University of Technology, Sydney. Faculty of Information Technology. / Over the last few years Peer-to-Peer (P2P) systems have emerged as highly attractive systems supporting many useful large-scale applications and services. They allow the exploitation of enormous untapped resources (such as idle processing cycles, storage, and bandwidth) available at Internet-connected devices, which were previously considered incapable of providing services to others. Participating nodes (peers) form an overlay network and communicate with each other without being controlled by a central authority. The structures and routing decisions of the most current P2P networks often do not correlate with the Internet infrastructure. In doing so, the tasks of overlay construction and routing become less complicated however, this results in high end-to-end delay for the P2P applications. As a consequence, the P2P networks may not be able to provide stringent Quality of Service (QoS) requirements for a new generation of P2P applications, and thus limit their benefits for the end users. Moreover, the infrastructure ignorance means P2P systems waste Internet resources by adding more than they should to the Internet traffic. This leads to the increase in Internet access costs substantially, and in turn the P2P systems do not scale well. The thesis presents a novel architecture for developing efficient P2P systems, and new schemes for constructing infrastructure-aware overlay networks. The main objective is first, to overcome the disparity between the overlay and Internet structures in order to maximize the use of network resources and reduce the overlay delay to the P2P applications; second, to provide efficient communication for P2P systems enabling deployment of any P2P applications while preserving decentralized, self-organizing and self-maintaining characteristics for the systems. To achieve these goals, we firstly developed Geographically Longest Prefix Matching (Geo-LPM) and Geographical Partitioning (Geo-Partitioning) schemes to cluster nodes that are close to each other in terms of network latency and network membership, and to determine links between neighboring clusters respectively. The developed schemes are efficient, generate low overhead, and help to produce excellent physically/infrastructure-aware overlay networks. Their distinctive features are self-organization, self-maintenance, and decentralization, which make them suitable to work in a P2P environment. Secondly we propose a novel architecture, called a physically-aware reference model (PARM) that captures desirable features for P2P systems by resolving major functional P2P system problems efficiently in a layered structure. For example, the application routing layer of PARM deals with routing inefficiency, meanwhile the infrastructure unawareness is resolved at the overlay network layer. We develop a useful P2P application, called a Peer Name Service (PNS) that interprets node names into their current IP addresses for any Internet-connected devices. Using the overlay networks, the PNS can support devices, which could be unreachable via the Domain Name Server (DNS), and mobile devices on-the-move without prior setup requirement in a distributed and timely fashion. Finally, to validate the whole concept of PARM, we simulate the PNS and a file transfer to a mobile node at the top layer of PARM, the P2P application layer. Since the PNS is sensitive to delay, it would be useful to evaluate the impacts of overlay delay factor and PARM on the performance of P2P applications. The simulation results show that the performance of the PARM-based applications is significantly improved while achieving decentralized and self-organizing features. The results also indicate that PARM can be a recommended reference model for developing scalable and efficient P2P systems.

Computing is increasingly pervading the activities of our everyday lives: at work, at home, and out on the town. When designing these pervasive systems there is a need to better understand and incorporate the context of use and yet there are limited empirical investigations into what constitutes this context. The user’s physical and social situation is an important part of their context when operating in an urban environment and thus needs to be understood and included in the interaction design of context-aware pervasive computing. This thesis has combined ideas from human computer interaction (HCI) and architecture to investigate indexicality in interface design as an instrument for incorporating physical and social context of the built environment into context-aware pervasive computing. Indexicality in interface design is a new approach to designing HCI for pervasive computing that relies on knowledge of current context to implicitly communicate between system and user. It reduces the amount of information that needs to be explicitly displayed in the interface while maintaining the usefulness and understandability of the communication.

Thesis (Ph.D.)--Electrical and Computer Engineering, Georgia Institute of Technology, 2008. / Committee Chair: Abler, Randal T.; Committee Member: Copeland, John A.; Committee Member: Haas, Kevin; Committee Member: Moore II, Elliot; Committee Member: Owen III, Henry L.

Thesis (Ph.D.)--Computing, Georgia Institute of Technology, 2009. / Committee Chair: Abowd, Gregory; Committee Member: Edwards, Keith; Committee Member: Grinter, Rebecca; Committee Member: LaMarca, Anthony; Committee Member: Starner, Thad.

No description available.

Aplicações cientes de contexto precisam de mecanismos para recuperar informações sobre o seu contexto de execução. Com base no contexto atual, tais aplicações são capazes de se autoadaptar para fornecer informações e serviços adequados aos seus usuários. A abordagem comum para infraestruturas de apoio às aplicações sensíveis ao contexto fornece serviços para a descoberta de recursos através da utilização de pares <chave-valor> e motores que executam apenas correspondência sintática. Esta abordagem não considera as possíveis relações semânticas entre as palavras-chave usadas. Portanto, a sua expressividade semântica limitada, leva a um serviço de descoberta que provê baixa taxa de recuperação e baixa acurácia. Este trabalho apresenta a utilização de uma outra abordagem para o serviço de contexto e descoberta, que utiliza ontologias para representar os recursos do contexto de execução e capturar a semântica da consulta do usuário, melhorando assim o processo de descoberta para a autoadaptação de sistemas sensíveis ao contexto. A abordagem proposta oferece também pontos de extensão para as aplicações clientes através da utilização de outras ontologias. Esta abordagem foi integrada à infraestrutura CDRF, de forma a adicionar semântica aos serviços desenvolvidos neste projeto. Exemplos de aplicações são também propostos para demonstrar a utilização dos novos serviços. / Context-aware applications demand ways of retrieving context information from the environment. Based on the current context, such applications are able to self-adapt to provide the correct information and services to its users. The usual approach for supporting infrastructures for context-aware applications provides facilities for resource discovery using <key-value> pairs and discovery engines that perform syntactic matching. This approach does not consider the possible semantic relations between the keywords used. So its limited semantic expressiveness often leads to poor discovery results. This paper presents the use of a different approach for service discovery that uses ontologies to represent resources and capture the semantics of the users query, improving the discovery process for self-adaptation of context-aware systems. The proposed approach also offers extension hooks to the client applications through the use of other ontologies. This approach is integrated into the CDRF framework and adds semantics to the services developed in that project. Example applications are also proposed to demonstrate the use of the new services.

Background: Unawareness of memory functioning is a key symptom of Alzheimer’s disease and dementia that has been demonstrated to be related to a number of important factors for the person with dementia (PwD) and their family caregivers including quality of life and depression. Understanding more about how awareness relates to these factors will help inform how PwD and their caregivers are best supported. Objective: A meta-analysis was conducted in order to examine the relationship between Awareness and depression in dementia. An empirical study was conducted to examine the contribution awareness provides to explaining PWDs’ Quality of Life (QoL). PwD have been found to be aware of factors that affect their caregiver and so caregiver wellbeing and quality of life and the quality of the caregiving relationship were also investigated as well as more established predictors of quality of life for PwD. Both PwD self-ratings and caregiver ratings of the PwD they care for of QoL were examined as they have been shown to be affected by different factors. Method: Meta-analysis: A search of electronic databases Psycinfo, Embase and Medline was conducted. A meta-analysis of correlations was undertaken examining the relationship between awareness and depression in dementia. Empirical study: 27 PwD and their caregivers were recruited. In order to assess the research aims the PwD completed measures of: Quality of life (Quality Of Life-Alzheimer’s Disease scale), awareness of memory functioning (Memory Awareness Rating Scale-Adjusted), cognitive functioning (Addenbrooke’s Cognitive Exam-R), depression and anxiety (Hospital Anxiety and Depression Scale). The caregiver completed measures of: PwD Quality of life (Quality Of Life-Alzheimer’s Disease scale proxy), Memory Functioning Scale (from MARSA), self-ratings of depression and anxiety (Hospital Anxiety and Depression Scale), activities of daily living (Disability Assessment in Dementia), Neuropsychiatric symptoms (Neuropsychiatric symptoms inventory-Questionnaire), caregiver burden (Zarit Burden Inventory), and rating of relationship quality with PwD (Burns Relationship Satisfaction Scale). Results: Meta-analysis: Thirty-one studies were identified. A small association was found between awareness and depression with substantial amount of heterogeneity (- 0.23). Analysing the studies that excluded major depression demonstrated that mild depression had a moderate negative relationship with awareness (-0.42). Subgroup analysis showed that the different measures of awareness used seemed to suggest different effects with depression for different measures. Empirical study: Awareness was not found to predict PwD rated or caregiver rated QoL. No caregiver variables predicted PwD QoL. Depression and neuropsychiatric symptoms predicted PwD QoL. Caregiver rated QoL was predicted by activities of daily living and caregiver rated quality of caregiving relationship. Conclusions: Meta-analysis: The effect between mild depression and lack of awareness but not major depression supports the assertion that unawareness is a psychological response to decline in memory functioning in dementia. Neither depression nor awareness appear to be unitary constructs in PwD. Empirical study: Awareness not related to PwD QoL. The quality of caregiving relationship is important to QoL in a dementia context. PwD and their caregivers rate the QoL of PwD differently.

Um dos principais desafios de pesquisa na UbiComp consiste em fornecer mecanismos para a ciência de contexto que promovam o desenvolvimento de aplicações que reajam de acordo com a dinâmica do ambiente de interesse do usuário. Para manter o conhecimento a respeito desse ambiente, a área da UbiComp pressupõe a utilização de informações produzidas e disponibilizadas em diferentes localizações, o tempo todo. Nesse sentido, os recentes avanços na área da Internet das Coisas (IoT) têm proporcionado uma crescente disponibilidade de sensores conectados em rede, os quais são potenciais produtores de informações contextuais do ambiente para aplicações ubíquas. Com essa motivação, nessa tese é apresentado o COIOT, um middleware para Internet das Coisas concebido com o objetivo de gerenciar a coleta e o processamento das informações contextuais do ambiente físico, bem como a atuação remota sobre o mesmo. O COIOT foi idealizado considerando os trabalhos previamente desenvolvidos pelo grupo de pesquisa GPPD (Grupo de Processamento Paralelo e Distribuído) da UFRGS, particularmente o middleware EXEHDA (Execution Environment for Highly Distributed Applications). Na concepção do COIOT foi adotada uma abordagem distribuída de processamento de contexto que contempla tanto as premissas da IoT quanto as demandas das aplicações da UbiComp. A arquitetura proposta também contempla o gerenciamento de eventos distribuídos através de regras e triggers para tratar as mudanças de estados dos contextos de interesse. Além disso, a arquitetura proposta gerencia outros aspectos importantes nos cenários da IoT, como o tratamento da interoperabilidade, da heterogeneidade, apoio ao controle da escalabilidade e descoberta de recursos. As principais contribuições desta tese são: (i) a concepção de uma arquitetura para IoT capaz de realizar de forma distribuída tanto a coleta e processamento das informações contextuais, como a atuação remota no meio a fim de atender as aplicações da UbiComp e, (ii) a proposição de um modelo de processamento de eventos distribuídos adequado aos cenários da IoT. Para avaliar a arquitetura do COIOT foram realizados dois estudos de caso na área da agricultura. O primeiro estudo de caso foi desenvolvido em ambiente de produção a partir de demandas de pesquisadores da área da agricultura, particularmente da análise de sementes. Já o segundo estudo de caso teve como cenário de testes ambientes da viticultura de precisão. / One of the main research challenges in UbiComp is to provide mechanisms for context-aware to promote the development of applications that react according to the dynamics of user interest environment. To keep the knowledge of this environment, the area of UbiComp presupposes the use of information produced and made available in different locations, all the time. In this sense, the recent advances in the field of Internet of Things (IoT) have provided an increasing availability of sensors and actuators networked. These sensors are potential producers of contextual information. With this motivation, this thesis is presented the CoIoT, a middleware for Internet of Things (IoT) designed in order to manage the collect and processing of contextual information of the physical environment as well as remote actuation on it. The CoIoT was designed considering the work previously developed by the research group GPPD (Parallel Processing Group and distributed) of UFRGS, particularly middleware EXEHDA (Execution Environment for Highly Distributed Applications). In designing the CoIoT it was adopted a distributed approach of context processing that includes both the principles of IoT as the demands of the applications of UbiComp. The proposed architecture also includes rules based and triggers mechanisms to deal with events that characterize the changes of states of the contexts of interest. In addition, the proposed architecture manages other important aspects of IoT scenarios such as the treatment of interoperability, heterogeneity, support the control of scalability and resource discovery. Until now, the central contributions of this thesis include: (i) the design of an architecture for IoT able to perform distributed way both the collect and processing of contextual information, such as remote actuation in the environment in order to meet UbiComp applications and, (ii) the proposition of a distributed event processing model appropriate to the IoT scenarios. In order to evaluate the CoIoT architecture, two case studies were carried out in the area of agriculture. The first case study was developed in a production environment based on the demands of agricultural researchers, particularly seed analysis. On the other hand, the second case study was based on precision testing of viticulture environments.

A Computação Pervasiva tem sido tema de diversos trabalhos nos últimos anos. Essa emergente área de pesquisa propõe uma visão de futuro onde serviços computacionais são oferecidos para os usuários através de inúmeros dispositivos espalhados pelo ambiente. Os serviços são disponibilizados, tanto através da infraestrutura existente dos computadores ligados fisicamente à rede quanto através de dispositivos móveis. Esse espalhamento da computação deve acontecer de maneira natural e imperceptível ao usuário. Dados pessoais, programas e arquivos de dados poderão ser acessados de qualquer lugar em qualquer momento. O poder de processamento será um recurso do ambiente, acessado quando necessário, da mesma forma que é hoje a eletricidade. O usuário não precisará ter ciência de qual máquina realiza o processamento necessário às suas aplicações, contanto que o resultado esperado seja obtido. Acredita-se que essa realidade será atingida através da aliança entre áreas de pesquisa como a Computação em Grade, Computação Móvel e a Computação Consciente do Contexto. A Computação Consciente do Contexto busca enriquecer a comunicação entre os seres humanos e os dispositivos computacionais, tornando sua atuação mais eficaz. As aplicações conscientes do contexto conseguem perceber as modificações que ocorrem no ambiente e adaptar seu comportamento ao novo estado. Esse processo pode ser dividido em três etapas: monitoramento, reconhecimento de contexto e adaptação. Na etapa de monitoramento são coletadas, através de sensores, informações sobre o ambiente. Essas informações, entretanto, são geralmente, de baixo nível de abstração e, portanto, dificilmente usadas diretamente por aplicações. A etapa de reconhecimento de contexto relaciona os dados obtidos do ambiente e transforma-os para que possam ser úteis às aplicações no processo de escolha do comportamento mais adequado à cada circunstância, habilitando a etapa de adaptação a efetivar a transformação do comportamento da aplicação de acordo com a nova situação do ambiente. Este trabalho propõe um servidor de contexto chamado MultiS que tem como objetivo a resolução dos problemas relativos à etapa de reconhecimento de contexto: a produção de dado de contexto baseado em informações de diversos sensores e a capacidade de reagir a modificações no ambiente. Também é proposta uma linguagem para composição de dados do contexto chamada CD-XML utilizada pelas aplicações para descrever ao servidor de contexto os dados aos quais elas são sensíveis. / The Pervasive Computing has been studied on several papers in the last years. This emergent research area presents a vision of future where computational services will be available through uncountable devices scattered across the environment. This service network will be exposed to the users by both traditional wired computers and mobile devices. This distribution of the computing is going to happen smoothly and transparently to the users. Personal data, computer programs, and data files will be available anywhere, anytime. The processing power will be an environment resource and will be accessed whenever needed, in the same way which is the electricity nowadays. The users will no longer need to worry about where their program is being executed, as long as he gets the needed result. The ISAM group believes that this new reality will be achieved through the alliance of research areas such as Grid Computing, Mobile Computing and Context-Aware Computing. The Context-Aware Computing aims to enrich the communication between human being and computer devices. Context-aware applications are capable of recognize the changes on the environment and adapt its own behavior to the new context state. This process can be divided in tree steps: monitoring, context recognition and adaptation. On the monitoring layer, environment information is collected from sensors. Those sensors, however, usually return only low level information, which is hardly used by the applications on its original form. The context recognition layer processes the data acquired from the context and transforms into information aimed to be useful to the adaptation process. With that information the adaptation system can identify the correct behavior for the application on each different context situation. This dissertation propose a context server named MultiS, which target is to solve the problems related to context recognition layer: the production of new context data based on the information of several sensors and the capability of react to changes on the environment. It also presents a new programming language for composition of contextual information, named CD-XML. This language is used by the context-aware applications to communicate to the context server describing which information the application is sensible to.

It is well acknowledged in the AI community that finding a good reward function for reinforcement learning is extremely challenging. Apprenticeship learning (AL) is a class of “learning from demonstration” techniques where the reward function of a Markov Decision Process (MDP) is unknown to the learning agent and the agent uses inverse reinforcement learning (IRL) methods to recover expert policy from a set of expert demonstrations. However, as the agent learns exclusively from observations, given a constraint on the probability of the agent running into unwanted situations, there is no verification, nor guarantee, for the learnt policy on the satisfaction of the restriction. In this dissertation, we study the problem of how to guide AL to learn a policy that is inherently safe while still meeting its learning objective. By combining formal methods with imitation learning, a Counterexample-Guided Apprenticeship Learning algorithm is proposed. We consider a setting where the unknown reward function is assumed to be a linear combination of a set of state features, and the safety property is specified in Probabilistic Computation Tree Logic (PCTL). By embedding probabilistic model checking inside AL, we propose a novel counterexample-guided approach that can ensure both safety and performance of the learnt policy. This algorithm guarantees that given some formal safety specification defined by probabilistic temporal logic, the learnt policy shall satisfy this specification. We demonstrate the effectiveness of our approach on several challenging AL scenarios where safety is essential.