<p>Our main results are that splay trees are faster for sorted insertion, where AVL trees are faster for random insertion. For searching, skip lists are faster than single class top-down splay trees, but two-class and multi-class top-down splay trees can behave better than skip lists.</p>

Magister Scientiae - MSc / Our main results are that splay trees are faster for sorted insertion, where AVL trees are faster for random insertion. For searching, skip lists are faster than single class top-down splay trees, but two-class and multi-class top-down splay trees can behave better than skip lists. / South Africa

No description available.

With the recent development of reliable autonomous technologies for small unmanned air vehicles (UAVs), the algorithms utilizing teams of these vehicles are becoming an increasingly important research area. Unfortunately, there is no unified framework into which all (or even most) cooperative control problems fall. Five factors that affect the development of cooperative control algorithms are objective coupling, communication, completeness, robustness, and efficiency. We classify cooperative control algorithms by these factors and then present three algorithms with application to target and perimeter surveillance and a method for decentralized algorithm design. The primary contributions of this research are the development and analysis of decentralized algorithms for perimeter and target surveillance. We pose the cooperative perimeter surveillance problem and offer a decentralized solution that accounts for perimeter growth (expanding or contracting) and insertion/deletion of team members. By identifying and sharing the critical coordination information and by exploiting the known communication topology, only a small communication range is required for accurate performance. Convergence of the algorithm to the optimal configuration is proven to occur in finite-time. Simulation and hardware results are presented that demonstrate the applicability of the solution. For single target surveillance, a team of UAVs angularly spaced (i.e. in the splay state configuration) provides the best coverage of the target in a wide variety of circumstances. We propose a decentralized algorithm to achieve the splay state configuration for a team of UAVs tracking a moving target and derive the allowable bounds on target velocity to generate a feasible solution as well as show that, near equilibrium, the overall system is exponentially stable. Monte Carlo simulations indicate that the surveillance algorithm is asymptotically stable for arbitrary initial conditions. We conclude with high fidelity simulation and actual flight tests to show the applicability of the splay state controller to unmanned air systems.

Deep-marine sandstones containing significant (> 10%) detrital mud (silt and clay) matrix have become increasingly recognized, but mostly in drill core or poorly exposed outcrops where details of their vertical and lateral variability are poorly captured. Exceptional vertical and along-strike exposures of matrix-rich and associated matrix-poor deposits in deep-marine strata of the passive margin Neoproterozoic Windermere Supergroup and foreland basin Ordovician Cloridorme Formation, provide an unparalleled opportunity to document such characteristics. In both study areas, strata form a 100s m long depositional continuum that at its upflow end consists of thick-bedded matrix-poor sandstone (<20% matrix) that transforms progressively downflow to medium- to thick-bedded muddy sandstone (20 – 50% matrix) to medium-bedded bipartite facies with a basal sandy (30 – 60% matrix) part overlain sharply by a muddier part (40 – 80% matrix), and then to thin-bedded sandy mudstone (50 – 90% matrix). This depositional continuum is then overlain everywhere by a thin- to very thin-bedded traction-structured sandstone and/or silty mudstone cap. This consistent lithofacies change is interpreted to reflect particle settling in a rapidly but systematically evolving, negligibly-sheared sand-mud suspension developed along the margins (Windermere) and downflow terminus (Cloridorme) of a high-energy, mud-enriched avulsion jet. Stratigraphically upward, beds of similar lithofacies type succeed one another vertically and transform to the next facies in the depositional continuum at about the same along-strike position, forming stratal units 2–9 beds thick whose grain-size distribution gradually decreases upward. This spatial and temporal regularity is interpreted to be caused by multiple surges of a single, progressively waning turbidity current, with sufficient lag between successive surges for the deposition of a traction-structured sandstone overlain by mudstone cap. Furthermore, the systematic backstepping or side-stepping recognized at the stratal unit scale in both the Windermere and Cloridorme is interpreted to be driven by a combination of knickpoint migration and local topographic steering of the flows, which continued until the supply of mud from local seafloor erosion became exhausted, the main channel avulsed elsewhere, or a new stratal element developed.

No description available.

First distinguished from other sedimentary successions in 1928, the Entrada Sandstone has been the subject of numerous studies. The western extent of the formation was initially described as laterally continuous "earthy" red beds, and categorized as sub- to supratidal marine-influenced sediments. Recent workers have reexamined the sedimentary facies hosted by the Entrada Sandstone, and findings suggest purely terrestrial depositional environments. Several outcrops of the upper Entrada hosted peculiar sedimentary features, including undulatory and convex-upward, parallel-laminated bedforms, reminiscent of hummocky cross-stratification- unexpected features in a terrestrial environment. The purpose of this study was to collect detailed outcrop measurements of these and other facies present in the upper Entrada Sandstone and to place them in context within a regional sedimentary system. Measured section data was analyzed and divided into sixteen primary facies based on textures, features, bedforms, grain size, and other characteristics. Surfaces were also noted and described. Each facies and surface was recognized to have developed under specific depositional or flow conditions, including eolian, paleosol, and fluvial subcritical, critical, supercritical, and waning flow. Primary facies were grouped into observed and interpreted facies associations. A depositional environment was then assigned to each facies association. These environments included sabkha, overbank splay/paleosol, distal terminal splay, and hyper-distal terminal splay. Ancient analogs were found in the Blomidon, Skagerrak, and Ormskirk Formations, which have been described as dryland fluvial systems that terminated onto saline mudflats (sabkhas). Modern analogs were found in the central Australian continent, in the form of fluvial terminal splays in ephemeral Lakes Eyre and Frome. The sedimentary system of the upper Entrada Sandstone of the San Rafael Swell is interpreted as an interfingering fluvial terminal splay and inland sabkha system. These are marked by a wide array of sedimentary structures representing stark extremes, from hyperarid to flash flooding conditions. During arid conditions, the only source of water was evaporative pumping of a high water table. During the rare occasions when surface water flowed through the system, flash flooding events produced the highest stage of supercritical flow described in geological literature. The succession of these facies reveals allogenic and autogenic processes active at the time of deposition, including episodes of tectonic uplift and fluvial avulsions.

The Permo-Triassic lower Beaufort Group fluvial deposits extend over 100s of kilometres within the Karoo Basin, South Africa. A detailed study of the depositional architecture and stacking patterns of sand bodies within a 900 m thick succession has enabled interpretation of the controls on ancient river channel and overbank processes. Facies include very fine- to medium-grained sandstone, intra-formational conglomerate, mudstone and palaeosols. Channel-belts are dominated by upper flow regime structures, consistent with a flashy to ephemeral fluvial system. The overbank deposits comprise splays interbedded with purple, green and grey mudstone; these floodplain colour changes signify water table fluctuations. A hierarchy of channel-related elements has been established that recognises beds, bedsets, storeys, channel-belts, complexes and complex sets. Each channel-belt may be single- or multi-storey, whereby one storey represents the complete cut and fill cycle of a single migrating river, comprising bar accretion elements and channel-abandonment fill. The abandonment fill elements often consist of heterolithic plugs of climbing ripple-laminated very fine-grained sandstone, or interbedded claystone with siltstone. The Beaufort channel-belts preserve either lateral- or downstream-accretion patterns, or a combination. Each belt has either a lenticular or tabular geometry, recognisable by an erosional base overlain by intra-formational conglomerate lag and barform deposits. Genetically related channel-belts cluster to form complexes, of which two broad styles have been identified: Type A) laterally and vertically stacked channel-belts, and Type B) sub-vertically stacked channel-belts. There is evidence of localised clustering of sub-vertically stacked channel-belts adjacent to extensive overbank mudstone deposits. The apparent lack of a well-defined ‘container’ surface with mappable margins, suggests that this stacked channel-belt architecture represents an avulsion complex rather than a palaeovalley-fill. The lateral and stratigraphic variability in fluvial-overbank architecture is interpreted as the interplay of several controls. Allogenic forcing factors include, tectonic subsidence that influences accommodation, sediment supply, and high frequency climate cycles associated with the flashy discharge regime and expressed in the mudrock colour changes and distribution of palaeosols. The depositional river style, variability in channel-belt stacking patterns and compensational stacking of some channel-belt/splay complexes is interpreted to be the result of autogenic channel avulsion, supported by an absence of significant erosion. The relative merits of basin-axial trunk river and distributive fluvial system (DFS) models are assessed from detailed architectural and stratigraphic outcrop studies.

Mon travail de thèse se propose d'étudier la structure, les propriétés physiques et les processus géodynamiques de la zone de subduction d'Equateur-Colombie grâce à l'adaptation et le développement d'outils d'imagerie sismique (inversion de formes d'ondes 'alias' tomographie en diffraction) et à leur application aux données de sismique marine multitrace (MCS) et grand-angle OBS (WA) acquises en Equateur-Colombie pendant les campagnes SISTEUR et SALIERI. Ces outils m'ont permis de réaliser une imagerie fine et quantitative à trois niveaux : l'imagerie superficielle (~ 0-3 km), l'imagerie à profondeur intermédiaire (~ 3-10 km) et l'imagerie profonde (~ 10-30 km). Dans le domaine superficiel, j'ai effectué une cartographie fine et quantitative des propriétés physiques des sédiments au voisinage du BSR (Bottom Simulating Reflector), interpreté comme la base de stabilité des hydrates de gaz. Sur le profil SIS-40 situé sur la marge sud de la Colombie, j'ai pu identifier la présence de failles qui perturbent localement le BSR. Les résultats présentés sous la forme d' une série de logs adjacents de l'image migrée en profondeur, montrent que certaines régions du BSR sont caractérisées par une augmentation de la vitesse (1470-1650 m/s), indiquant la présence d'une faible quantité d'hydrates de gaz au dessus du BSR; d'autres zones situées immédiatement sous le BSR sont caractérisées par une diminution de la vitesse (~1200 m/s), liée à la présence de gaz libres piégés sous la couche d'hydrate de gaz. A des profondeurs moyennes j'ai étudié la structure du chenal de subduction (profil SIS-72). Le chenal constitue la limite mécanique entre la plaque chevauchante et la plaque plongeante. Il est délimité à son toit par un fort réflecteur interprété comme le décollement interplaque et à sa base par le toit très réflectif de la croûte océanique en subduction. L'imagerie fine et quantitative des propriétés physiques du décollement interplaque permet de mieux comprendre le rôle de la circulation des fluides et des variations lithologiques et physiques, sur le couplage mécanique inter-plaque. En raison de la sensibilité de la méthode de tomographie en diffraction au macro-modèle de vitesse, un code de correction de ce modèle a été implémenté, afin d'obtenir des images tomographiques fiables (i.e. géométrie et amplitudes correctes). Du fait de la bande passante limitée de la source et de la longueur du dispositif d'acquisition limitée à 4.5 km, les images tomographiques ont une résolution spatiale limitée : l'image tomographique présente un déficit des petits et grands nombre d'onde (fréquences spatiales) limitant ainsi l' interprétation géologique des paramètres physiques cartographiés. Un traitement spécifique basé sur la modélisation des traces sismiques a été implémenté. L'image tomographique, traitée comme une série de traces verticales, constitue la donnée observée. L'espace des modèles est constitué par un ensemble de modèles impulsionnels et unidimensionnels de Terre construits aléatoirement. Ces modèles sont dégradés par convolution avec une estimation de l'ondelette source afin de fournir une représentation synthétique de l'image tomographique « observée ». La minimisation de la fonction coût entre les traces migrées et les traces synthétiques est effectuée dans le cadre d'une inversion globale par recuit simulé (VFSA= « Very Fast Simulated Annealing »). Le modèle moyen issu de cette procédure fournit un modèle 2D fin de vitesse, fonction de la profondeur et comparable à la limite de la résolution théorique de la source. A l'issue de ce traitement, des perturbations de vitesse positives sont mises en évidence au toit de la croûte, et d'autres négatives accompagnent certains segments du niveau du décollement. Ces dernières sont probablement associées à la présence de fluides. Le domaine plus profond a été étudié à partir des données MCS et WA dans le double but (1) d'améliorer la résolution spatiale des images sismiques du Moho et du contact interplaque en relation avec la zone sismogène, et (2) de détecter la présence d'anomalies crustales de vitesse et d'analyser leur relation avec les zones d'aspérité sismologiques. L'utilisation conjointe des données de sismique MCS et WA a été mise en oeuvre pour prolonger vers le bas les images de sismique verticale et tenter ainsi d'établir une relation entre les processus profonds et les manifestations en surface. L'application de la chaîne de traitement au profil SIS-44 a permis d'obtenir un modèle de vitesse bien contraint jusqu'à 25 km de profondeur. Ce modèle met en évidence des réflecteurs profonds (Moho et contact interplaque ) et des réflecteurs plus superficiels (splay fault), dont l'interprétation était initialement incertaine sur les images migrées en temps.