published_or_final_version / Botany / Master / Master of Philosophy

This dissertation deals with collapsing soils that are prevalent in Tucson, Arizona. Upon wetting, such soils generally swell under small loads but collapse under large loads. Since the recognition of such collapsing soils in Tucson, before about two decades, more collapsing soils were encountered due to booming construction. Therefore, the main goal of this research was to study in depth the mechanism by which these soils collapse and to investigate the effect of certain mechanical and chemical treatment on that mechanism. The research included studies of undisturbed, compacted, and lime-treated samples. Both mechanical and physicochemical tests were conducted. The mechanical tests included collapse, swell, and unconfined compressive strength. The physicochemical tests involved X-ray diffraction and scanning electron microscopy. Various sites of highly collapsing soils were classified with respect to collapse according to existing criteria and the soil of one site was selected for a detailed investigation. A predictive collapse criterion was developed and used to classify the collapse susceptibility of soils in Tucson. The microstructure of the selected soil was investigated before and after collapse. A physical model was proposed to explain the mechanism of collapse. The effects of initial water content, sequence of loading and wetting, and level of loading on the engineering behavior of the selected soil were investigated. Stabilization by compaction was studied using impact and static methods at seven points on the Standard Compaction Curve. The benefits of hydrated-lime additive and the short-term reactions of lime-treated samples were also studied. The research results indicated that the microstructure of the soil is highly porous due to many interassemblage pores. Fine clay particles were found either clothing or buttressing the larger silt particles. The collapse was due mainly to weakening or failure of the clay connectors between the larger soil particles due to swelling of the expansive clay minerals, reduction of the strength of clay connectors due to wetting, dispersion of the supporting buttresses, and reduction of capillary tension. Compaction by both impact and static methods minimized the collapse but not the swell of the soil. Lime treatment completely suppressed the soil's tendency toward collapse and swell.

This thesis applies lessons drawn from a historical-cultural analysis of rural power structures in Afghanistan to understand the nature of the threat posed by that country's opiate economy and to assess the counter-narcotics policies of the United Kingdom, the Government(s) of Afghanistan, and the United States. It argues that that the opiate economy should be considered an Afghan-specific problem involving narcotics, not a "drug war" problem involving Afghanistan. Specific lessons are taken from a chapter dedicated to Afghan culture, history, and rural power structures, and applied in chapters analyzing the opiate economy and current counter-narcotics policies. Several insights that are critical to sound policy, and that are not found in existing literature, are developed. Overall, the current policy emphasis on aggressive eradication of opium poppy is incongruent with local cultural and political realities and undermines central government stability. Counter-narcotics policy makers should adopt a roll-back strategy, eliminating cultivation from minor-cultivation provinces first for democraticgovernance, cultural, and counter-narcotics reasons. Counter-trafficking should be prioritized over eradication efforts and should particularly target anti-government forces, many of which are legacy groups of the anti-Soviet jihad and are not accountable to or culturally integral to rural society.

O pólen apícola apresenta elevadas porcentagens de nutrientes e pode ser utilizado como suplemento nutricional na alimentação humana. Este trabalho teve por objetivo principal avaliar a estabilidade das vitaminas antioxidantes (vitamina C, E e &#946;-caroteno) em pólen apícola durante um ano de estocagem. Foram adquiridos entre os meses de março e abril 2007 seis lotes de pólen apícola in natura e desidratado, diretamente de entrepostos de comercialização de produtos apícolas. Foram analisadas as concentrações das três vitaminas no tempo zero e em seguida amostras foram armazenadas, em embalagens fornecidas pelo produtor, sob três formas: a temperatura ambiente; a temperatura ambiente, porém protegida da luz; em freezer. Foi utilizado o método títulométrico para análise de vitamina C. Para &#946;-caroteno utilizou-se a cromatografia em coluna aberta no tempo zero e cromatografia líquida de alta eficiência após 6 e 12 meses de estocagem. Esta última foi utilizada para as análises da vitamina E. Foram realizadas ainda análises polínica e de composição centesimal. Foram encontradas as seguintes variações: 14&#177;0,25 a 119&#177;1,961 &#181;g/g para vitamina C, 19,43&#177;1,70 a 45,00&#177;3,61&#181;g/g para vitamina E 3,77&#177;0,10 a 99,27&#177;2,45 &#181;g/g para &#946;-caroteno em amostras frescas. Após processo de desidratação, houve uma alteração de 67,1% para mais na vitamina C (diferença significativa p<0,05), uma perda de 18,7% para vitamina E e de 15,6% para &#946;-caroteno. O valor pró-vitamínico A das amostras desidratadas variou de 0,26 a 6,48 &#181;g/g. A composição centesimal das amostras estudadas está de acordo com as especificações estabelecidas pela legislação brasileira em vigor (Instrução Normativa N° 3, de 19/01/2001). Houve grande variabilidade dos tipos polínicos encontrados nas amostras e alguns deles estiveram fortemente correlacionados com os teores de vitamina C (Myrtaceae), de &#946;-caroteno (Arecaceae, Cecropia e Fabaceae) e de lipídeos (Arecaceae e Fabaceae). Outros estiveram correlacionados de forma negativa, como é o caso dos Mimosa caesalpineafolia e Poacease com os níveis de &#946;-caroteno, do tipo Arecaceae com as proteínas e do tipo Mimosa caesalpineafolia com os lipídeos. A estocagem em freezer foi a condição mais eficiente na conservação das três vitaminas e a perda na estocagem a temperatura ambiente exposto a luz e protegido da luz foram semelhantes. Considerando-se as três condições estudadas, a vitamina E parece ser mais preservada durante estocagem quando comparada à vitamina C e ao &#946;-caroteno. Entretanto, conforme teste estatístico realizado, houve perdas significativas (p<0,05) apenas para vitamina C em todas as condições estudadas quando comparadas a sua concentração inicial (tempo 0). / Bee pollen contains high percentages of nutrients and it can be used as a nutritional supplement for human feeding. The aim of this work was to evaluate the stability the antioxidant vitamins (vitamin C, E and &#946;-carotene) in bee pollen during one year of storage. Six batches of fresh and dried bee pollen pellets were acquired in 2007 March and April from bee products warehouses. The three vitamins were quantified and then stored under three forms in packages supplied by the producer: in room temperature, in room temperature protected from light and frozen. Vitamin C was quantified by potentiometric titration. The open column chromatography was used for &#946;-carotene analyses in the zero time and the high performance liquid chromatography after 6 and 12 months storage. This last one was used for the vitamin E analyses. The centesimal composition and botanical characterization of the bee pollen were obtained. Vitamin content in fresh samples varied between 14&#177;0.25 and 119&#177;1.96&#181;g/g for vitamin C; 19.43&#177;1.70 and 45.00&#177;3.61&#181;g/g for vitamin E and 3.77&#177;0.10 and 99.27&#177;2.45&#181;g/g for &#946;-carotene. After the drying process a significant alteration 67.1 % for more in the vitamin C (p<0.05), a losses of 18.7% for vitamin E and 15.6% for &#946;-carotene were observed. The provitamin A value was between 0.26 and 6.48&#181;g/g. The proximal composition of the samples studied presented results which were ali accordance to the specifications established for the Brazilian regulation (Normative Instruction N° 3, 19/01/2001). A great variability of the pollen types was found in the samples and some of them were strongly correlated with the vitamin C (Myrtaceae), &#946;-carotene (Arecaceae, Cecropia and Fabaceae) and lipids (Arecaceae and Fabaceae). Other ones were negatively correlated, such as Mimosa caesalpineafolia and Poaceae types with &#946;-carotene, Arecaceae type with proteins and Mimosa caesalpineafolia type with lipids. Storage in freezer was more efficient to keep the vitamins and the losses at room temperature storage when exposed to light and in the dark were similar. Vitamin E was more preserved during the storage when compared to vitamin C and &#946;-carotene. However, only vitamin C presented significant statistical losses (p<0.05) in ali of the studied conditions when compared to its initial content.

by Leung Hiu Tung, Vivien. / Thesis (M.Phil.)--Chinese University of Hong Kong, 1996. / Includes bibliographical references (leaves 174-184). / Chapter Chapter One: --- Introduction --- p.1 / Chapter 1.1) --- Economic Restructuring in Hong Kong --- p.1 / Chapter 1.2) --- An Agenda Of and For Working Class Families --- p.2 / Chapter 1.3) --- Synopsis of Arguments --- p.5 / Chapter 1.4) --- "Methodology, Sampling and Limitation" --- p.13 / Chapter 1.5) --- Structure of This Thesis --- p.16 / Chapter Chapter Two: --- "Gender, Family and the Economy" --- p.18 / Chapter 2.1) --- Social Dimension of Economic Restructuring --- p.18 / Chapter 2.2) --- "Family, Gender and the Economy" --- p.20 / Chapter 2.3) --- Unpacking the Chinese Family --- p.24 / Chapter 2.4) --- Reconceptualizing Family Strategy --- p.26 / Chapter 2.5) --- Framework and Conceptualization --- p.32 / Chapter Chapter Three: --- The Socio-Political Context of Economic Restructuring in Hong Kong / Chapter 3.1) --- The Hong Kong Way of Economic Restructuring --- p.39 / Chapter 3.2) --- Economic Restructuring: A Private Problem of Workers --- p.42 / Chapter 3.3) --- The Gendered Labour Market Under Economic Restructuring --- p.46 / Chapter Chapter Four: --- Gender Embeddedness of Strategization --- p.50 / Chapter 4.1) --- Major Concerns of Coping Responses --- p.52 / Chapter 4.2) --- Strategic Orientation --- p.55 / Chapter 4.3) --- Familial Resources and Constraints in Devising Coping Responses --- p.59 / Chapter 4.4) --- Subjective Engagement of the Actors --- p.63 / Chapter Chapter Five: --- Familial Embeddedness I-- Strategization of Impoverished Families --- p.68 / Chapter 5.1) --- The Predicaments of Impoverished Families --- p.69 / Chapter 5.2) --- Impoverished Traditional Families: Patriarchal Household Resource Mobilization --- p.73 / Chapter 5.3) --- Impoverished Flexible Families: Negotiating Household Resource --- p.85 / Chapter 5.4) --- Impoverished Solitary Families: Relying on External Resource --- p.93 / Chapter 5.5) --- Strategization in Impoverished Families --- p.96 / Chapter Chapter Six: --- Familial Embeddedness II-- Strategization of Sustainable Families --- p.99 / Chapter 6.1) --- Sustainable Traditional Families: Securing Breadwinner's Employment --- p.101 / Chapter 6.2) --- Sustainable Flexible Families: Negotiating Couple's Employment --- p.113 / Chapter 6.3) --- Strategization in Sustainable Families --- p.123 / Chapter Chapter Seven: --- Familial Embeddedness III-- Strategization of Affluent Families --- p.127 / Chapter 7.1) --- Mobilizing Breadwinner's Employment --- p.128 / Chapter 7.2) --- Trivializing Wife's Employment --- p.131 / Chapter 7.3) --- Strategization in Affluent Families --- p.138 / Chapter Chapter Eight: --- Conclusion --- p.140 / Chapter 8.1) --- The Gender Embeddedness and Familial Embeddedness of Strategization --- p.141 / Chapter 8.2) --- The Social Impacts of Economic Restructuring -- Gender and Intraclass Differences --- p.143 / Chapter 8.3) --- Theoretical Implication: Family Strategy Revisited --- p.147 / Endnotes --- p.155 / Appendix / Chapter 1: --- Tables --- p.162 / Chapter 2: --- Profile of the Informants and of their Families --- p.164 / Chapter 3: --- Question Set for In-depth Interviews --- p.169 / Bibliography --- p.174 / LIST OF TABLES / Table 1.1 Gender Difference in Strategization --- p.5 / Table 1.2 Familial Embeddedness in Strategization --- p.7 / Table 4.1 Gender Difference in Strategization (Same as Table 1.1) --- p.51 / Table 4.2 Categorization of Informants Across Attitudes and Major Concerns in Strategization --- p.51 / Table 5.1 Familial Embeddedness in Strategization (Same as Table 1.2) --- p.68 / Table 5.2 Categorization of Families According to Different Familial Contexts --- p.69 / Appendix 1 / "Table I Persons and Percentage Engaged in Selected Economic Sectors, 1985-94" --- p.162 / Table II Establishments and Persons Engaged in the Manufacturing Sector --- p.162 / Table III Nominal and Real Wage Indices of Payroll per Person Engaged --- p.164

Les films de ZrO2 pur sont déposés par MOCVD (Metal-Organic Chemical Vapor Deposition) en variant de nombreux paramètres du processus. L’influence des conditions de dépôt sur l’évolution de la microstructure (morphologies, structure cristalline/phase, texture et contrainte résiduelle) a été étudiée et clarifiée. Par des analyses approfondies des résultats expérimentaux, trois mécanismes typiques de croissance de dépôt de ZrO2 ont été proposées. Les contraintes de croissance de compression sont en relation directe avec la diffusion atomique et la quantité d’espèces piégées dans les films. La formation de la texture cristallographique est complexe et deux types de textures ont été analysées dans la phase tétragonale : la texture de fibre {1 1 0}t est contribuée par l’effet superplastique des nano-cristallites de ZrO2 et par la contrainte de croissance de compression ; tandis que la morphologie en facette est due à la croissance concurrentielle de différents plans cristallographiques. La stabilisation de la phase tétragonale de ZrO2 a été analysée et discutée. En plus de la taille critique des cristallites, la stabilisation de la phase tétragonale est favorisée par deux autres mécanismes : la grande quantité des défauts cristallins et la morphologie des cristallites. / Pure ZrO2 films were deposited by MOCVD (Metal-Organic Chemical Vapor Deposition) by varying the deposition parameters over large range. The influence of deposition conditions on the evolution of the microstructure (morphology, crystal structure / phase, texture and residual stress) was studied and clarified. Through careful study and analysis of experimental results, three typical mechanisms of deposition of ZrO2 have been proposed. The compressive growth stresses are directly related to atomic diffusion and the trapped-in effects during deposition. The formation of crystallographic texture is complex and two types of textures were analyzed in the tetragonal phase: the fiber texture {1 1 0}t is supposed to be the result of the effect of superplastic of ZrO2 nano-crystallites and the compressive growth stress, while the facet morphology (the {0 1 1}t fiber) is due to the competitive growth of different crystallographic planes. The stabilization of the tetragonal phase of ZrO2 was analyzed and discussed. In addition to the critical size of crystallites, the stabilization of the tetragonal phase can be favored by two mechanisms: the large amount of crystal defects and morphology of crystallites.

A hybrid system is a dynamical system that exhibits both continuous and discrete dynamic behavior. Hybrid systems arise in a wide variety of important applications in diverse areas, ranging from biology to computer science to air traffic dynamics. The interaction of continuous- and discrete-time dynamics in a hybrid system often leads to very rich dynamical behavior and phenomena that are not encountered in purely continuous- or discrete-time systems. Investigating the dynamical behavior of hybrid systems is of great theoretical and practical importance. The objectives of this thesis are to develop the qualitative theory of nonlinear hybrid systems with impulses, time-delay, switching modes, and stochastic disturbances, to develop algorithms and perform analysis for hybrid systems with an emphasis on stability and control, and to apply the theory and methods to real-world application problems. Switched nonlinear systems are formulated as a family of nonlinear differential equations, called subsystems, together with a switching signal that selects the continuous dynamics among the subsystems. Uniform stability is studied emphasizing the situation where both stable and unstable subsystems are present. Uniformity of stability refers to both the initial time and a family of switching signals. Stabilization of nonlinear systems via state-dependent switching signal is investigated. Based on assumptions on a convex linear combination of the nonlinear vector fields, a generalized minimal rule is proposed to generate stabilizing switching signals that are well-defined and do not exhibit chattering or Zeno behavior. Impulsive switched systems are hybrid systems exhibiting both impulse and switching effects, and are mathematically formulated as a switched nonlinear system coupled with a sequence of nonlinear difference equations that act on the switched system at discrete times. Impulsive switching signals integrate both impulsive and switching laws that specify when and how impulses and switching occur. Invariance principles can be used to investigate asymptotic stability in the absence of a strict Lyapunov function. An invariance principle is established for impulsive switched systems under weak dwell-time signals. Applications of this invariance principle provide several asymptotic stability criteria. Input-to-state stability notions are formulated in terms of two different measures, which not only unify various stability notions under the stability theory in two measures, but also bridge this theory with the existent input/output theories for nonlinear systems. Input-to-state stability results are obtained for impulsive switched systems under generalized dwell-time signals. Hybrid time-delay systems are hybrid systems with dependence on the past states of the systems. Switched delay systems and impulsive switched systems are special classes of hybrid time-delay systems. Both invariance property and input-to-state stability are extended to cover hybrid time-delay systems. Stochastic hybrid systems are hybrid systems subject to random disturbances, and are formulated using stochastic differential equations. Focused on stochastic hybrid systems with time-delay, a fundamental theory regarding existence and uniqueness of solutions is established. Stabilization schemes for stochastic delay systems using state-dependent switching and stabilizing impulses are proposed, both emphasizing the situation where all the subsystems are unstable. Concerning general stochastic hybrid systems with time-delay, the Razumikhin technique and multiple Lyapunov functions are combined to obtain several Razumikhin-type theorems on both moment and almost sure stability of stochastic hybrid systems with time-delay. Consensus problems in networked multi-agent systems and global convergence of artificial neural networks are related to qualitative studies of hybrid systems in the sense that dynamic switching, impulsive effects, communication time-delays, and random disturbances are ubiquitous in networked systems. Consensus protocols are proposed for reaching consensus among networked agents despite switching network topologies, communication time-delays, and measurement noises. Focused on neural networks with discontinuous neuron activation functions and mixed time-delays, sufficient conditions for existence and uniqueness of equilibrium and global convergence and stability are derived using both linear matrix inequalities and M-matrix type conditions. Numerical examples and simulations are presented throughout this thesis to illustrate the theoretical results.

In this study, a parametric model for a main battle tank electric gun turret drive system stabilization controller has been developed. Main scope was the study of the muzzle deviation due to barrel flexibility. Traverse and elevation dynamics has been modeled to include the drive-line and barrel flexibilities. Order of the models has been kept large enough to cover the frequencies dominant in the interest scope but at the same time low enough to create a parametric model which can be used in real-time fire control computers. Therefore a 5-dof elevation and a 7-dof traverse models have been implemented. These models have been used to design a classical feedback and feedforward controllers which performed good enough to meet 0.5mrad stabilization accuracies. After satisfactory results have been obtained from the stabilization controller, a special coincidence algorithm has been implemented by time-series analysis of the disturbance signal which is constantly being measured by the feedforward gyro. Necessity of predicting the future muzzle angular orientation due to the latency in fire is discussed and by using autoregressive modeling of the disturbance signal, future values of the disturbance signal has been entered into the observer model. The prediction horizon has been set to the time delay value between the trigger is pulled by the gunner and the ammunition exit from the muzzle. By checking the future coincidence within a very narrow windowing (0.05mrad) a 100% first round hit probability in theory has been achieved. This is assured since the coincidence inhibited the fire signals which were to miss the aiming point with a large error.

A hybrid system is a dynamical system that exhibits both continuous and discrete dynamic behavior. Hybrid systems arise in a wide variety of important applications in diverse areas, ranging from biology to computer science to air traffic dynamics. The interaction of continuous- and discrete-time dynamics in a hybrid system often leads to very rich dynamical behavior and phenomena that are not encountered in purely continuous- or discrete-time systems. Investigating the dynamical behavior of hybrid systems is of great theoretical and practical importance. The objectives of this thesis are to develop the qualitative theory of nonlinear hybrid systems with impulses, time-delay, switching modes, and stochastic disturbances, to develop algorithms and perform analysis for hybrid systems with an emphasis on stability and control, and to apply the theory and methods to real-world application problems. Switched nonlinear systems are formulated as a family of nonlinear differential equations, called subsystems, together with a switching signal that selects the continuous dynamics among the subsystems. Uniform stability is studied emphasizing the situation where both stable and unstable subsystems are present. Uniformity of stability refers to both the initial time and a family of switching signals. Stabilization of nonlinear systems via state-dependent switching signal is investigated. Based on assumptions on a convex linear combination of the nonlinear vector fields, a generalized minimal rule is proposed to generate stabilizing switching signals that are well-defined and do not exhibit chattering or Zeno behavior. Impulsive switched systems are hybrid systems exhibiting both impulse and switching effects, and are mathematically formulated as a switched nonlinear system coupled with a sequence of nonlinear difference equations that act on the switched system at discrete times. Impulsive switching signals integrate both impulsive and switching laws that specify when and how impulses and switching occur. Invariance principles can be used to investigate asymptotic stability in the absence of a strict Lyapunov function. An invariance principle is established for impulsive switched systems under weak dwell-time signals. Applications of this invariance principle provide several asymptotic stability criteria. Input-to-state stability notions are formulated in terms of two different measures, which not only unify various stability notions under the stability theory in two measures, but also bridge this theory with the existent input/output theories for nonlinear systems. Input-to-state stability results are obtained for impulsive switched systems under generalized dwell-time signals. Hybrid time-delay systems are hybrid systems with dependence on the past states of the systems. Switched delay systems and impulsive switched systems are special classes of hybrid time-delay systems. Both invariance property and input-to-state stability are extended to cover hybrid time-delay systems. Stochastic hybrid systems are hybrid systems subject to random disturbances, and are formulated using stochastic differential equations. Focused on stochastic hybrid systems with time-delay, a fundamental theory regarding existence and uniqueness of solutions is established. Stabilization schemes for stochastic delay systems using state-dependent switching and stabilizing impulses are proposed, both emphasizing the situation where all the subsystems are unstable. Concerning general stochastic hybrid systems with time-delay, the Razumikhin technique and multiple Lyapunov functions are combined to obtain several Razumikhin-type theorems on both moment and almost sure stability of stochastic hybrid systems with time-delay. Consensus problems in networked multi-agent systems and global convergence of artificial neural networks are related to qualitative studies of hybrid systems in the sense that dynamic switching, impulsive effects, communication time-delays, and random disturbances are ubiquitous in networked systems. Consensus protocols are proposed for reaching consensus among networked agents despite switching network topologies, communication time-delays, and measurement noises. Focused on neural networks with discontinuous neuron activation functions and mixed time-delays, sufficient conditions for existence and uniqueness of equilibrium and global convergence and stability are derived using both linear matrix inequalities and M-matrix type conditions. Numerical examples and simulations are presented throughout this thesis to illustrate the theoretical results.

Cette thèse se positionne dans le cadre de l'algorithmique distribuée tolérante aux pannes adaptée aux réseaux mobiles à grande échelle.L'auto-stabilisation est une approche de tolérance aux pannes satisfaisante dans les systèmes ayant des perturbations transitoires, mais pas dans les réseaux très dynamiques à grande échelle. La faute est due à l'éventuelle absence totale de service lorsque les perturbations sont fréquentes.Pour remédier à cet inconvénient, nous avons introduit l'approche auto-stabilisation robuste apportant une garantie de service pendant la phase de stabilisation.La garantie de service offerte par l'auto-stabilisation robuste est assurée via : (1) le délai de reprise d'un service minimum, et(2) la préservation du service minimum pendant la convergence vers un service optimum en dépit de l'occurrence de certaines perturbations hautement tolérées.L'intérêt d'avoir la propriété auto-stabilisation robuste est d'assurer une haute disponibilité du système en dépit de l'occurrence des perturbations et changements topologiques.Dans cette thèse, nous proposons, prouvons et évaluons une suite protocolaire auto-stabilisante robuste.Dans un premier temps, nous proposons deux protocoles auto-stabilisants robustes pour les problèmes de partitionnement, et l'établissement et le maintien de la connaissance des clusters voisins.Les deux protocoles sont écrits dans le modèle à états et fonctionnent sous l'hypothèse d'un démon distribué faiblement équitable.Le protocole de partitionnement, baptisé R-BSC, permet de partitionner le réseau en clusters à 1-saut. Les noeuds choisis pour être leaders sont les plus aptes à ce rôle, et les clusters construits sont de taille bornée dans le but d'équilibrer la charge entre leaders.Le protocole R-BSC fournit rapidement, en 4 rounds seulement, un service minimum où le réseau est complètement partitionné en clusters de taille bornée.Pendant la convergence vers un service optimum, où les leaders seront bien les noeuds les plus aptes et leur nombre sera réduit localement, le service minimum restera préservé. Le protocole de connaissance des clusters voisins, baptisé R-CNK, permet à chaque leader de connaître l'identité des leaders des clusters voisins, les chemins menant vers eux, ainsi que la composition (liste des noeuds ordinaires) des clusters voisins.Le service minimum de notre protocole R-CNK, atteint après 4 rounds seulement, garantit que tout leader connaît toujours des chemins vers tous les leaders des clusters voisins. Ce service minimum est maintenu en dépit des changements de la structure hiérarchique : création / destruction des clusters, changement de composition des clusters suite au départ / arrivé des noeuds ordinaires.Un deuxième aspect de nos travaux concerne l'évaluation des protocoles conçus (R-BSC et R-CNK) dans le contexte des réseaux mobiles.Nous avons mené une étude expérimentale sous le simulateur NS2 pour évaluer les performances de nos protocoles, ainsi que ceux des protocoles auto-stabilisants correspondants.Cette étude a montré que nos protocoles R-BSC et R-CNK offrent de meilleurs performances en terme de garantie de service, d'où l'efficacité de l'approche auto-stabilisation robuste par rapport à l'auto-stabilisation classique. / This dissertation is focused on fault-tolerant distributed algorithms adapted to large scale mobile networks.Self-stabilization is a fault-tolerance approach suited for systems with transient disruptions, but not for large scale dynamic networks.The fault is due to the eventual total lack of service when faults occur frequently.To address this drawback, we have introduced the robust self-stabilization approach that improves the service guarantee during the stabilization phase.The service guarantee provided by the robust self-stabilization is achieved via:(1) fast recovery to a minimum service and(2) preservation of minimum service during the convergence to an optimum service despite the occurrence of highly tolerated disruptions.Having the robust self-stabilization property ensures a high availability of the system despite the occurrence disruptions and topological changes in the network.In this thesis, we propose, evaluate and prove a series of robust self-stabilizing protocols.At first, we propose two robust self-stabilizing protocols for both problems : clustering and the maintain of knowledge about neighbor clusters.The two protocols are written in the local shared memory model and operate under the assumption of a weakly fair distributed daemon.The clustering protocol, called R-BSC, gathers the network nodes into 1-hop clusters.It allows a best choice of leaders, and it builds clusters with limited size in order to balance the load between leaders.The protocol R-BSC quickly provides, after at most 4 rounds, a minimum service where the network is completely partitioned into bounded-size clusters.During the convergence towards an optimum service, in which leaders will be the most appropriate nodes and their number will be reduced locally, the minimum service is preserved.The protocol for knowledge of neighbor clusters, called R-CNK, allows each leader to know the identity of leaders of neighbor clusters, paths leading to them, and the composition (list of ordinary nodes) of its neighbor clusters.The minimum service provided by of R-CNK protocol, reached after 4 rounds, ensures that every leader always knows paths towards all the leaders of neighbor clusters.We conducted an experimental study using the simulator NS2 to evaluate and to compare the performance of our protocols (R-BSC and R-CNK) with those of their self-stabilizing version in the context of mobile networks.This study confirmed that our protocols R-BSC and R-CNK offer a better service guarantee.