Introdução: O adenocarcinoma gástrico é uma doença de elevada incidência e alta mortalidade. A gastrectomia com linfadenectomia é tratamento potencialmente curativo, promovendo controle loco-regional da doença e fornecendo material para análise histopatológica. Para o adequado estadiamento dos pacientes é recomendado que pelo menos 16 linfonodos sejam examinados pela patologia, entretanto, espera-se maior sobrevida quando >= 30 linfonodos são avaliados, mesmo em pacientes com tumores precoces. A justificativa para este achado é o sub-estadiamento de pacientes com poucos linfonodos examinados. Linfonodos pequenos são particularmente difíceis de serem encontrados, mas podem conter metástases e impactar negativamente na sobrevida. Visando facilitar sua identificação, soluções clareadoras de gordura foram propostas, entretanto não há evidência clara de seu benefício clínico. Objetivos: Comparar as soluções de Carnoy e de formalina neutra tamponada em relação ao número absoluto de linfonodos encontrados na peça cirúrgica de pacientes submetidos a gastrectomia. Averiguar se linfonodos retirados cirurgicamente são perdidos com a fixação em formalina e, caso isso ocorra, se este fato é relevante para o estadiamento. Observar se o protocolo de pesquisa influenciou o número de linfonodos encontrados. Métodos: Cinquenta produtos de gastrectomia subtotal com linfadenectomia D2 por adenocarcinoma gástrico foram randomizados para fixação em Carnoy ou formalina com posterior dissecção da peça em busca de linfonodos. Após a dissecção do grupo Formalina, a gordura residual a ser desprezada foi imersa em Carnoy e reavaliada posteriormente. Os dados de 25 gastrectomias D2 operadas previamente ao estudo também foram avaliados. Resultados: A média de linfonodos encontrados nos grupos Carnoy e Formalina foi de 50,4 e 34,8; respectivamente (p < 0,001). Na gordura residual foram encontrados linfonodos em todos os casos (média 16,9 linfonodos), elevando a média do grupo Formalina para 51,7 (valor similar ao do grupo Carnoy, p=0,809). Com exceção de 1 linfonodo de 7mm, todos os demais encontrados na gordura residual mediram <= 3mm. Treze linfonodos metastáticos passaram despercebidos com a fixação em formalina e a revisão da gordura residual determinou a mudança de estadiamento de 2 (8%) pacientes. Os linfonodos encontrados no Carnoy possuíam tamanho significativamente menor quando comparados aos do grupo Formalina (p=0,01). A média de linfonodos encontrados no grupo retrospectivo foi similar ao do grupo Formalina prospectivo (p=0,802). Conclusões: Quando comparada à formalina, a solução de Carnoy permite encontrar número maior de linfonodos no espécime cirúrgico de gastrectomias com linfadenectomia. Linfonodos milimétricos foram perdidos após a fixação em formalina, estes foram identificados com o Carnoy e são clinicamente relevantes, pois podem conter metástases modificando assim, o estádio clínico e prognóstico do paciente. A implementação de protocolo de pesquisa não influenciou o número de linfonodos encontrados neste estudo / Background: Gastric adenocarcinoma is a frequent disease with high mortality ratio. Gastrectomy with lymphadenectomy is potentially curative, allows local control of the disease and provides material for TNM classification. While pathology examination of at least 16 lymph nodes is recommended following surgery, longer survival rates are expected when >=30 lymph nodes are examined, even for early gastric cancer. The understaging of patients with less examined lymph nodes justifies this findings. Small lymph nodes are particularly difficult to identify and fat clearing solutions have been proposed to improve this, but there is no evidence of their clinical benefit. Objectives: Compare Carnoy\'s solution (CS) and formalin in terms of the total number of examined lymph nodes following gastrectomy. Verify if surgically retrieved lymph nodes are lost with the formalin fixation and if this fact is clinically significant. Observe if a research protocol influences the number of examined lymph nodes. Methods: Fifty specimens of gastrectomy with D2 lymphadenectomy were randomized for fixation in CS or formalin with posterior dissection in search for lymph nodes. In the Formalin group, the residual fat to be discarded was immersed in CS and dissected again. Data from 25 D2 gastrectomies performed previously the present study were retrospectively analyzed. Results: The medium number of examined lymph nodes was 50.4 and 34.8 for CS and formalin, respectively (p < 0.001). Lost lymph nodes were found in all cases in the Residual Fat group (medium 16.9), this increased the Formalin group average to 51.7 (which is similar to the CS group, p=0.809). With one exception (7mm), all other examined lymph nodes in the Residual Fat group measured <= 3mm. Thirteen lymph nodes from this group were metastatic, this determined the upstaging of 2 (8%) patients. Lymph nodes from the CS group were smaller than those found in the formalin group (p=0.01). The medium number of retrieved lymph nodes in the retrospective group was similar to the formalin group (p=0.802). Conclusions: When compared to formalin, Carnoy\'s solution increases lymph node detection following gastrectomy with lymphadenectomy. CS identifies small lymph nodes lost with formalin fixation and that are clinically significant, since they may contain metastasis, modifying the TNM classification. No influence of the research protocol over the number of examined lymph nodes was observed in the present study.

A crescente demanda de tráfego em redes de acesso pressiona a melhor utilização das redes backbone, que são utilizadas para transporte de grandes taxas de dados em diversos domínios (Sistemas Autônomos, SAs). Com o aumento destas redes, aumenta-se a complexidade de topologia das interligações entre domínios. Desta maneira, roteamento de tráfego e pontos de interconexão de SAs (nós de borda) são questões importantes para o desempenho destas redes, que são operadas por diversos provedores que podem utilizar protocolos de comunicação distintos. Neste sentido, o roteamento interdomínio apresenta desafios como a publicação ou não de informações de parâmetros de rede de SAs e como tratar esta questão de maneira globalizada, com novos protocolos e suas especificações. Em termos de pontos de interconexão de SAs, a especificação dos locais onde enlaces inter-redes são conectados aos domínios são importantes para seu desempenho, já que são responsáveis por toda troca de tráfego entre redes distintas. O trabalho considera redes ópticas opacas e translúcidas em cenário multidomínio com bandas multigranulares. Neste cenário é estudado um algoritmo de roteamento multidomínio. No trabalho também é feito um planejamento, especificando em quais nós serão conectados enlaces interdomínio. A principal contribuição deste trabalho é o estudo de planejamento de enlaces interdomínio, com a proposta de um método para escolha de nós de borda (sistematização), com objetivo de diminuir a probabilidade de bloqueio interdomínio. A sistematização é baseada em estudos de resultados de algoritmo genético desenvolvido para o mesmo propósito e sua utilização diminui em até 42% o bloqueio interdomínio. Um algoritmo de alocação de banda também foi desenvolvido para redes multidomínio, que considera parâmetros da camada de rede e óptica para o cálculo de peso de enlaces para encontrar caminhos ópticos entre nós fonte e destino. Os resultados mostram diminuição de até 35% no bloqueio interdomínio com a modificação feita em algoritmo proposto na literatura. / The huge demand for traffic in last mile networks push the better utilization of backbone networks, which are used to transport large data rates in several domains (Autonomous Systems, ASs). With this growth, the topology complexity of interdomain links increases. Then, traffic routing and interconnection points of ASs (border nodes) are relevant questions for the performance of these networks, which are managed by several providers that can use distinct communications protocols. Thus, the interdomain routing presents challenges such as the decision on publishing or not the network´s parameters from ASs and how to deal with this issue in a global way, with new protocols and its specifications. For interconnection points between ASs, the points where interdomain links are connected are important for their performances, since they are responsible for all traffic exchange between distinct networks. This work considers opaque and translucent optical networks in a multidomain scenario with multigranular data rates. In this scenario a multidomain routing algorithm is studied and a network planning is developed, specifying the nodes where interdomain links are connected. The main contribution of this work is the planning of interdomain links, with the proposal of a method for border nodes selection (systematization), with the objective of decreasing the interdomain blocking probability. The systematization is based on the results from a genetic algorithm developed for the same purpose and its utilization decrease up to 42% of the interdomain blocking. A bandwidth allocation algorithm was also created for multidomain scenarios, that considers parameters from network and optical layer for the link weight calculation in order to find optimal paths. The results show a decreasing of up to 35% for interdomain blocking with a contribution based on literature\'s work.

Orientador: Luiz Carlos Zeferino / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Ciencias Medicas / Made available in DSpace on 2018-08-12T08:55:33Z (GMT). No. of bitstreams: 1 Vieira_SabasCarlos_D.pdf: 1418388 bytes, checksum: 1d70cc451697cfbd93e48319a0284d4e (MD5) Previous issue date: 2008 / Resumo: Objetivos: Avaliar a detecção do linfonodo sentinela em pacientes com câncer do colo do útero utilizando a combinação de azul patente com tecnécio99m e complicações associadas ao uso do azul patente. Sujeitos e métodos: Este foi um estudo de uma série de casos, para o qual foram selecionadas 56 mulheres com diagnóstico de câncer do colo do útero estádios Ia2,Ib1,Ib2 e IIa da FIGO, que se submeteram ao procedimento de identificação do linfonodo sentinela. O período de realização do estudo foi de maio de 2006 a dezembro de 2007. O estudo é apresentado em dois artigos: o primeiro consiste na detecção do linfonodo sentinela no câncer do colo do útero pela combinação do azul patente com tecnécio 99m e avalia a concordância entre a linfocintigrafia pré-operatória e o mapeamento linfático intra-operatório com o gama probe; o segundo consiste na avaliação das alterações da oximetria de pulso das pacientes submetidas à cirurgia após a injeção do azul patente. Resultados: No primeiro artigo identificou-se pelo menos um linfonodo sentinela em 83,13% das pacientes e a localização mais freqüente destes linfonodos foi na cadeia ilíaca externa. A sensibilidade, especificidade, valor preditivo positivo e valor preditivo negativo foram, respectivamente, 80%, 100%, 100% e 97,67% no histopatológico de congelação. Além disso, observou-se que linfocintigrafia pré operatória detecta um número consideravelmente menor de linfonodos sentinelas quando comparado ao mapeamento linfático intra-operatório com o gama probe. No segundo estudo observou-se que somente uma paciente apresentou reação anafilática. Treze pacientes apresentaram queda de oximetria de pulso (menor que 96% de saturação) após a injeção do azul patente no colo do útero, que durou em média cinco minutos e sem repercussões clínicas; essa queda se associou de forma limítrofe com tumores maiores e localizados ao redor do orifício externo do canal cervical. Conclusões: Concluiu-se que a combinação do azul patente com o tecnécio99m demonstrou excelentes resultados na detecção do linfonodo sentinela; a linfocintigrafia pré-operatória não oferece qualquer vantagem em relação ao mapeamento linfático intra-operatório com azul patente e tecnécio99m. Não houve repercussões clínicas devido à queda da oximetria de pulso e essas alterações se correlacionaram, embora com significância limítrofe, com tumores maiores e localizados ao redor do orifício cervical externo. / Abstract: Objectives: To evaluate sentinel lymph node detection in cervical cancer patients using a combination of patent blue dye and technetium99m and assess complications associated with the use of patent blue dye. Subjects and methods: This study investigated a case series that selected 56 women diagnosed with FIGO stage Ia2, Ib1, Ib2 and IIa cervical cancer who underwent a procedure for sentinel lymph node identification. The study was conducted from May 2006 to December 2007 and was described in two articles. The first article was about a study of sentinel lymph node detection in cervical cancer using a combination of patent blue dye and technetium99m. It assessed the agreement between preoperative lymphoscintigraphy and intraoperative lymphatic mapping with a gamma probe. The second article focused on the evaluation of changes in pulse oximetry readings in patients undergoing surgery after patent blue injection. Results: In the first article, at least one sentinel lymph node was identified in 83.13% of the patients and the most frequent site for finding sentinel lymph nodes was the external iliac chain. The sensitivity, specificity, positive predictive value and negative predictive value were 80%, 100%, 100% and 97.67% respectively on histopathology examination of frozen biopsy. In addition, it was observed that preoperative lymphoscintigraphy detected a substantially lower number of sentinel lymph nodes when compared to intraoperative lymphatic mapping with a gamma probe. In the second study, only one patient presented with an anaphylactic reaction. Thirteen patients showed a decrease in pulse oximetry readings (less than 96% saturation) after patent blue injection into the cervix, which lasted an average of five minutes and had no clinical repercussions. There was a borderline association between this decline in oxygen saturation values and tumors that were larger and located around the external cervical os. Conclusions: It was concluded that a combination of patent blue dye and technetium99m demonstrated excellent results in the detection of sentinel lymph nodes. Preoperative lymphoscintigraphy offers no advantage in relation to intraoperative lymphatic mapping with patent blue dye and technetium99m. There were no clinical repercussions due to lower oxygen saturation values. These changes correlated with tumors that were larger and located around the external cervical os, although the significance of this correlation was borderline. / Doutorado / Ciencias Biomedicas / Doutor em Tocoginecologia

In this thesis, we consider stochastic systems which arise in different real-world application contexts. The first problem we consider is based on product adoption and pricing. A monopolist selling a product has to appropriately price the product over time in order to maximize the aggregated profit. The demand for a product is uncertain and is influenced by a number of factors, some of which are price, advertising, and product technology. We study the influence of price on the demand of a product and also how demand affects future prices. Our approach involves mathematically modelling the variation in demand as a function of price and current sales. We present a simulation-based algorithm for computing the optimal price path of a product for a given period of time. The algorithm we propose uses a smoothed-functional based performance gradient descent method to find a price sequence which maximizes the total profit over a planning horizon. The second system we consider is in the domain of sensor networks. A sensor network is a collection of autonomous nodes, each of which senses the environment. Sensor nodes use energy for sensing and communication related tasks. We consider the problem of finding optimal energy sharing policies that maximize the network performance of a system comprising of multiple sensor nodes and a single energy harvesting(EH) source. Nodes periodically sense a random field and generate data, which is stored in their respective data queues. The EH source harnesses energy from ambient energy sources and the generated energy is stored in a buffer. The nodes require energy for transmission of data and and they receive the energy for this purpose from the EH source. There is a need for efficiently sharing the stored energy in the EH source among the nodes in the system, in order to minimize average delay of data transmission over the long run. We formulate this problem in the framework of average cost infinite-horizon Markov Decision Processes[3],[7]and provide algorithms for the same.

Prostate cancer (PC) is the most common cancer in Sweden. Most patients have slow growing tumors that will not cause them any harm within their lifetime, but some have aggressive tumors and will die from their disease. The ability of current clinical practice to predict tumor behavior and disease outcome is limited leading to both over- and undertreatment of PC patients. The men who die from their disease are those that develop metastases. It is therefore of great value to find better and more sensitive prognostic techniques, so that metastatic spread can be detected (or predicted) at an early time point, and so that appropriate treatment can be offered to each subgroup of patients. The aim of this thesis was to investigate if, and by what means, highly aggressive prostate tumors influence extratumoral tissues such as the non-malignant parts of the prostate and regional lymph nodes (LN), and also if any of our findings could be of prognostic importance. Gene- and protein expression analysis were the main methods used to address these questions. Our research group has previously introduced the expression Tumor Instructed (Indicating) Normal Tissue (TINT), and we use the term TINT-changes when referring to alterations in non-malignant tissue due to the growth of a tumor nearby or elsewhere in the body. In the Dunning rat PC-model we found that MatLyLu (MLL)-tumors, having a high metastatic ability, caused pre-metastatic TINT-changes that differ from those caused by AT1-tumors who have low metastatic ability. Prostate-TINT surrounding MLL-tumors had elevated immune cell infiltration, and gene ontology enrichment analysis suggested that biological functions promoting tumor growth and metastasis were activated in MLL- while inhibited in AT1-prostate-TINT. In the regional LNs we found signs of impaired antigen presentation, and decreased quantity of T cells in the MLL-model. One of the downregulated genes in the MLL-LNs was Siglec1 (also known as Cd169), expressed by LN resident macrophages that are important for antigen presentation. When examining metastasis-free LN tissue from PC patients we found CD169 expression to be a prognostic factor for PC-specific survival, and reduced expression was linked to an increased risk of PC-specific death. Some of our findings in prostate- and LN-TINT could be seen already when the tumors were very small suggesting that differences in TINT-changes between tumors with different metastatic capability can be detected early in tumor progression. However, before coming of use in the clinic more research is needed to better define a suitable panel of prognostic TINT-factors as well as the right time window of when to use them. / Populärvetenskaplig sammanfattning Prostatacancer är den i särklass vanligaste cancerformen hos män i Sverige. De flesta patienter har en mycket långsamt växande tumör som inte orsakar dem några större besvär under deras livstid, men enbart i Sverige dör ca 2500 patienter/år av sjukdomen. Det är först vid uppkomst av metastaser som sjukdomen blir dödlig. Befintliga diagnos- och prognosmetoder är otillräckliga när det gäller att uppskatta och förutse tumörens aggressivitet och risk för att bilda metastaser. Detta gör att vissa patienter inte får tillräcklig behandling eller behandlas försent medan andra behandlas i onödan. Behovet av förbättrad diagnostik är därför stort. Om vi kan hitta markörer för potentiellt metastaserande sjukdom, och i bästa fall också behandla innan metastaser uppstår, skulle det förbättra chansen för överlevnad markant. För att kunna växa och spridas behöver en tumör inte bara förbereda närliggande vävnader utan förmodligen hela kroppen. Vår hypotes är att potentiell dödliga tumörer sannolikt är bättre på detta än mer ofarliga. Man vet från studier av andra cancerformer att farliga tumörer orsakar förändringar i det organ dit cancern senare sprids. Dessa förändringar sker för att de tumörceller som senare anländer ska kunna överleva, och processen har fått namnet pre-metastatisk nisch. Bl.a. har man sett att immunsystemet hämmas och nybildning av kärl ökar. Det är vanligt att metastaser uppstår i närliggande lymfkörtlar innan uppkomst av metastaser i andra organ. Dock är väldigt lite känt om pre-metastatiska förändringar i lymfkörtlar eftersom den forskning som hittills är gjord främst har tittat på andra organ. Inom prostatacancer finns det förvånande få studier av premetastatiska nischer överhuvudtaget, och man vet därför inte om de alls förekommer eller vilka förändringar som i så fall sker. Vår grupp har tidigare myntat uttrycket TINT som står för Tumor Instructed (Indicating) Normal Tissue (TINT är ett engelskt verb som betyder färga) och syftar på förändringar i normal vävnad som inducerats av tumören, dvs. att tumörer färgar av sig på omgivningen. Det kan vara förändringar i normal vävnad nära tumören, som i det här fallet resten av prostatan, eller i vävnad långt ifrån tumören som till exempel regionala lymfkörtlar, lungor och benmärg. Syftet med det här avhandlingsarbetet var att undersöka TINT-förändringar inducerade av aggressiv cancer och se om dessa skiljer sig från TINT-förändringar inducerade av mindre farliga tumörer, samt att utvärdera om någon TINT-förändring skulle kunna användas för att prognostisera vilka patienter som har hög risk att få metastaser. Vi har använt oss av en prostatacancer-modell i råtta där vi analyserat genoch proteinuttryck i pre-metastatiska regionala lymfkörtlar, tumörer och prostata-TINT (dvs. prostatavävnad utanför tumören). TINT-förändringar inducerade av MatLyLu (MLL), en tumör med hög metastaserande förmåga, jämfördes mot TINT-förändringar inducerade av AT1, en snabbväxande tumör men med låg förmåga att bilda metastaser. Vi kunde vi se flera skillnader mellan modellerna. Genuttrycket i MLL-prostata-TINT indikerade en aktivering av cellulära funktioner som visat sig stimulera tumörväxt och spridning såsom celldelning, viabilitet, migration, invasion, och angiogenes (nybildning av kärl). I AT1-prostata-TINT var genuttrycket kopplat till samma funktioner men verkade istället inhibera dessa. Genom att titta på vävnaderna i mikroskop kunde vi se att MLL-tumörer rekryterade färre T-celler (som har en viktig funktion i immunsvaret mot tumören), men istället fler makrofager och granulocyter till både tumören och prostata-TINT (dessa typer av immunceller har visats kunna hjälpa tumörer att växa och sprida sig). MLL-tumörer hade också fler blodkärl och lymfkärl strax utanför tumören. I de regionala lymfkörtlarna från djur med MLL-tumörer visade genuttrycket tecken på försämrad antigenpresentation, samt immunhämning och/eller induktion av immuntolerans. Immuntolerans innebär att immuncellen inte längre reagerar mot det specifika antigen den blivit tolerant emot. Detta är vanligt förekommande hos individer med cancer och är ett sätt för tumören att undkomma immunförsvaret. I vävnadsprover av lymfkörtlarna kunde vi se färre antigenpresenterande celler, och liksom i tumörerna fanns det färre T-celler i MLL-modellen, något vi kunde se redan när tumörerna var väldigt små. CD169 är ett protein som bl.a. uttrycks av sinus-makrofager i lymfkörtlar. Dessa makrofager har en central funktion i att aktivera ett tumör-specifikt immunsvar. I råttmodellen kunde vi se att regionala lymfkörtlar från djur med MLL-tumörer hade lägre nivåer av CD169 än regionala lymfkörtlar från djur med AT1-tumörer, och då antalet sinus-makrofager visat sig ha prognostiskt värde i t.ex. tjocktarmscancer, ville vi se om det kunde vara så även i prostatacancer. Därför kvantifierade vi uttrycket av CD169 i metastasfria regionala lymfkörtlar från prostatacancerpatienter och såg att låga nivåer av CD169 medförde en ökad risk för att dö i prostatacancer. Sammantaget tyder resultaten på att MLL-tumören jämfört med AT1- tumören bättre lyckas förbereda omgivande vävnad för att gynna tumörväxt och spridning, både lokalt i prostatan men också längre bort från tumören i de regionala lymfkörtlarna. Våra fynd stämmer väl överens med aktuell tumörbiologisk forskning om hur tumörer påverkar sin omgivning. Något som inte visats tidigare är att miljön utanför tumören verkar skilja sig drastiskt beroende på tumörens metastaserande förmåga, samt att dessa skillnader går att se relativt tidigt under sjukdomsförloppet och förmodligen även långt bort från tumören. Vi har också visat att särskilt aggressiv prostatacancer verkar inducera en pre-metastatisk nisch i tumördränerande lymfkörtlar likt det som beskrivits i andra modellsystem och i andra cancertyper, men hittills inte i prostatacancer. Fler studier behövs för att bättre karaktärisera de förändringar som en potentiellt dödlig prostatacancer orsakar i andra vävnader, och för att ta reda på hur denna kunskap kan användas för att förbättra diagnostik och behandling.

Sensor nodes in a wireless sensor network (WSN) have a small, non-rechargeable power supply. Each message transmission or reception depletes a sensor node’s energy. Many WSN applications are ad-hoc deployments where a sensor node is only aware of its immediate neighbours. The lack of a predefined route path and the need to restrict the amount of communication that occurs within the application area impose constraints on WSNs not prevalent in other types of networks. An area of active research has been how to notify the central sink (or monitoring hub) about an event in real-time by utilising the minimum number of messages to route a message from a source node to the destination sink node. In this thesis, strategies to limit communication within a WSN application area, while ensuring that events are reported on and responded to in real-time, is presented. A solution based on modelling a WSN as a small world network and then transmitting an initialisation message (IM) on network start-up to create multiple route paths from any sensor node to one or more sinks is proposed. The reason for modelling a WSN as a small world network is to reduce the number of nodes required to re-transmit a message from a source sensor node to a sink. The purpose of sending an IM at network start-up is to ensure that communication within the WSN is minimised. When routing a message to a static sink, the nodes closest to the static sink receive a disproportionate number of messages, resulting in their energy being consumed earlier. The use of mobile sinks has been proposed but to our knowledge no studies have been undertaken on the paths these mobile sinks should follow. An algorithm to determine the optimum path for mobile sinks to follow in a WSN application area is described. The purpose of an optimum path is to allow more equitable usage of all nodes to transfer an event message to a mobile sink. The idea of using multiple static sinks placed at specific points in the small world model is broadened to include using multiple mobile sinks called actors to move within a WSN application area and respond to an event in real-time. Current coordination solutions to determine which actor(s) must respond to the event result in excessive message communication and limit the real-time response to an event. An info gap decision theory (IGDT) model to coordinate which actor or set of actors should respond to the event is described. A comparison of the small world routing (SWR) model against routing using flooding and gossiping shows that the SWR model significantly reduces the number of messages transmitted within the network. An analysis of the number of IMs transmitted and received at individual node level shows that prudent selection of the hop count (number of additional nodes required to route a message to sink) to a sink node will result in a reduced number of messages transmitted and received per node within the network. The use of the IGDT model results in a robust decision on the actor(s) chosen to respond to an event even when uncertainty about the location and available energy of other actor(s) exists. / Thesis (PhD(Eng))--University of Pretoria, 2012. / Electrical, Electronic and Computer Engineering / unrestricted

Wireless Sensor Networks (WSN) may consist of hundreds or even thousands of nodes and could be used for a multitude of applications such as warfare intelligence or to monitor the environment. A typical WSN node has a limited and usually irreplaceable power source and the efficient use of the available power is of utmost importance to ensure maximum lifetime of each WSN application. Each of the nodes needs to transmit and communicate sensed data to an aggregation point for use by higher layer systems. Data and message transmission among nodes collectively consume the largest amount of the energy available in a WSN. The network routing protocols ensure that every message reaches the destination and has a direct impact on the amount of transmissions to deliver a messages successfully. To this end the transmission protocol within the WSN should be scalable, adaptable and optimized to consume the least possible amount of energy to suite different network architectures and application domains. This dissertation proposes a Mobile Tolerant Hybrid Energy Efficient Routing Protocol (MT-HEER), where hybrid refers to the inclusion of both flat and hierarchical routing architectures as proposed by Page in the Hybrid Energy Efficient Routing Protocol (HEER). HEER was previously developed at the University of Pretoria and forms the starting point of this research. The inclusion of mobile nodes in the WSN deployment proves to be detrimental to protocol performance in terms of energy efficiency and message delivery. This negative impact is attributable to assuming that all nodes in the network are statically located. In an attempt to adapt to topological changes caused by mobile nodes, too much energy could be consumed by following traditional network failure algorithms. MT-HEER introduces a mechanism to pro-actively track and utilise mobile nodes as part of the routing strategy. The protocol is designed with the following in mind: computational simplicity, reliability of message delivery, energy efficiency and most importantly mobility awareness. Messages are propagated through the network along a single path while performing data aggregation along the same route. MT-HEER relies on at least 40% of the nodes in the network being static to perform dynamic route maintenance in an effort to mitigate the risks of topological changes due to mobile nodes. Simulation results have shown that MT-HEER performs as expected by preserving energy within acceptable limits, while considering the additional energy overhead introduced by dynamic route maintenance. Mobile node tolerance is evident in the protocol's ability to provide a constant successful message delivery ratio at the sink node with the introduction and increase in the number of mobile nodes. MT-HEER succeeds in providing tolerance to mobile nodes within a WSN while operating within acceptable energy conservation limits. AFRIKAANS : Koordlose Sensor Netwerke mag bestaan uit honderde of selfs duisende nodes en kan gebruik word vir 'n legio van toepassings soos oorlogs intellegensie of om die omgewing te monitor. 'n Tipiese node in so 'n netwerk het 'n beperkte en soms onvervangbare energie bron. Die effektiewe gebruik van die beskikbare energie is dus van uiterste belang om te verseker dat die maksimum leeftyd vir 'n koordlose sensor network behaal kan word. Elkeen van die nodes in the network moet die waargeneemde data aanstuur oor die netwerk na 'n versamelings punt vir latere gebruik deur applikasie vlak stelsels. Informasie en boodskap transmissie tussen die nodes is wel een van die aktiwiteite wat die meeste energie verbruik in the netwerk. Die roeterings protokol verseker dat die boodskappe die eindbestemming behaal en het 'n direkte impak op die hoeveelheid transmissies wat kan plaas vind om dit te bewerkstellig. Die roeterings protokol moet dus skaleerbaar, aanpasbaar en verfyn word om die minste moontlike energie te verbruik in verskillende toepassings velde. Hierdie verhandeling stel 'n Bewegings Tolerante Hybriede Netwerk Roeterings Protokol vir Koordlose Sensor Netwerke (“MT-HEER”) voor. In hierdie konteks verwys hybried na die samesmelting van beide plat en hierargiese roeterings beginsels soos voor gestel deur Page in Hybriede Netwerk Roeterings Protokol (“HEER”). HEER was ontwikkel by die Universiteit van Pretoria en vorm die begin punt van hierdie navorsing. Die insluiting van bewegende nodes in 'n Koordlose Sensor Netwerk toon 'n negatiewe tendens in terme van energie effektiwiteit en suksesvolle boodskap aflewerings by die eindbestemming. Die grootste rede vir hierdie negatiewe tendens is die aanname deur gepubliseerde werke dat alle nodes in die netwerk staties is. Te veel energie sal vermors word indien tradisionele fout korregerende meganismes gevolg word om aan te pas by die bewegende nodes. MT-HEER stel 'n meganisme voor om die bewegende nodes te gebruik as deel van die roetering strategie en gevolglik ook hierdie nodes te volg soos hulle beweeg deur die netwerk. Die protokol is ontwikkel met die volgende doelstellings: rekenkundig eenvoudigheid, betroubare boodskap aflewering, energie effektiwiteit en bewustheid van bewegende nodes. Boodskappe word langs 'n enkele pad gestuur deur die netwerk terwyl boodskap samevoeging bewerkstellig word om die eind bestemming te bereik. MT-HEER vereis wel dat ten minste 40% van die netwerk nodes staties bly om die dienamiese roeterings instandhouding te bewerkstellig. Simulasie toetse en resultate het bewys dat MT-HEER optree soos verwag in gevalle waar daar bewegende nodes deel uit maak van die netwerk. Energie bewaring is binne verwagte parameters terwyl die addisionele energie verbruik binne rekening gebring word om te sorg vir bewegende nodes. Die protokol se toleransie teen bewegende nodes word ten toon gestel deur die vermoë van die protokol om konstant 'n hoë suksesvolle bookskap aflewerings verhouding te handhaaf. MT-HEER behaal die uitgesette doel om 'n toleransie teen bewegende nodes beskikbaar te stel, terwyl die protokol steeds funksioneer binne verwagte energie besparings limiete. Copyright / Dissertation (MEng)--University of Pretoria, 2010. / Electrical, Electronic and Computer Engineering / unrestricted

New PCI-e flash cards and SSDs supporting over 100,000 IOPs are now available, with several usecases in the design of a high performance storage system. By using an array of flash chips, arranged in multiple banks, large capacities are achieved. Such multi-banked architecture allow parallel read, write and erase operations. In a raw PCI-e flash card, such parallelism is directly available to the software layer. In addition, the devices have restrictions such as, pages within a block can only be written sequentially. The devices also have larger minimum write sizes (>4KB). Current flash translation layers (FTLs) in Linux are not well suited for such devices due to the high device speeds, architectural restrictions as well as other factors such as high lock contention. We present a FTL for Linux that takes into account the hardware restrictions, that also exploits the parallelism to achieve high speeds. We also consider leveraging the parallelism for garbage collection by scheduling the garbage collection activities on idle banks. We propose and evaluate an adaptive method to vary the amount of garbage collection according to the current I/O load on the device. For large scale distributed storage systems, flash memories are an excellent choice because flash memories consume less power, take lesser floor space for a target throughput and provide faster access to data. In a traditional distributed filesystem, even distribution is required to ensure load-balancing, balanced space utilisation and failure tolerance. In the presence of flash memories, in addition, we should also ensure that the numbers of writes to these different flash storage nodes are evenly distributed, to ensure even wear of flash storage nodes, so that unpredictable failures of storage nodes are avoided. This requires that we distribute updates and do garbage collection, across the flash storage nodes. We have motivated the distributed wearlevelling problem considering the replica placement algorithm for HDFS. Viewing the wearlevelling across flash storage nodes as a distributed co-ordination problem, we present an alternate design, to reduce the message communication cost across participating nodes. We demonstrate the effectiveness of our design through simulation.

Many sequential decision making problems under uncertainty arising in engineering, science and economics are often modelled as Markov Decision Processes (MDPs). In the setting of MDPs, the goal is to and a state dependent optimal sequence of actions that minimizes a certain long-term performance criterion. The standard dynamic programming approach to solve an MDP for the optimal decisions requires a complete model of the MDP and is computationally feasible only for small state-action MDPs. Reinforcement learning (RL) methods, on the other hand, are model-free simulation based approaches for solving MDPs. In many real world applications, one is often faced with MDPs that have large state-action spaces whose model is unknown, however, whose outcomes can be simulated. In order to solve such (large) MDPs, one either resorts to the technique of function approximation in conjunction with RL methods or develops application specific RL methods. A solution based on RL methods with function approximation comes with the associated problem of choosing the right features for approximation and a solution based on application specific RL methods primarily relies on utilizing the problem structure. In this thesis, we investigate the problem of choosing the right features for RL methods based on function approximation as well as develop novel RL algorithms that adaptively obtain best features for approximation. Subsequently, we also develop problem specie RL methods for applications arising in the areas of wireless sensor networks and road traffic control. In the first part of the thesis, we consider the problem of finding the best features for value function approximation in reinforcement learning for the long-run discounted cost objective. We quantify the error in the approximation for any given feature and the approximation parameter by the mean square Bellman error (MSBE) objective and develop an online algorithm to optimize MSBE. Subsequently, we propose the first online actor-critic scheme with adaptive bases to find a locally optimal (control) policy for an MDP under the weighted discounted cost objective. The actor performs gradient search in the space of policy parameters using simultaneous perturbation stochastic approximation (SPSA) gradient estimates. This gradient computation however requires estimates of the value function of the policy. The value function is approximated using a linear architecture and its estimate is obtained from the critic. The error in approximation of the value function, however, results in sub-optimal policies. Thus, we obtain the best features by performing a gradient descent on the Grassmannian of features to minimize a MSBE objective. We provide a proof of convergence of our control algorithm to a locally optimal policy and show numerical results illustrating the performance of our algorithm. In our next work, we develop an online actor-critic control algorithm with adaptive feature tuning for MDPs under the long-run average cost objective. In this setting, a gradient search in the policy parameters is performed using policy gradient estimates to improve the performance of the actor. The computation of the aforementioned gradient however requires estimates of the differential value function of the policy. In order to obtain good estimates of the differential value function, the critic adaptively tunes the features to obtain the best representation of the value function using gradient search in the Grassmannian of features. We prove that our actor-critic algorithm converges to a locally optimal policy. Experiments on two different MDP settings show performance improvements resulting from our feature adaptation scheme. In the second part of the thesis, we develop problem specific RL solution methods for the two aforementioned applications. In both the applications, the size of the state-action space in the formulated MDPs is large. However, by utilizing the problem structure we develop scalable RL algorithms. In the wireless sensor networks application, we develop RL algorithms to find optimal energy management policies (EMPs) for energy harvesting (EH) sensor nodes. First, we consider the case of a single EH sensor node and formulate the problem of finding an optimal EMP in the discounted cost MDP setting. We then propose two RL algorithms to maximize network performance. Through simulations, our algorithms are seen to outperform the algorithms in the literature. Our RL algorithms for the single EH sensor node do not scale when there are multiple sensor nodes. In our second work, we consider the problem of finding optimal energy sharing policies that maximize the network performance of a system comprising of multiple sensor nodes and a single energy harvesting (EH) source. We develop efficient energy sharing algorithms, namely Q-learning algorithm with exploration mechanisms based on the -greedy method as well as upper confidence bound (UCB). We extend these algorithms by incorporating state and action space aggregation to tackle state-action space explosion in the MDP. We also develop a cross entropy based method that incorporates policy parameterization in order to find near optimal energy sharing policies. Through numerical experiments, we show that our algorithms yield energy sharing policies that outperform the heuristic greedy method. In the context of road traffic control, optimal control of traffic lights at junctions or traffic signal control (TSC) is essential for reducing the average delay experienced by the road users. This problem is hard to solve when simultaneously considering all the junctions in the road network. So, we propose a decentralized multi-agent reinforcement learning (MARL) algorithm for solving this problem by considering each junction in the road network as a separate agent (controller) to obtain dynamic TSC policies. We propose two approaches to minimize the average delay. In the first approach, each agent decides the signal duration of its phases in a round-robin (RR) manner using the multi-agent Q-learning algorithm. We show through simulations over VISSIM (microscopic traffic simulator) that our round-robin MARL algorithms perform significantly better than both the standard fixed signal timing (FST) algorithm and the saturation balancing (SAT) algorithm over two real road networks. In the second approach, instead of optimizing green light duration, each agent optimizes the order of the phase sequence. We then employ our MARL algorithms by suitably changing the state-action space and cost structure of the MDP. We show through simulations over VISSIM that our non-round robin MARL algorithms perform significantly better than the FST, SAT and the round-robin MARL algorithms based on the first approach. However, on the other hand, our round-robin MARL algorithms are more practically viable as they conform with the psychology of road users.

This thesis proposes and analyzes data fusion algorithms that operate on the physical layer of a wireless sensor network, in the context of three applications of cognitive radios: 1. Cooperative spectrum sensing via binary consensus; 2. Multiple transmitter localization and communication footprint identification; 3.Target self-localization using beacon nodes. For the first application, a co-phasing based data combining scheme is studied under imperfect channel knowledge. The evolution of network consensus state is modeled as a Markov chain, and the average transition probability matrix is derived. Using this, the average hitting time and average consensus duration are obtained, which are used to determine and optimize the performance of the consensus procedure. Second, using the fact that a typical communication footprint map admits a sparse representation, two novel compressed sensing based schemes are proposed to construct the map using 1-bit decisions from sensors deployed in a geographical area. The number of transmitters is determined using the K-means algorithm and a circular fitting technique, and a design procedure is proposed to determine the power thresholds for signal detection at sensors. Third, an algorithm is proposed for self-localization of a target node using power measurements from beacon nodes transmitting from known locations. The geographical area is overlaid with a virtual grid, and the problem is treated as one of testing overlapping subsets of grid cells for the presence of the target node. The column matching algorithm from group testing literature is considered for devising the target localization algorithm. The average probability of localizing the target within a grid cell is derived using the tools from Poisson point processes and order statistics. This quantity is used to determine the minimum required node density to localize the target within a grid cell with high probability. The performance of all the proposed algorithms is illustrated through Monte Carlo simulations.