torbellino longitudinal con regiones con altos gradientes de velocidad y alta intensidad de turbulencia. Esta estructura de flujo es, sin duda, un reto para los modelos computacionales y puede considerarse un factor de riesgo para la operación de los aerogeneradores. Se obtuvieron y analizaron distribuciones espaciales de los esfuerzos de Reynolds mediante 3CHW y PIV. Este tipo de parámetros no constituyen parte de los resultados habituales en los ensayos en túnel sobre topografías y son muy útiles para los modelizadores que utilizan simulación de grades torbellinos (LES). Se proporciona una interpretación de los resultados obtenidos en el túnel aerodinámico en términos de utilidad para los diseñadores de parques eólicos. La evolución y variación de los parámetros del flujo a lo largo de líneas, planos y superficies han permitido identificar como estas propiedades del flujo podrían afectar la localización de los aerogeneradores y a la clasificación de emplazamientos. Los resultados presentados sugieren, bajo ciertas condiciones, la robustez de los ensayos en túnel para estudiar la topología sobre terreno complejo y su comparabilidad con otras técnicas de simulación, especialmente considerando el nivel de acuerdo del conjunto de resultados presentados con los resultados a escala real. De forma adicional, algunos de los parámetros del flujo obtenidos de las medidas en túnel son difícilmente determinables en ensayos a escala real o por medios computacionales, considerado el estado del arte. Este trabajo fue realizado como parte de las actividades subvencionadas por la Comisión Europea como dentro del proyecto FP7-PEOPLE-ITN-2008WAUDIT (Wind Resource Assessment Audit and Standardization) dentro de la FP7 Marie-Curie Initial Training Network y por el Ministerio Español de Economía y Competitividad dentro del proyecto ENE2012-36473, TURCO (Determinación en túnel aerodinámico de la distribución espacial de parámetros estadísticos de la turbulencia atmosférica sobre topografías complejas) del Plan Nacional de Investigación (Subprograma de investigación fundamental no orientada 2012). El informe se ha organizado en siete capítulos y un conjunto de anexos. En el primer capítulo se introduce el problema. En el capítulo dos se describen los medios experimentales utilizados. Seguidamente, en el capítulo tres, se analizan en detalle las condiciones de referencia del principal túnel aerodinámico utilizado en esta investigación. En el capítulo tres se presentan resultados de ensayos de presión superficial sobre un modelo de la isla. Los principales resultados del experimento de Bolund se reproducen en el capítulo cinco. En el capítulo seis se identifican diferentes estructuras del flujo sobre la isla y, finalmente, en el capitulo siete, se recogen las conclusiones y una propuesta de lineas de trabajo futuras. ABSTRACT The main objective of this work is to contribute to answer the question: to which extend can the wind tunnel testing contribute to determine the flow characteristics that affect the dynamic response of wind turbines operating in highly complex terrains?. This question is not new, indeed, the debate in the scientific community was opened in the first third of the past century and it is still intensely alive. The accepted approach to face this problem consists in analysing a given case study where full-scale tests, computational modelling and wind tunnel testing are applied to the same topography. This is neither easy nor cheap. This is is the reason why since the Askervein experience in 1988, the atmospheric flow modellers community had to wait till 2007 when the Bolund experiment was setup with a deployment of technical means equivalent (considering the evolution of the sensor and computing techniques). The problem is so manifold that both experiences were restricted to neutral conditions without Coriolis effects in order to reduce the complexity. This is the framework in which this PhD has been carried out. The flow topology over the Bolund Island has been studied by replicating the Bolund experiment in the IDR A9 and ACLA16 wind tunnels. Two mock-ups of the Bolund island were manufactured at two scales of 1:230 and 1:115. The in-flow in the empty wind tunnel simulating the incoming atmospheric boundary layer was in the transitionally rough regime and used as a reference case. The 1:230 model was tested in the A9 wind tunnel to measure surface pressure. The mapping of the pressure coefficient across the island gave a visualisation and estimation of a detachment region on the top of the escarpment in front of the island. Time resolved instantaneous pressure measurements illustrated the non-steadiness in the detachment region. The 1:115 model was tested using 3C hot-wires(HW) and 2C Particle Image Velocimetry(PIV). Measurements at met masts M3, M6, M7 and M8 and along Line 270°were taken to replicate the result of the Bolund experiment. The flow was characterised by the speed-up ratio, normalised increment of the turbulent kinetic energy, inclination angle and turning angle. Results along line 270°at heights of 2 m and 5 m compared very well with the full-scale results of the Bolund experiment. Vertical profiles at the met masts showed a significant agreement with the full-scale results. The analysis of the Reynolds stresses and the spectral analysis at the met mast locations gave a varied level of agreement at some locations while clear mismatch at others. The horizontal mapping of the flow field, for a 270°wind direction, allowed to characterise the behaviour of the intermittent recirculation bubble on top of the front escarpment followed by a relaxation region and the presence of a shear layer in the lee side of the island. Further detailed velocity measurements were taken at cross-flow planes over the island to study the flow structures on the island. A longitudinal vortex-like structure with high mean velocity gradients and high turbulent kinetic energy was characterised on the escarpment and evolving downstream. This flow structure is a challenge to the numerical models while posing a threat to wind farm designers when siting wind turbines. Spatial distribution of Reynold stresses were presented from 3C HW and PIV measurements. These values are not common results from usual wind tunnel measurements and very useful for modellers using large eddy simulation (LES). An interpretation of the wind tunnel results in terms of usefulness to wind farm designers is given. Evolution and variation of the flow parameters along measurement lines, planes and surfaces indicated how the flow field could affect wind turbine siting. Different flow properties were presented so compare the level of agreement to full-scale results and how this affected when characterising the site wind classes. The results presented suggest, under certain conditions, the robustness of the wind tunnel testing for studying flow topology over complex terrain and its capability to compare to other modelling techniques especially from the level of agreement between the different data sets presented. Additionally, some flow parameters obtained from wind tunnel measurements would have been quite difficult to be measured at full-scale or by computational means considering the state of the art. This work was carried out as a part of the activities supported by the EC as part of the FP7- PEOPLE-ITN-2008 WAUDIT project (Wind Resource Assessment Audit and Standardization) within the FP7 Marie-Curie Initial Training Network and by the Spanish Ministerio de Economía y Competitividad, within the framework of the ENE2012-36473, TURCO project (Determination of the Spatial Distribution of Statistic Parameters of Flow Turbulence over Complex Topographies in Wind Tunnel) belonging to the Spanish National Program of Research (Subprograma de investigación fundamental no orientada 2012). The report is organised in seven chapters and a collection of annexes. In chapter one, the problem is introduced. In chapter two the experimental setup is described. Following, in chapter three, the inflow conditions of the main wind tunnel used in this piece of research are analysed in detail. In chapter three, preliminary pressure tests results on a model of the island are presented. The main results from the Bolund experiment are replicated in chapter five. In chapter six, an identification of specific flow strutures over the island is presented and, finally, in chapter seven, conclusions and lines for future works related to the presented one are included.

Real time motion estimation using a neural architecture implemented on GPUs

This work describes a neural network based architecture that represents and estimates object motion in videos. This architecture addresses multiple computer vision tasks such as image segmentation, object representation or characterization, motion analysis and tracking. The use of a neural network architecture allows for the simultaneous estimation of global and local motion and the representation of deformable objects. This architecture also avoids the problem of finding corresponding features while tracking moving objects. Due to the parallel nature of neural networks, the architecture has been implemented on GPUs that allows the system to meet a set of requirements such as: time constraints management, robustness, high processing speed and re-configurability. Experiments are presented that demonstrate the validity of our architecture to solve problems of mobile agents tracking and motion analysis.

Reliability estimation of second life battery system power electronic topologies for grid frequency response applications

This paper is part of a project which aims to research the opportunities for the re-use of batteries after their primary use in low and ultra low carbon vehicles on the electricity grid system. One potential revenue stream is to provide primary/secondary/high frequency response to National Grid through market mechanisms via DNO's or Energy service providers. Some commercial battery energy storage systems (BESS) already exist on the grid system, but these tend to use costly new or high performance batteries. Second life batteries should be available at lower cost than new batteries but reliability becomes an important issue as individual batteries may suffer from degraded performance or failure. Therefore converter topology design could be used to influence the overall system reliability. A detailed reliability calculation of different single phase battery-to-grid converter interfacing schemes is presented. A suitable converter topology for robust and reliable BESS is recommended.

Freeway travel time estimation and prediction using dynamic neural networks

Providing transportation system operators and travelers with accurate travel time information allows them to make more informed decisions, yielding benefits for individual travelers and for the entire transportation system. Most existing advanced traveler information systems (ATIS) and advanced traffic management systems (ATMS) use instantaneous travel time values estimated based on the current measurements, assuming that traffic conditions remain constant in the near future. For more effective applications, it has been proposed that ATIS and ATMS should use travel times predicted for short-term future conditions rather than instantaneous travel times measured or estimated for current conditions. ^ This dissertation research investigates short-term freeway travel time prediction using Dynamic Neural Networks (DNN) based on traffic detector data collected by radar traffic detectors installed along a freeway corridor. DNN comprises a class of neural networks that are particularly suitable for predicting variables like travel time, but has not been adequately investigated for this purpose. Before this investigation, it was necessary to identifying methods for data imputation to account for missing data usually encountered when collecting data using traffic detectors. It was also necessary to identify a method to estimate the travel time on the freeway corridor based on data collected using point traffic detectors. A new travel time estimation method referred to as the Piecewise Constant Acceleration Based (PCAB) method was developed and compared with other methods reported in the literatures. The results show that one of the simple travel time estimation methods (the average speed method) can work as well as the PCAB method, and both of them out-perform other methods. This study also compared the travel time prediction performance of three different DNN topologies with different memory setups. The results show that one DNN topology (the time-delay neural networks) out-performs the other two DNN topologies for the investigated prediction problem. This topology also performs slightly better than the simple multilayer perceptron (MLP) neural network topology that has been used in a number of previous studies for travel time prediction.^

Freeway Travel Time Estimation and Prediction Using Dynamic Neural Networks

Providing transportation system operators and travelers with accurate travel time information allows them to make more informed decisions, yielding benefits for individual travelers and for the entire transportation system. Most existing advanced traveler information systems (ATIS) and advanced traffic management systems (ATMS) use instantaneous travel time values estimated based on the current measurements, assuming that traffic conditions remain constant in the near future. For more effective applications, it has been proposed that ATIS and ATMS should use travel times predicted for short-term future conditions rather than instantaneous travel times measured or estimated for current conditions. This dissertation research investigates short-term freeway travel time prediction using Dynamic Neural Networks (DNN) based on traffic detector data collected by radar traffic detectors installed along a freeway corridor. DNN comprises a class of neural networks that are particularly suitable for predicting variables like travel time, but has not been adequately investigated for this purpose. Before this investigation, it was necessary to identifying methods for data imputation to account for missing data usually encountered when collecting data using traffic detectors. It was also necessary to identify a method to estimate the travel time on the freeway corridor based on data collected using point traffic detectors. A new travel time estimation method referred to as the Piecewise Constant Acceleration Based (PCAB) method was developed and compared with other methods reported in the literatures. The results show that one of the simple travel time estimation methods (the average speed method) can work as well as the PCAB method, and both of them out-perform other methods. This study also compared the travel time prediction performance of three different DNN topologies with different memory setups. The results show that one DNN topology (the time-delay neural networks) out-performs the other two DNN topologies for the investigated prediction problem. This topology also performs slightly better than the simple multilayer perceptron (MLP) neural network topology that has been used in a number of previous studies for travel time prediction.