Inverse problems associated with reservoir characterization are typically underdetermined and often have difficulties associated with stability and convergence of the solution. A common approach to address this issue is through the introduction of prior constraints, regularization or reparameterization to reduce the number of estimated parameters. We propose a dual scale approach to production data integration that relies on a combination of coarse-scale and fine-scale inversions while preserving the essential features of the geologic model. To begin with, we sequentially coarsen the fine-scale geological model by grouping layers in such a way that the heterogeneity measure of an appropriately defined 'static' property is minimized within the layers and maximized between the layers. Our coarsening algorithm results in a non-uniform coarsening of the geologic model with minimal loss of heterogeneity and the ?optimal? number of layers is determined based on a bias-variance trade-off criterion. The coarse-scale model is then updated using production data via a generalized travel time inversion. The coarse-scale inversion proceeds much faster compared to a direct fine-scale inversion because of the significantly reduced parameter space. Furthermore, the iterative minimization is much more effective because at the larger scales there are fewer local minima and those tend to be farther apart. At the end of the coarse-scale inversion, a fine-scale inversion may be carried out, if needed. This constitutes the outer iteration in the overall algorithm. The fine-scale inversion is carried out only if the data misfit is deemed to be unsatisfactory. We propose a fast and robust approach to calibrating geologic models by transient pressure data using a trajectory-based approach that based on a high frequency asymptotic expansion of the diffusivity equation. The trajectory or ray-based methods are routinely used in seismic tomography. In this work, we investigate seismic rays and compare them with streamlines. We then examine the applicability of streamline-based methods for transient pressure data inversion. Specifically, the high frequency asymptotic approach allows us to analytically compute the sensitivity of the pressure responses with respect to reservoir properties such as porosity and permeability. It facilitates a very efficient methodology for the integration of pressure data into geologic models.

The demand for aviation has been steadily growing over the past few decades and will keep increasing in the future. The anticipated growth of traffic demand will cause the current airspace system, one that is already burdened by heavy operations and inefficient usage, to become even more congested than its current state. Because busy airports in the United States (U.S.) are becoming "bottlenecks" of the National Airspace System (NAS), it is of great importance to discover the most efficient means of using existing facilities to improve airport operations. This dissertation aims at designing an efficient airport surface operations management system that substantially contributes to the modernized NAS. First, a global comparison is conducted in the major airports within the U.S. and Europe in order to understand, compare, and explore the differences of surface operational efficiency in two systems. The comparison results are then presented for each airport pair with respect to various operational performance metrics, as well as airport capacity and different demand patterns. A detailed summary of the associated Air Traffic Management (ATM) strategies that are implemented in the U.S. and Europe can be found towards the end of this work. These strategies include: a single Air Navigation Service Provider (ANSP) in the U.S. and multiple ANSPs in Europe, airline scheduling and demand management differences, mixed usage of Instrument Flight Rule (IFR) and Visual Flight Rules (VFR) operations in the U.S., and varying gate management policies in two regions. For global comparison, unimpeded taxi time is the reference time used for measuring taxi performance. It has been noted that different methodologies are currently used to benchmark taxi times by the performance analysis groups in the U.S. and Europe, namely the Federal Aviation Authority (FAA) and EUROCONTROL. The consistent methodology to measure taxi efficiency is needed for the facilitation of global benchmarking. Therefore, after an in-depth factual comparison conducted for two varying methodologies, new methods to measure unimpeded taxi times are explored through various tools, including simulation software and projection of historical surveillance data. Moreover, a sophisticated statistical model is proposed as a state-of-the-art method to measure taxi efficiency while quantifying the impact of various factors to taxi inefficiency and supporting decision-makers with reliable measurements to improve the operational performance. Lastly, a real-time integrated airport surface operations management (RTI-ASOM) is presented to fulfil the third objective of this dissertation. It provides optimal trajectories for each aircraft between gates and runways with the objective of minimizing taxi delay and maximizing runway throughput. The use of Mixed Integer Linear Programming (MIP) formulation, Dynamic Programming for decomposition, and CPLEX optimization can permit the use of an efficient solution algorithm that can instantly solve the large-scale optimization problem. Examples are shown based on one-day track data at LaGuardia Airport (LGA) in New York City. In additional to base scenarios with historical data, simulation through MATLAB is constructed to provide further comparable scenarios, which can demonstrate a significant reduction of taxi times and improvement of runway utilization in RTI-ASOM. By strategically holding departures at gates, the application of RTI-ASOM also reduces excess delay on the airport surface, decreases fuel consumption at airports, and mitigates the consequential environmental impacts.

Continuous Descent Arrival (CDA) procedures have been shown to minimize the thrust required during landing, thereby reducing noise, emissions, and fuel usage for commercial aircraft. Thus, implementation of CDA at Atlanta's Hartsfield-Jackson International Airport, the world's busiest airport, would result in significant reductions in environmental impact and airline operating costs. The Air Transportation Laboratory at Georgia Tech, Delta Air Lines, and the local FAA facilities (Atlanta Center and Atlanta TRACON) collaborated to design CDA procedures for early morning arrivals from the west coast. Using the Tool for Analysis of Separation and Throughput (TASAT), we analyzed the performance of various aircraft types over a wide range of weights and wind conditions to determine the optimum descent profile parameters and to find the required spacing between aircraft types at a fixed metering point to implement the procedure. However, to see the full benefits of CDA, these spacing targets must be adhered, lest there will be a loss in capacity or negation of the noise, emissions, and fuel savings benefits. Thus a method was developed to determine adjustments to cruise speeds while aircraft are still en route, to achieve these spacing targets and to optimize fleet wide fuel burn increase. The tool in development, En route Speed Change Optimization Relay Tool (ESCORT), has been shown to solve the speed change problem quickly, incorporating aircraft fuel burn information and dividing the speed changes fairly across multiple airlines. The details of this tool will be explained in this thesis defense. Flight tests were conducted in April-May of 2007, where it was observed that the spacing targets developed by TASAT were accurate but that delivery of these aircraft to the metering point with the desired spacing targets was very challenging without automation. Thus, further flight tests will be conducted in 2008 using the en route spacing tool described above to validate the improvement it provides in terms of accurately delivering aircraft to the metering point.

La congestion dans les Terminal Manoeuvring Area (TMA) des aéroports en hub est le principal problème dans le transport aérien chinois. Face au trafic extrêmement dense dans les TMAs, nous pouvons envisager d'automatiser une grande partie des opérations de routine, comprenant la planification, le séquencement et la séparation. Nous proposons dans cette thèse un nouveau système automatisé de séquencement des avions et de fusion des flux vers des pistes parallèles, qui sont utilisées dans la plupart des aéroports chinois. Notre méthodologie intègre un réseau de route 3D nommé Multi-Level and Multi-Point Merge System (MLMPMS) basé sur le concept de l'Area Navigation (RNAV) et un algorithme d'optimisation heuristique hybride pour trouver une solution correcte, opérationnellement acceptable. Un algorithme de Simulated Annealing (SA) spécifique et un module de génération de trajectoire collaborent pour rechercher la solution quasi optimale. Notre objectif est de générer en temps réel des trajectoires sans conflit, minimisant la consommation de carburant et permettant des méthodes de contrôle faciles et flexibles. Dans ce but, nous explorons en permanence les solutions possibles avec le moins de retard et assuront l'atterrissage le plus rapide. Nous déterminons quatre variables de décision pour contrôler chaque vol : l'heure et la vitesse d'entrée dans la TMA, le temps de vol sur l'arc de séquencement et le choix de la piste utilisée. La simulation de trajectoire dans les différentes phases de vol est basée sur le modèle de performances BADA. Dans le cas de l'aéroport de Beijing Capital International Airport (BCIA), les résultats numériques montrent que notre système d'optimisation de bonnes performances sur le séquencement et la fusion des trajectoires. Tout d'abord, il permet d'assurer des performances de résolution conflit très stables pour gérer les flux de trafic continuellement denses. Par rapport à l'algorithme Hill Climbing (HC), le SA peut toujours trouver une solution sans conflit, non seulement pour l'approche parallèle mixte ou séparée~(pour les arrivées), mais aussi pour les configurations parallèles indépendantes~ (départs et arrivées intégrés). Ensuite, avec un réseau d'itinéraires Multi-Level Point Merge (ML-PM) unique, il peut fournir une bonne solution de contrôle de la trajectoire pour traiter efficacement et économiquement différents types de flux d'arrivée. Il peut réaliser un temps de vol plus court et une descente vers le bas en Continuous Descent Approach (CDA) pour l'avion d'arrivée. Il peut réaliser un re-séquencement plus facile des avions avec un déplacement de position plus relâché. Théoriquement, les Maximum Position Shifting (MPS) peuvent atteindre 6 positions, surpassant la contrainte difficile de 3 positions. Troisièmement, l'algorithme montre son efficacité dans un modèle d'approche parallèle séparé avec une capacité de séquencement plus relâché. Par rapport au décalage de position forcé dur, qui est souvent utilisé dans le système actuel Arrival Manager (AMAN) et la méthode First Come First Served (FCFS) utilisé par les contrôleurs, il peut réduire le délai et le temps de transit moyens dans une situation d'arrivée très dense. Le palier par vol est inférieur à 12\% du temps de transit total dans la TMA. Quatrièmement, en configuration parallèle indépendant, il peut fournir des informations différentes concernant la valeur objectif associée, le temps de vol moyen, les trajectoires de croisement en point chaud entre les arrivées et les départs, l'efficacité avec différents arcs de séquencement conçus dans le réseau de route ML-PM etc.. / Congestion in Terminal Manoeuvring Area (TMA) at hub airports is the main problem in Chinese air transportation system. Facing extremely dense operations in complex TMA, we can consider reducing traffic complexity by solving all potential conflicts in advance with a feasible trajectory control for controllers, or automating a large proportion of routine operations, such as sequencing, merging and spacing. As parallel runways are a common structure of Chinese hub airports, in this thesis, we propose a novel system to integrated sequencing and merging aircraft to parallel runways. Our methodology integrates a Area Navigation (RNAV)-based 3D Multi-Level and Multi-Point Merge System (MLMPMS), a hybrid heuristic optimization algorithm and a simulation module to find good, systematic, operationally-acceptable solutions. First, a Receding Horizon Control (RHC) technique is applied to divide 24-hour traffic optimization problem into several sub- problems. Then, in each sub-problem, a tailored Simulated Annealing (SA) algorithm and a trajectory generation module worn together to find a near-optimal solution. Our primary objective is to rapidly generate conflict-free and economical trajectories with easy, flexible and feasible control methods. Based on an initial solution, we continuously explore possible good solutions with less delay and shorter landing interval on runway. Taking Beijing Capital International Airport (BCIA) as a case to study, numerical results show that our optimization system performs well. First, it has very stable de-conflict performance to handle continuously dense traffic flows. Compared with Hill Climbing (HC), the tailored SA algorithm can always guarantee a conflict-free solution not only for the mixed or segregated parallel approach (arrivals only) pattern, but also for the independent parallel operation (integrated departures and arrivals) pattern. Second, with its unique Multi-Level Point Merge (ML-PM) route network, it can provide a good trajectory control solution to efficiently and economically handle different kinds of arrival flows. It can realize a shorter flying time and a near-Continuous Descent Approach (CDA) descent for arrival aircraft, compared with baseline. It also realizes an easier re-sequencing of aircraft with more relaxed position shifting as well, compared with conventional sequencing method. Theoretically, the Maximum Position Shifting (MPS) can be up to 6 positions, overcoming the hard constraint of 3 position shifts (MPS <= 3). Third, it is efficient for the segregated parallel approach patterns. Compared with hard constrained position shifting, which is often used in current Arrival Manager (AMAN) system and controller's manual-control First Come First Served (FCFS) method, it can reduce the average delay, average additional transit time in super dense arrival situations. The average time flown level per flight is less than 12% of total transit time in TMA. Fourth, in independent parallel patterns, it can provide a range of useful information concerning the associated objective value, the average flying time, crossing trajectories in hot spots between arrivals and departures, the efficiency of using different designed sequencing legs in ML-PM route network.

This thesis explores motion trajectory-based approaches to recognize human actions in real-world, unconstrained videos. Recognizing actions is an important task in applications such as video retrieval, surveillance, human-robot interactions, analysis of sports videos, summarization of videos, behaviour monitoring, etc. There has been a considerable amount of research done in this regard. Earlier work used to be on videos captured by static cameras where it was relatively easy to recognise the actions. With more videos being captured by moving cameras, recognition of actions in such videos with irregular camera motion is still a challenge in unconstrained settings with variations in scale, view, illumination, occlusion and unrelated motions in the background. With the increase in videos being captured from wearable or head-mounted cameras, recognizing actions in egocentric videos is also explored in this thesis. At first, an effective motion segmentation method to identify the camera motion in videos captured by moving cameras is explored. Next, action recognition in videos captured in normal third-person view (perspective) is discussed. Further, the action recognition approaches for first-person (egocentric) views are investigated. First-person videos are often associated with frequent unintended camera motion. This is due to the motion of the head resulting in the motion of the head-mounted cameras (wearable cameras). This is followed by recognition of actions in egocentric videos in a multicamera setting. And lastly, novel feature encoding and subvolume sampling (for “deep” approaches) techniques are explored in the context of action recognition in videos. The first part of the thesis explores two effective segmentation approaches to identify the motion due to camera. The first approach is based on curve fitting of the motion trajectories and finding the model which best fits the camera motion model. The curve fitting approach works when the trajectories generated are smooth enough. To overcome this drawback and segment trajectories under non-smooth conditions, a second approach based on trajectory scoring and grouping is proposed. By identifying the instantaneous dominant background motion and accordingly aggregating the scores (denoting the “foregroundness”) along the trajectory, the motion that is associated with the camera can be separated from the motion due to foreground objects. Additionally, the segmentation result has been used to align videos from moving cameras, resulting in videos that seem to be captured by nearly-static cameras. In the second part of the thesis, recognising actions in normal videos captured from third-person cameras is investigated. To this end, two kinds of descriptors are explored. The first descriptor is the covariance descriptor adapted for the motion trajectories. The covariance descriptor for a trajectory encodes the co-variations of different features along the trajectory’s length. Covariance, being a second-order encoding, encodes information of the trajectory that is different from that of the first-order encoding. The second descriptor is based on Granger causality. The novel causality descriptor encodes the “cause and effect” relationships between the motion trajectories of the actions. This type of interaction descriptors captures the causal inter-dependencies among the motion trajectories and encodes complimentary information different from those descriptors based on the occurrence of features. The causal dependencies are traditionally computed on time-varying signals. We extend it further to capture dependencies between spatiotemporal signals and compute generalised causality descriptors which perform better than their traditional counterparts. An egocentric or first-person video is captured from the perspective of the personof-interest (POI). The POI wears a camera and moves around doing his/her activities. This camera records the events and activities as seen by him/her. The POI who is performing actions or activities is not seen by the camera worn by him/her. Activities performed by the POI are called first-person actions and third-person actions are those done by others and observed by the POI. The third part of the thesis explores action recognition in egocentric videos. Differentiating first-person and third-person actions is important when summarising/analysing the behaviour of the POI. Thus, the goal is to recognise the action and the perspective from which it is being observed. Trajectory descriptors are adapted to recognise actions along with the motion trajectory ranking method of segmentation as pre-processing step to identify the camera motion. The motion segmentation step is necessary to remove unintended head motion (camera motion) during video capture. To recognise actions and corresponding perspectives in a multi-camera setup, a novel inter-view causality descriptor based on the causal dependencies between trajectories in different views is explored. Since this is a new problem being addressed, two first-person datasets are created with eight actions in third-person and first-person perspectives. The first dataset is a single camera dataset with action instances from first-person and third-person views. The second dataset is a multi-camera dataset with each action instance having multiple first-person and third-person views. In the final part of the thesis, a feature encoding scheme and a subvolume sampling scheme for recognising actions in videos is proposed. The proposed Hyper-Fisher Vector feature encoding is based on embedding the Bag-of-Words encoding into the Fisher Vector encoding. The resulting encoding is simple, effective and improves the classification performance over the state-of-the-art techniques. This encoding can be used in place of the traditional Fisher Vector encoding in other recognition approaches. The proposed subvolume sampling scheme, used to generate second layer features in “deep” approaches for action recognition in videos, is based on iteratively increasing the size of the valid subvolumes in the temporal direction to generate newer subvolumes. The proposed sampling requires lesser number of subvolumes to be generated to “better represent” the actions and thus, is less computationally intensive compared to the original sampling scheme. The techniques are evaluated on large-scale, challenging, publicly available datasets. The Hyper-Fisher Vector combined with the proposed sampling scheme perform better than the state-of-the-art techniques for action classification in videos.

As drones are becoming ubiquitous, robust detection and tracking of potentially hostile drones is becoming a necessity. Among the many approaches being investigated in this relatively new research area, one cost effective option is the use of optical cameras equipped with computer vision algorithms. With the use of deep learning, it becomes possible to achieve high accuracy by generalizing from large datasets. However, drones are small and visually similar to birds, which has proven to be a major difficulty for purely vision based systems. This thesis investigates the utility of trajectory information (velocity and acceleration) in addition to temporal appearance features for detection and tracking of drones. While both kinds of information has been used in a variation of ways, work combining the two is largely lacking. Our approach uses background subtraction to generate candidate objects that initialize an LSTM which in turn combines trajectory and appearance information over multiple frames for joint detection and tracking of drones. While our specific implementation fails to outperform a traditional object detector in the form of YOLOv8, this could change with the solution of two problems identified with our approach. First problem being how to effectively incorporate large amounts of background data into the training of our network. Second being how to avoid repeatedly proposing the same non-drone candidates, while still being able to quickly resume tracking of a lost drone. / I takt med att drönare blir allt vanligare stiger kraven på robusta system som kan upptäcka och spåra hotfulla drönare. Bland de flertal tillvägagångssätt som undersöks i detta relativt nya forskningsområde är användandet av optiska kameror utrustade med datorseende-algoritmer ett kostnadseffektivt val. Genom användningen av djupinlärning har det blivit möjligt att uppnå hög pricksäkerhet genom att generalisera utifrån stora dataset. Men, drönare är små och utseendemässigt sett lika fåglar vilket är ett svåröverkomligt problem för system som endast förlitar sig på datorseende. I detta examensarbete undersöks vilken nytta som kan fås om man även tar hänsyn till information om drönarens trajektoria i form av hastighet och acceleration. Trots att både visuellt utseende och trajektoria är välstuderat när det kommer till drönardetektering, saknas det till stor del forskning som behandlar båda tillsammans. Vi använder bakgrundssubtraktion för att generera kandidater som startpunkt för en LSTM för att sedan kombinera trajektoria med utseende för förenad detektering och spårning av drönare. Fastän vår specifika implementation inte lyckas överträffa en traditionell objektdetekterare i form av YOLOv8, skulle detta kunna ändras givet en lösning på två identifierade problem med vårt tillvägagångssätt. Det första problemet är att hitta ett effektivt sätt att inkorporera stora mängder bakgrundsdata i träningen av vårt nätverk. Det andra är att undvika att gång på gång föreslå samma kandidater och samtidigt kunna snabbt återuppta spårningen av en förlorad drönare.

Not only is Quantum Mechanics (QM) conceptually rich, it is also a theory that physics students have found abstract and technically formidable. Nevertheless, compared to other classical topics of physics, university students’ understanding of QM has received minimal attention in the physics education literature. The principal purpose of this study was to characterize the variation in the ways that undergraduate physics students depict the basic concepts of QM and to extrapolate the results to scaffold possible changes to instructional practices at the university that provided the context for the study. In so doing, an adaptation of a developmental phenomenographic perspective was chosen. Empirically, the study was approached through in-depth interviews with 35 physics students from two Ethiopian governmental universities after they had been exposed to the traditional QM course for one-third of a semester. Interview responses were analyzed using phenomenographic approach where a picture of students’ depictions was established for each quantum concept by expounding the given responses. For each basic quantum concept addressed, the structure of the description categories was separately constructed, and overall, it was found that naive, quasi-classical ontology and/or variants of classical ways of visualization are dominant in students’ responses. For example, it was found that students’ depictions of the photon concept could be described with three distinct categories of description, which are (a) classical intuitive description, (b) mixed model description and (c) quasi-quantum model description. Similarly, the findings revealed that it is possible to establish three qualitatively different categories of description to picture students’ depictions of matter waves, namely, (a) classical and trajectory-based description, (b) an intricate blend of classical and quantum description and (c) incipient quantum model description. Likewise, it was found that students’ depictions of uncertainty principle can be described as: (a) uncertainty as classical ignorance, (b) uncertainty as measurement disturbance and (c) uncertainty as a quasi-quantum principle. With regard to learning QM, the categories of description made clear several issues: most students did not have enough knowledge to depict the basic concepts of QM properly; they were influenced by the perspective of classical physics and their perceptions in making explanations about QM; and they also applied mixed ideas, one based on their classical model and the other from newly introduced QM. These results are also supported by the findings of previous studies in similar domains. Findings from the study were used to guide the design of multiple representations-based instructions and interactive learning tutorials on the conceptual aspects of QM that has been shown to address specific difficulties identified in the study. Theoretical and practical implications of the study, as well as potential future considerations are drawn. / Mathematics, Science and Technology Education / D. Phil. (Mathematics, Science and Technology Education)

Not only is Quantum Mechanics (QM) conceptually rich, it is also a theory that physics students have found abstract and technically formidable. Nevertheless, compared to other classical topics of physics, university students’ understanding of QM has received minimal attention in the physics education literature. The principal purpose of this study was to characterize the variation in the ways that undergraduate physics students depict the basic concepts of QM and to extrapolate the results to scaffold possible changes to instructional practices at the university that provided the context for the study. In so doing, an adaptation of a developmental phenomenographic perspective was chosen. Empirically, the study was approached through in-depth interviews with 35 physics students from two Ethiopian governmental universities after they had been exposed to the traditional QM course for one-third of a semester. Interview responses were analyzed using phenomenographic approach where a picture of students’ depictions was established for each quantum concept by expounding the given responses. For each basic quantum concept addressed, the structure of the description categories was separately constructed, and overall, it was found that naive, quasi-classical ontology and/or variants of classical ways of visualization are dominant in students’ responses. For example, it was found that students’ depictions of the photon concept could be described with three distinct categories of description, which are (a) classical intuitive description, (b) mixed model description and (c) quasi-quantum model description. Similarly, the findings revealed that it is possible to establish three qualitatively different categories of description to picture students’ depictions of matter waves, namely, (a) classical and trajectory-based description, (b) an intricate blend of classical and quantum description and (c) incipient quantum model description. Likewise, it was found that students’ depictions of uncertainty principle can be described as: (a) uncertainty as classical ignorance, (b) uncertainty as measurement disturbance and (c) uncertainty as a quasi-quantum principle. With regard to learning QM, the categories of description made clear several issues: most students did not have enough knowledge to depict the basic concepts of QM properly; they were influenced by the perspective of classical physics and their perceptions in making explanations about QM; and they also applied mixed ideas, one based on their classical model and the other from newly introduced QM. These results are also supported by the findings of previous studies in similar domains. Findings from the study were used to guide the design of multiple representations-based instructions and interactive learning tutorials on the conceptual aspects of QM that has been shown to address specific difficulties identified in the study. Theoretical and practical implications of the study, as well as potential future considerations are drawn. / Mathematics, Science and Technology Education / D. Phil. (Mathematics, Science and Technology Education)