In radio frequency (RF), filtering is an essential part of RF transceivers. They are employed for different purposes of band selection, channel selection, interference cancellation, image rejection, etc. These are all translated in selecting the wanted signal while mitigating the rest. This can be performed by either selecting the desired frequency range by a band pass filter or rejecting the unwanted part by a band stop filter. Although there has been tremendous effort to design RF tunable filters, there is still lack of designs with frequency and bandwidth software-tuning capability at frequencies above 4 GHz. This prevents the implementation of Software Defined Radios (SDR) where software tuning is a critical part in supporting multiple standards and frequency bands. Designing a tunable integrated filter will not only assist in realization of SDR, but it also causes an enormous shrinkage in the size of the circuit by replacing the current bulky off-chip filters. The main purpose of this research is to design integrated band pass and band stop filters aimed to perform interference cancellation. In order to do so, two systems are proposed for this thesis. The first system is a band pass filter capable of frequency and band with tuning for C band frequency range (4-8 GHz) and is implemented in 0.13 µm BiCMOS technology. Frequency tunability is accomplished by using a variable capacitor (varactor) and bandwidth tuning is carried out by employing a negative transconductance cell to compensate for the loss of the elements. Additional circuitry is added to the band pass filter to enhance the selectivity of the filter. The second system is a band stop filter (notch) with the same capability as the band pass filter in terms of tuning. This system is implemented in C band, similar to its band stop counterpart and is capable of tuning its depth by using a negative transconductance in an LC tank. A negative feedback is added to the circuit to improve the bandwidth. While implemented in the same process as the band pass filter, it only employs CMOS transistors since it is generally more attractive due to its lower cost and scalability. Both of the systems mentioned use a varactor for changing the center frequency which is a nonlinear element. Therefore, the nonlinearity of it is modelled using two different methods of nonlinear feedback and Volterra series in order to gain further understanding of the nonlinear process taking place in the LC tank. After the validation of the models proposed using Cadence Virtuoso simulator, two methods of design and tuning are suggested to improve the linearity of the system. After post layout-extraction, the band pass filter is capable of Q tuning in the range of 3 to 270 and higher. With the noise figure of 10 to 14 dB and input 1-dB compression point as high as 2 dBm, the system shows a reasonably good performance along its operating frequency of 4 to 8 GHz. The band stop filter which is designed in the same frequency band can achieve better than 55 dB of rejection with the noise figure of 6.7 to 8.8 dB and 1-dB compression point of -4 dBm. With the power consumption of 39 to 70 mW, the band stop filter can be used in a low power receiver to suppress unwanted signals. The technique used in the band stop filter can be applied to higher frequency ranges if the circuit is implemented in a more advanced silicon technology. Implementing the mentioned filters in a receiver along with other elements of low noise amplifiers, mixers, etc. would be a major step toward full implementation of SDR systems. Studying the linearity theory of varactors would help future designers identify the sources of nonlinearity and suggest more efficient tuning techniques to improve the linearity of RF electronic systems. / Master of Science

<p dir="ltr">Brute force guessing attacks continue to pose a significant threat to user passwords. To protect user passwords against brute force attacks, many organizations impose restrictions aimed at forcing users to select stronger passwords. Organizations may also adopt stronger hashing functions in an effort to deter offline brute force guessing attacks. However, these defenses induce trade-offs between security, usability, and the resources an organization is willing to investigate to protect passwords. In order to make informed password policy decisions, it is crucial to understand the distribution over user passwords and how policy updates will impact this password distribution and/or the strategy of a brute force attacker.</p><p dir="ltr">This first part of this thesis focuses on developing rigorous statistical tools to analyze user password distributions and the behavior of brute force password attackers. In particular, we first develop several rigorous statistical techniques to upper and lower bound the guessing curve of an optimal attacker who knows the user password distribution and can order guesses accordingly. We apply these techniques to analyze eight password datasets and two PIN datasets. Our empirical analysis demonstrates that our statistical techniques can be used to evaluate password composition policies, compare the strength of different password distributions, quantify the impact of applying PIN blocklists, and help tune hash cost parameters. A real world attacker may not have perfect knowledge of the password distribution. Prior work introduced an efficient Monte Carlo technique to estimate the guessing number of a password under a particular password cracking model, i.e., the number of guesses an attacker would check before this particular password. This tool can also be used to generate password guessing curves, but there is no absolute guarantee that the guessing number and the resulting guessing curves are accurate. Thus, we propose a tool called Confident Monte Carlo that uses rigorous statistical techniques to upper and lower bound the guessing number of a particular password as well as the attacker's entire guessing curve. Our empirical analysis also demonstrate that this tool can be used to help inform password policy decisions, e.g., identifying and warning users with weaker passwords, or tuning hash cost parameters.</p><p dir="ltr">The second part of this thesis focuses on developing stronger password hashing algorithms to protect user passwords against offline brute force attacks. In particular, we establish that the memory hard function Scrypt, which has been widely deployed as password hash function, is maximally bandwidth hard. We also present new techniques to construct and analyze depth robust graph with improved concrete parameters. Depth robust graph play an essential rule in the design and analysis of memory hard functions.</p>

AC and DC impedances of switching power converters are used for the stability analysis of modern power electronics systems at three-phase AC and single-phase DC interfaces. Therefore, a small-signal characterization algorithm for switching power converter, which is based on FFT, will be presented and explained. The presented extraction algorithm is general and can be used to obtain other small-signal transfer functions of arbitrary power converter switching simulation models. Furthermore, FFT algorithm is improved by using cross power spectral density functions for identification, resulting in an algorithm, which is more noise immune. Both small-signal identification algorithms are validated in simulations, and CPSD algorithm is used in experimental measurement procedure. Several wide bandwidth injection signals, among which are chirp, multi-tone, pulse and white noise, are compared and theoretically analyzed. Several hardware examples are included in the analysis. The second part of the dissertation will focus on the modeling of small-signal input dq admittance of multi-pulse diode rectifiers, providing comparison between well-known averaged value models (AVMs), parametric averaged value models (PAVM), the switching simulation model and hardware measurements. Analytical expressions for all four admittances present in the dq matrix are derived and analyzed in depth, revealing the accuracy range of the averaged models. Furthermore, a hardware set-up is built, measured and modeled, showing that the switching simulation model captures nonlinear sideband effects accurately. In the end, a multi-pulse diode rectifier feeding a constant power load is analyzed with modified AVM and through detailed simulations of switching model, proving effectiveness of the proposed modifications. The third part describes implementation and design of a single-phase multi-level single-phase shunt current injection converter based on cascaded H-bridge topology. Special attention is given toward the selection of inductors and capacitors, trying to optimize the selected component values and fully utilize operating range of the converter. The proposed control is extensively treated, including inner current, outer voltage loop and voltage balancing loops. The designed converter is constructed and integrated with measurement system, providing experimental verification. The proposed multi-level single-phase converter is a natural solution for single-phase shunt current injection with the following properties: modular design, capacitor energy distribution, reactive element minimization, higher equivalent switching frequency, capability to inject higher frequency signals, suitable to perturb higher voltage power systems and capable of generating cleaner injection signals. Finally, a modular interleaved single-phase series voltage injection converter, consisting of multiple paralleled H-bridges is designed and presented. The decoupling control is proposed to regulate ac injection voltage, providing robust and reliable strategy for series voltage injection. The designed converter is simulated using detailed switching simulation model and excellent agreement between theory and simulation results are obtained. The presented control analysis treats different loads, examining robustness of the circuit to load variations. Simulation model and hardware prototype results verify the effectiveness of the proposed wide-bandwidth identification of small-signal dq impedances via single-phase injections. / Ph. D.

Terahertz (THz) spectroscopy is a powerful tool that uses ultrashort pulses of light to study the properties of materials on picosecond time scales. THz light can be generated through a variety of methods. In our lab, we generate THz through the process of optical rectification in nonlinear optical (NLO) organic crystals. THz light can be used to study several phenomena in materials, such as spin precession, electron acceleration, vibrational and rotational motion. The work presented in this dissertation is divided into two parts: (1) the generation of THz light and (2) applications of THz light. The first portion of this work shows how THz light is generated, with an emphasis on the generation through optical rectification. We also show how to improve the generation of THz light by creating heterogenous multi-layer structures with yellow organic THz generation crystals. Additionally, we show that crystals used for THz generation can also be used to generate second-harmonic light. In the second half of this work, we show that THz light can be used to study the vibrational motion of molecular systems. We model how resonant vibrational modes in a fluorobenzene molecule can be excited with a multi-THz pump to transfer energy anharmonically to non-resonant modes. We also show that we can use two-dimensional (2D) THz spectroscopy to excite infrared-active vibrational modes and probe Raman-active modes in a CdWO4 crystal to obtain a nonlinear response. We show that the nonlinear response is due to anharmonic coupling between vibrational modes and we can quantify the relative strengths of these anharmonic couplings, which previously was only accessible through first-principles calculations.

In recent years, there has been a growing commitment to studying the economic lives of the poor by better understanding their psychological lives (Banerjee and Duflo, 2007; Schilbach et al., 2016). These developments stem from the failure to empirically detect poverty traps, which have been at the core of the development literature for decades (Dean et al., 2017). Instead, empirical studies document the existence of very large rates of returns to investment, which, however, are not matched by observed consumption growth rates (Kremer et al., 2019). Several behaviors of the poor, which do not fit with traditional models, puzzled economists. Why do poor micro-entrepreneurs keep borrowing at extremely high interest rates instead of saving some of their profit to borrow less with each passing day (Ananth et al., 2007)? If using fertilizer has such high rates of return, why don't poor farmers purchase it (Duflo et al., 2008)? If the poor remain poor because they do not get enough calories, why do they spend their money on other things besides food (Banerjee and Duflo, 2007)? Such questions led to the rise of the subfield of Behavioral Development Economics, which applies insights from psychology and behavioral economics to study the economic behavior of the poor; trying to explain why and how it departs from standard economic models. Behavioral biases, studied extensively in Behavioral Economics, may be much more consequential for the poor. Failing to resist to the temptation of a hedonistic reward after a hard day of work will have very different implications for a poor person than for a rich one. This thesis aims to contribute to this new strand of literature, in particular to one of its branches titled "the psychology of poverty", which studies the impact poverty has on cognitive function and economic behavior. One influential theory in this field is the scarcity/mental bandwidth theory (Mullainathan and Shafir, 2013), which states that poverty implies not only lack of financial resources, but also lack of mental resources to focus on other things besides pressing concerns. At any time, a poor person's mind will be preoccupied with worries about bills, school fees or health problems; and how to best manage all of them with very limited resources. While this makes the poor better at decisions regarding the pressing issue at hand (Mullainathan and Shafir, 2013), it also makes them neglect other important domains which may not appear urgent enough (Shah et al., 2012, 2015). While the theory may help explain many puzzling behaviors of the poor, up to now there has been little evidence on real-world economic outcomes. The first two chapters of this thesis try to bring the framework closer to real-world economic decisions even though restricted to the lab setting. The main challenge with studying the psychology of poverty outside the lab is the fact that even exogenous changes in income will affect several other channels besides mental bandwidth, making it very challenging to pin down the precise mechanism. Given this, the first two chapters are limited to varying mental bandwidth in a lab setting, keeping income fixed. The novel aspect is that the decisions participants make mimic closely everyday life purchasing decisions, involving real products. I note however, that due to limited funding and ethical considerations, in both chapters decisions are only weakly incentivized: only 1% of participants actually receive the goods they selected. The first chapter explores the relationship between the psychology of poverty, investment in human capital, and financial incentives. Empirical evidence indicates that the poor are less attentive parents, investing less in the human capital of their children (McLoyd, 1998; Evans, 2004). This contributes to the inter-generational transmission of poverty because investing in human capital has extremely high rates of return, highest in early childhood (Cunha and Heckman, 2007; Cunha et al., 2010). The question is why don't the poor invest more? Traditional answers to this question put the blame on lack of knowledge of parenting practices, wrong beliefs on the expected returns or lower altruism. We propose an alternative explanation based on the scarcity theory. Poor parents may fail to invest the required time and resources in their child because their minds are preoccupied with other more urgent concerns. When there is uncertainty about how the next bill will be paid, spending time doing educational activities with the child may shift out of focus. When such behaviors keep repeating on a regular basis, a gap emerges between poor and non-poor children in the amount of cognitive and emotional stimulation they receive. The challenge is how to test this hypothesis. Given the identification issues with disentangling such channels with observational data, we bring it to the lab. Parents of toddlers, living in the UK, are invited to participate in an online experiment. First, they are asked to answer how their family would deal with various hypothetical financial scenarios which vary in severity (hard for the treatment group, easy for the control group). Among the treated, the scenarios aim to bring financial worries to mind, trying to capture what people living in poverty experience on a regular basis. After completing the scenarios, parents receive a budget of pounds 30 to be spent as they choose in an experimental market on 3 types of goods: necessities, child investment goods, and luxury goods. Half of parents are incentivized to purchase more child investment goods by being offered a 50% discount. This treatment investigates if financial worries change how parents respond to such incentives, and is motivated by the results in Das et al. (2013) which find that accounting for household re-optimization in response to a policy is crucial when evaluating its effects. We find that the incentive increases investment in human capital among lower income parents only when financial worries are not salient. When worries become salient, low income parents do not invest more but instead use the additional money to increase their demand of necessities. In addition, they also lower their demand for luxury goods to zero. When no discount is offered, we do not find financial worries to lower investment, which is likely to be explained by floor effects. Among higher income parents, financial worries do not affect behavior. The effects among lower income participants are driven by those who were further away from their last paycheck at the time of the experiment - an indicator of real world monetary scarcity. This finding increases the external validity of our main results. The second chapter departs from studying the human capital of children, focusing instead on the human capital of adultsootnote{However, the behavior studied is likely to have negative externalities also on children (e.g. domestic violence).}. Addictive (or temptation) goods have been at the core of academic and policy debates for decades. With Becker and Murphy (1988), addiction was rationalized as a utility maximizing decision where the individual fully internalizes the costs of consuming such goods. In this framework, the only scope for intervention is to balance out the externalities -- the costs that individuals place on society through consumption decisions (e.g. healthcare costs). Gruber (2001) questioned theoretically and empirically the rational framework, showing that with inconsistent time preferences, individuals do not fully internalize the cost of their behavior. Further studies have confirmed these findings which increased the scope of policy interventions (Gruber and Kőszegi, 2004; O’Donoghue and Rabin, 2006; Allcott et al., 2019a). The most widely used policy tools to limit the over-consumption of temptation are "sin" taxes, popular among governments because they bring large revenues. However, such taxes have sparked debates regarding their effects on income distribution. Since the poor tend to spend a higher share of their budget on temptation, they are likely to pay a higher cost. On the other hand, they are also the ones expected to benefit more in terms of health by consuming less. Traditionally, such taxes were placed on tobacco and alcohol. Recently, several governments have started adding taxes also on the consumption of unhealthy foods, such as sugary drinks and beverages. Crucial to determining the effect of the tax is the elasticity of demand with respect to price and the degree to which individuals are not internalizing their choices (Allcott et al., 2019a). The second chapter integrates the economics of temptation with the scarcity theory, and investigates if financial worries affect (i) the demand for temptation and (ii) the elasticities of demand with respect to price (sin taxes). The first question is not straightforward in the scarcity framework. While poverty is scarcity of financial resources, it is also scarcity of immediate gratification. The poor have stressful lives and jobs which are often less rewarding and highly physically demanding. Compensating for these struggles is harder since they can only access a small set of potential alternatives to addictive goods (e.g. going to nice restaurant and travelling are not really in the choice set of the poor). Following a similar design as in the first chapter but with a less specific population (adults living in the UK), we randomly trigger financial worries before asking participants to choose between necessities and temptation goods in an experimental market. The basket of temptation goods offered includes tobacco, alcohol and unhealthy foods and we simulate "sin taxes" by randomly increasing the price of temptation by 10% or 20%. We find that triggering financial worries lowers the demand for temptation but also dampens demand elasticities. The effects are stronger among low income participants. When financial worries are salient, their demand curve is actually slightly upward sloping. The finding is puzzling: financial worries appear to limit over-consumption of temptation, but they also hurt the poor the most when additional taxes are introduced. We find suggestive evidence that both effects are mediated by an increased focus on urgent necessities. The first two chapters integrated the scarcity framework into public policies. The results are very consistent across studies and have clear policy implications. Among the poor, when monetary concerns are top of mind: (i) incentivizing investments in human capital may not achieve its desired outcome, (ii) (dis)incentivizing consumption of temptation through new taxes may harm the poor the most since they do not lower their demands in response to price increases, which leads, through taxation, to a transfer of funds from the poor to the nonpoor without having any corrective effects (see Bernheim and Rangel, 2004; Bernheim and Taubinsky, 2018). However, I must note that both chapters make only speculative policy recommendations given that they lack the normative counterfactual. Further research is needed to rigorously establish the welfare implications of financial worries. The third chapter takes a step back from economic decisions to studying how violence exposure affects cognitive function in children. Unfortunately violence and poverty are closely linked in a vicious cycle. Economically deprived neighborhoods are in general also more violent. In addition to monetary concerns, the minds of the poor are likely to be preoccupied with safety concerns. This study attempts to apply the framework in Mullainathan and Shafir (2013), focusing on security concerns instead of monetary ones. While the link between the scarcity framework and violence as scarcity of security is novel and up for debate, the chapter is closely connected with the literature on the impact of emotions on cognition and decision making (Loewenstein and Lerner, 2003; Lerner et al., 2003, 2015; Callen et al., 2014; Bogliacino et al., 2017). In a lab-in-the-field experiment, primary school children in El Salvador are randomly assigned to recall episodes of violence exposure before or after taking cognitive tests. I find that recalling violence exposure before taking the tests, increases cognitive performance by 0.2 standard deviations, effect significantly stronger for children reporting higher exposure. The estimates contrast previous findings on the effect of violence and cognitive function (Sharkey, 2010; Sharkey et al., 2012; Bogliacino et al., 2017) and call for further research in the field.

Long Term Evolution (LTE) is fourth generation broadband wireless access technology. As LTE networks continue to expand, their energy consumption has become a concern due to increased costs and environmental impact. To address this issue, we present a novel approach in this thesis that focuses on energy-efficient LTE network operation by optimizing cell bandwidth deployment. Only historical data extracted from a real-world LTE network is provided by Ericsson in the thesis, so the model has to learn the pattern by exploiting the data. The model formulates the task as an offline contextual bandits problem and implements offline learning and evaluation algorithms. In addition, the model incorporates two agents: an energy-driven agent and a teacher agent within the bandwidth recommendation engine. This enables energy savings while ensuring uninterrupted network service. Users have the flexibility to customize the desired percentage of energy savings by adjusting a hyper-parameter that determines the relative weight of the two agents in the final bandwidth decision. Through comprehensive offline evaluation, the model is assessed and demonstrates its ability to conserve energy without compromising network performance. / Long Term Evolution (LTE) är en fjärde generationens bredbandslösning för trådlös access. Medan LTE-nätverk fortsätter att expandera har energiförbrukningen blivit en oro på grund av ökade kostnader och miljöpåverkan. För att hantera denna fråga presenterar vi i denna avhandling en innovativ metod som fokuserar på energieffektiv drift av LTE-nätverk genom att optimera tilldelningen av cellbandbredd. Endast historiska data som utvunnits från ett verkligt LTE-nätverk tillhandahålls av Ericsson i avhandlingen, vilket innebär att modellen måste lära sig mönster genom att utnyttja dessa data. Modellen formulerar uppgiften som ett offline-contextual bandits-problem och implementerar algoritmer för offline-inlärning och utvärdering. Dessutom inkluderar modellen två agenter: en energistyrd agent och en läraragent i bandbreddrekommendationssystemet. Detta möjliggör energibesparingar samtidigt som kontinuerlig nätverkstjänst garanteras. Användare har flexibiliteten att anpassa den önskade procentandelen av energibesparingar genom att justera en hyperparameter som bestämmer den relativa vikten av de två agenterna vid beslutsfattandet om slutlig bandbredd. Genom omfattande utvärdering offline bedöms modellen och dess förmåga att spara energi utan att äventyra nätverksprestanda bekräftas.

The ability to control the path a signal takes through microwave circuitry is crucial when designing RF systems. The component that allows for the control of the signal path is called a switch, and it is the microwave component that this thesis will focus on. Switches are widely used in the growing defense and space industries, both of which have strict requirements on their systems. This thesis aims to investigate four switch topologies and compare them to establish which topology is most suitable in high-performance systems where high isolation and wide bandwidths are essential parameters. The different topologies were simulated using Cadence AWR Microwave Studio to evaluate the parameters of interest. Built-in models for PIN diodes were employed to capture the nonlinear nature of the diodes and quantify the linearity of the switches. Measurements of PIN diodes using a spectrum analyzer and a vector network analyzer were used to confirm that the models of the PIN diode resembled the actual characteristics of PIN diodes. Out of the four topologies investigated, the series, shunt, series-shunt and double shunt, it was concluded that the series-shunt topology was most suitable for wideband multiple-throw switches. The double shunt topology was more suitable for SPST switches and narrowband SPDT switches. From this thesis, it will be possible to conclude which topology is most suitable for a particular need and what key design parameters will impact the switch’s performance. / Förmågan att styra vägen en signal tar genom mikrovågskretsar är en avgörande förmåga i konstruktionen av de flesta RF-system. Komponenten som tillåter styrning av signalvägen kallas en omkopplare och är den mikrovågskomponent som detta examensarbete kommer att fokusera på. RFomkopplare kan realiseras på många olika sätt och kan integreras i flera typer av vågledande strukturer såsom mikrostrips, striplines och vågledare. Omkopplingselementen kan realiseras på många sätt, med användning av transistorer och dioder är de mest etablerade sätten. Sådana omkopplare används i stor utsträckning inom försvars- och rymdindustrin som båda har strikta krav på dess komponenter. Avhandlingen syftar därför till att undersöka ett antal switchtopologier och jämföra dessa för att fastställa vilken topologi som är mest lämplig i högpresterande system där hög isolering och breda bandbredder är viktiga parametrar. De olika topologierna simulerades med hjälp av Cadence AWR Microwave Studio för att utvärdera parametrarna av intresse. Den inbyggda modellen för PIN-dioder användes också för att fånga diodernas olinjära natur för att kvantifiera omkopplarnas linjäritet. Mätningar av PIN-dioder med användning av en spektrumanalysator och en Nätverksanalysator gjordes för att bekräfta att modellerna av PIN-dioderna efterliknade PIN-diodernas faktiska beteende. Det fastslogs att series-shunt-topologin var mest lämpad för bredbandsomkopplare med flera utgångar. Dubbel-shunt-topologin var mer lämplig i SPSTomkopplare och i smalbandiga SPDT-omkopplare. Från detta examensarbete kommer det att vara möjligt att dra slutsatser kring vilken topologi som är mest lämplig för ett särskilt behov tillsammans med vilka designparametrar som kommer att påverka omkopplarens prestanda.

The research initiatives addressed in this thesis are geared towards improving the performance of passive optical networks (PONs) through the development of advanced dynamic bandwidth allocation (DBA) protocols. In particular, the aim of the research undertaken is to enhance the quality of service (QoS) offered by standard PONs by means of providing subscribers with service level agreement (SLA) to fulfil applications and associated bandwidth requirements on demand. To accomplish the research objectives, a novel service and bandwidth focused DBA protocol is developed for standard time division multiplexing (TDM), gigabit-capable PONs (GPONs) by flexibly assigning a guaranteed minimum bandwidth to each optical network unit (ONU),terminated at subscribers premises. Modelling and simulation of the developed algorithms have displayed a tenfold enhancement in network performance, showing a superior performance to other published DBA protocols, in terms of mean packet delay. To accomplish protocol optimisation, the ONU upstream transmission properties of TDM-PONs have been further analysed and subsequently the ONU data transfer order in each communication cycle has been dynamically configured to increase the network upstream throughput by overlapping the upstream transmission period with parts of the bandwidth request-allocation process between OLT and ONUs. In addition, with the objective of extending the application of the developed protocol to long-reach PONs by means of reducing the augmented propagation delays due to the network’s extensive reach, the concept of virtual communication cycles has been incorporated into the optimised DBA algorithm. This approach demonstrates comparable transmission efficiency in the context of subscriber throughput and packet delay as in a standard PON but at much longer distances from the network exchange. To overcome the inevitably limited communication capacity of single wavelength TDM protocols and with the transportation of the ever increasing, time-sensitive, multi-media services in mind, a novel multi-wavelength DBA protocol is then developed to be applied to a wavelength division multiplexing–PON. With this protocol, both the downstream and upstream network capacity is dynamically adjusted according to subscribers’ service level and bandwidth demand in each polling cycle as opposed to a fixed upstream network capacity in TDM-PONs. It therefore also demonstrates improved upstream transmission efficiency.

L’émergence de nouvelles applications et de nouveaux services (tels que les applications multimédias, la voix-sur-IP, la télévision-sur-IP, la vidéo-sur-demande, etc.) et le besoin croissant de mobilité des utilisateurs entrainent une demande de bande passante de plus en plus croissante et une difficulté dans sa gestion dans les réseaux cellulaires sans fil (WCNs), causant une dégradation de la qualité de service. Ainsi, dans cette thèse, nous nous intéressons à la gestion des ressources, plus précisément à la bande passante, dans les WCNs. Dans une première partie de la thèse, nous nous concentrons sur la prédiction de la mobilité des utilisateurs des WCNs. Dans ce contexte, nous proposons un modèle de prédiction de la mobilité, relativement précis qui permet de prédire la destination finale ou intermédiaire et, par la suite, les chemins des utilisateurs mobiles vers leur destination prédite. Ce modèle se base sur : (a) les habitudes de l’utilisateur en terme de déplacements (filtrées selon le type de jour et le moment de la journée) ; (b) le déplacement courant de l’utilisateur ; (c) la connaissance de l’utilisateur ; (d) la direction vers une destination estimée ; et (e) la structure spatiale de la zone de déplacement. Les résultats de simulation montrent que ce modèle donne une précision largement meilleure aux approches existantes. Dans la deuxième partie de cette thèse, nous nous intéressons au contrôle d’admission et à la gestion de la bande passante dans les WCNs. En effet, nous proposons une approche de gestion de la bande passante comprenant : (1) une approche d’estimation du temps de transfert intercellulaire prenant en compte la densité de la zone de déplacement en terme d’utilisateurs, les caractéristiques de mobilité des utilisateurs et les feux tricolores ; (2) une approche d’estimation de la bande passante disponible à l’avance dans les cellules prenant en compte les exigences en bande passante et la durée de vie des sessions en cours ; et (3) une approche de réservation passive de bande passante dans les cellules qui seront visitées pour les sessions en cours et de contrôle d’admission des demandes de nouvelles sessions prenant en compte la mobilité des utilisateurs et le comportement des cellules. Les résultats de simulation indiquent que cette approche réduit largement les ruptures abruptes de sessions en cours, offre un taux de refus de nouvelles demandes de connexion acceptable et un taux élevé d’utilisation de la bande passante. Dans la troisième partie de la thèse, nous nous penchons sur la principale limite de la première et deuxième parties de la thèse, à savoir l’évolutivité (selon le nombre d’utilisateurs) et proposons une plateforme qui intègre des modèles de prédiction de mobilité avec des modèles de prédiction de la bande passante disponible. En effet, dans les deux parties précédentes de la thèse, les prédictions de la mobilité sont effectuées pour chaque utilisateur. Ainsi, pour rendre notre proposition de plateforme évolutive, nous proposons des modèles de prédiction de mobilité par groupe d’utilisateurs en nous basant sur : (a) les profils des utilisateurs (c’est-à-dire leur préférence en termes de caractéristiques de route) ; (b) l’état du trafic routier et le comportement des utilisateurs ; et (c) la structure spatiale de la zone de déplacement. Les résultats de simulation montrent que la plateforme proposée améliore la performance du réseau comparée aux plateformes existantes qui proposent des modèles de prédiction de la mobilité par groupe d’utilisateurs pour la réservation de bande passante. / The emergence of new applications and services (e.g., multimedia applications, voice over IP and IPTV) and the growing need for mobility of users cause more and more growth of bandwidth demand and a difficulty of its management in Wireless Cellular Networks (WCNs). In this thesis, we are interested in resources management, specifically the bandwidth, in WCNs. In the first part of the thesis, we study the user mobility prediction that is one of key to guarantee efficient management of available bandwidth. In this context, we propose a relatively accurate mobility prediction model that allows predicting final or intermediate destinations and subsequently mobility paths of mobile users to reach these predicted destinations. This model takes into account (a) user’s habits in terms of movements (filtered according to the type of day and the time of the day); (b) user's current movement; (c) user’s contextual knowledge; (d) direction from current location to estimated destination; and (e) spatial conceptual maps. Simulation results show that the proposed model provides good accuracy compared to existing models in the literature. In the second part of the thesis, we focus on call admission control and bandwidth management in WCNs. Indeed, we propose an efficient bandwidth utilization scheme that consists of three schemes: (1) handoff time estimation scheme that considers navigation zone density in term of users, users’ mobility characteristics and traffic light scheduling; (2) available bandwidth estimation scheme that estimates bandwidth available in the cells that considers required bandwidth and lifetime of ongoing sessions; and (3) passive bandwidth reservation scheme that passively reserves bandwidth in cells expected to be visited by ongoing sessions and call admission control scheme for new call requests that considers the behavior of an individual user and the behavior of cells. Simulation results show that the proposed scheme reduces considerably the handoff call dropping rate while maintaining acceptable new call blocking rate and provides high bandwidth utilization rate. In the third part of the thesis, we focus on the main limitation of the first and second part of the thesis which is the scalability (with the number of users) and propose a framework, together with schemes, that integrates mobility prediction models with bandwidth availability prediction models. Indeed, in the two first contributions of the thesis, mobility prediction schemes process individual user requests. Thus, to make the proposed framework scalable, we propose group-based mobility prediction schemes that predict mobility for a group of users (not only for a single user) based on users’ profiles (i.e., their preference in terms of road characteristics), state of road traffic and users behaviors on roads and spatial conceptual maps. Simulation results show that the proposed framework improves the network performance compared to existing schemes which propose aggregate mobility prediction bandwidth reservation models.

La popularité croissante des applications Internet très gourmandes en bande passante (P2P, streaming,...) nous pousse à considérer le problème suivant :Comment construire des systèmes de communications collectives efficaces sur une plateforme à grande échelle ? Le développement de schéma de communications collectives dans le cadre d'un réseau distribué à grande échelle est une tâche difficile, qui a été largement étudiée et dont de multiples solutions ont été proposées. Toutefois, une nouvelle approche globale et systématique est nécessaire, une approche qui combine des modèles de réseaux et la conception algorithmique.Dans ce mémoire nous proposons l'utilisation de modèles capables de capturer le comportement d'un réseau réel et suffisamment simples pour que leurs propriétés mathématiques puissentêtre étudiées et pour qu'il soit possible de créer des algorithmesoptimaux. Premièrement, nous considérons le problème d'évaluation de la bande passante disponible pour une connexion point-à-point donnée. Nousétudions la façon d'obtenir des jeux de données de bande passante, utilisant plateforme PlanetLab. Nous présentons aussi nos propres jeux de données, jeux obtenus avec bedibe, un logiciel que nous avons développé. Ces données sont nécessaires pour évaluer les performances des différents algorithmesde réseau. Bien qu'on trouve de nombreux jeux de données de latence,les jeux de données de bande passante sont très rares. Nous présentons ensuite un modèle, appelé LastMile, qui estime la bande passante. En profitant des jeux de données décrits précédemment, nous montrons que cet algorithme est capable de prédire la bande passante entre deux noeuds donnés avec une précision comparable au meilleur algorithme connu de prédiction (DMF). De plus le modèle LastMile s'étend naturellement aux prédictions dans le scénario de congestion (plusieurs connexions partageant un même lien). Nous sommes effectivement en mesure de démontrer, à l'aide des ensembles de données PlanetLab, que la prédiction LastMile est préférable dans des tels scénarios.Dans le troisième chapitre, nous proposons des nouveaux algorithmes pour résoudre le problème de diffusion. Nous supposons que le réseau est modélisé par le modèle LastMile. Nous montrons que, sous cette hypothèse, nous sommes en mesure de fournir des algorithmes avec des ratios d'approximation élevés. De plus nous étendons le modèle LastMile, de manière à y intégrer des artéfacts de connectivité, dans notre cas ce sont des firewalls qui empêchent certains nœuds de communiquer directement entre eux. Dans ce dernier cas, nous sommes également en mesure de fournir des algorithmes d'approximation avec des garanties de performances prouvables. Les chapitres 1 à 3 forment les trois étapes accomplies de notre programme qui visent trois buts. Premièrement, développer à partir dezéro un modèle de réseau de communication. Deuxièmement, prouver expérimentalement sa performance. Troisièmement, montrer qu'il peut être utilisé pour développer des algorithmes qui résolvent les problèmes de communications collectives. Dans le 4e chapitre, nous montrons comment on peut concevoir dessystèmes de communication efficaces, selon différents modèles decoûts, en utilisant des techniques combinatoires,tout en utilisant des hypothèses simplificatrices sur la structure duréseau et les requêtes. Ce travail est complémentaire au chapitre précédent puisque auparavant, nous avons adopté l'hypothèse que les connectionsétaient autonomes (i.e. nous n'avons aucun contrôle sur le routage des connexions simples). Dans le chapitre 4, nous montrons comment résoudre le problème du routage économe en énergie, étant donnée une topologie fixée. / The increasing popularity of Internet bandwidth-intensive applications prompts us to consider followingproblem: How to compute efficient collective communication schemes on large-scale platform?The issue of designing a collective communication in the context of a large scale distributed networkis a difficult and a multi-level problem. A lot of solutions have been extensively studied andproposed. But a new, comprehensive and systematic approach is required, that combines networkmodels and algorithmic design of solutions.In this work we advocate the use of models that are able to capture real-life network behavior,but also are simple enough that a mathematical analysis of their properties and the design of optimalalgorithms is achievable.First, we consider the problem of the measuring available bandwidth for a given point-topointconnection. We discuss how to obtain reliable datasets of bandwidth measurements usingPlanetLab platform, and we provide our own datasets together with the distributed software usedto obtain it. While those datasets are not a part of our model per se, they are necessary whenevaluating the performance of various network algorithms. Such datasets are common for latencyrelatedproblems, but very rare when dealing with bandwidth-related ones.Then, we advocate for a model that tries to accurately capture the capabilities of a network,named LastMile model. This model assumes that essentially the congestion happens at the edgesconnecting machines to the wide Internet. It has a natural consequence in a bandwidth predictionalgorithm based on this model. Using datasets described earlier, we prove that this algorithm is ableto predict with an accuracy comparable to best known network prediction algorithm (DistributedMatrix Factorization) available bandwidth between two given nodes. While we were unable toimprove upon DMF algorithm in the field of point-to-point prediction, we show that our algorithmhas a clear advantage coming from its simplicity, i.e. it naturally extends to the network predictionsunder congestion scenario (multiple connections sharing a bandwidth over a single link). We areactually able to show, using PlanetLab datasets, that LastMile prediction is better in such scenarios.In the third chapter, we propose new algorithms for solving the large scale broadcast problem.We assume that the network is modeled by the LastMile model. We show that under thisassumption, we are able to provide algorithms with provable, strong approximation ratios. Takingadvantage of the simplicity and elasticity of the model, we can even extend it, so that it captures theidea of connectivity artifacts, in our case firewalls preventing some nodes to communicate directlybetween each other. In the extended case we are also able to provide approximation algorithmswith provable performance.The chapters 1 to 3 form three successful steps of our program to develop from scratch amathematical network communication model, prove it experimentally, and show that it can beapplied to develop algorithms solving hard problems related to design of communication schemesin networks.In the chapter 4 we show how under different network cost models, using some simplifyingassumptions on the structure of network and queries, one can design very efficient communicationschemes using simple combinatorial techniques. This work is complementary to the previous chapter in the sense that previously when designing communication schemes, we assumed atomicityof connections, i.e. that we have no control over routing of simple connections. In chapter 4 weshow how to solve the problem of an efficient routing of network request, given that we know thetopology of the network. It shows the importance of instantiating the parameters and the structureof the network in the context of designing efficient communication schemes.