Stepping into the 21st century, we see more and more evidence of the growing trend towards the amalgamation of cyberspace and the physical world. This trend emerged as computing technologies moved o_ desktops and migrated into aspects of our lives through their ubiquitous presence in the physical world. As these technologies become enmeshed in our daily routines, they begin to `disappear' from our awareness and cease to be thought of as technologies and simply become tools of everyday use. Yet even as they disappear, these technologies afford a new way for us to interact with the environments of everyday life and with the ordinary objects within these environments. The furthering of this vision will require, in many cases, the tools and applications to possess greater levels of autonomy and an awareness of the user's context. As a result, the applications gradually depend more and more for their behaviour on the information (context information) that is relevant to user interactions. However, it is difficult to develop new context-aware applications that take into account the ever-increasing amount of context information. This is because: the context information sources vary not only in their types, but also in their availability in different environments; the developers have to spend significant programming efforts in gathering, pre-processing and managing the context information when designing and developing the new applications; and, the information sources can fail from time to time, resulting in operational disruptions or service degradation. To make such context information easily and widely available for to new context-aware applications, there is a need to provide information provisioning and management at the infrastructure level. This thesis explores the issues and challenges associated with the development of an autonomic middleware system that addresses the problems discussed earlier, with a particular focus on supporting fault-tolerant context information provisioning for multiple applications, providing the support of opportunistic use of the context sources (the sensors) and, maximising overall the system's interoperability for the open, dynamic computing environments (Ubiquitous computing, for example). The research presented in this thesis makes several key contributions. First, it introduces a novel standards-based approach to model heterogeneous information sources and data preprocessing components. Second, it details the design of a standards-based approach for supporting the dynamic composition of context information sources and pre-processing components. This approach plays an important role in supporting fault-tolerant information provisioning from the sensors and the opportunistic use of these sensors. More specifically, it enables any given piece of high-level context information, as required by applications, to be derived via multiple different pre-processing models, resulting in a higher degree of reliability. Third, it describes the design and development of an autonomic context management system (ACoMS), which harnesses the first two contributions above. Finally, the thesis shows how this autonomic context management system can support context-aware routing in wireless mesh networks. These contributions are evaluated through two corresponding case studies. The first is a practical firefighting scenario with three prototypical applications that validate the design and development of ACoMS. The second is an adaptive wireless mesh surveillance camera system that validates the concept of adopting ACoMS as a cross-layer information plane to ease the prototyping and development of new adaptive protocols and systems, and illustrates the needs of adaptive controls at the sensing layer to optimise resource usage.

Submitted by Cássia Santos (cassia.bcufg@gmail.com) on 2016-09-08T13:52:43Z No. of bitstreams: 2 Dissertação - Ernesto Fonseca Veiga - 2016.pdf: 3105414 bytes, checksum: 404f9ee6b395253e73e4c1b1fefe6d67 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2016-09-09T14:31:54Z (GMT) No. of bitstreams: 2 Dissertação - Ernesto Fonseca Veiga - 2016.pdf: 3105414 bytes, checksum: 404f9ee6b395253e73e4c1b1fefe6d67 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2016-09-09T14:31:54Z (GMT). No. of bitstreams: 2 Dissertação - Ernesto Fonseca Veiga - 2016.pdf: 3105414 bytes, checksum: 404f9ee6b395253e73e4c1b1fefe6d67 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2016-07-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Context Management Systems (CMS) are responsible for processing the context throughout its life cycle, which is performed in four steps defined in the literature: acquisition, modeling, reasoning and dissemination. The modeling step has great relevance in this process, being responsible for transforming the data acquired from sensors in highlevel context. However, this step has been impacted by the increasing deployment of sensors and the large volume of information that is available, creating the need to provide services to carry out the handling of such information, providing adequate representation and supporting the development of applications. In this sense, the ontology-based representation has major advantages, including: high expressiveness, standardization and interoperability. Before highlighted requirements by the literature, this work presents the construction of services related to the context life cycle modeling step: representation, aggregation and history, based on ontologies. These services have been designed and implemented as components that integrated a CMS called Hermes. Therefore, the scope of this work is composed by: i) Hermes Widget: ontological representation of independent domain context, based on Stimulus-Sensor-Observation standard; ii) Hermes Aggregator: ontologically represented context aggregation and its conversion for application domains; and iii) Hermes History: context history for the different levels of expression produced by other components. These services have been validated and evaluated in a human vital signs monitoring scenario. / Sistemas de Gerenciamento de Contexto (CMS) são responsáveis pelo processamento do contexto ao longo do seu ciclo de vida, que é realizado em quatro etapas definidas pela literatura: aquisição, modelagem, raciocínio e disseminação. A etapa de modelagem possui grande relevância neste processo, sendo responsável por transformar os dados adquiridos de sensores em contexto de alto nível. Porém, esta etapa tem sido impactada pela crescente implantação de sensores e o grande volume de informações que são disponibilizadas, gerando a necessidade de prover serviços que realizem a manipulação dessas informações, oferecendo representação adequada e apoio ao desenvolvimento de aplicações. Neste sentido, a representação baseada em ontologias apresenta grandes vantagens, entre as quais destacam-se: alta expressividade, padronização e interoperabilidade. Diante dos requisitos destacados pela literatura, este trabalho apresenta a construção dos serviços relacionados à etapa de modelagem do ciclo de vida do contexto: representação, agregação e histórico, baseados em ontologias. Estes serviços foram projetados e implementados como componentes que integram um CMS denominado Hermes. Sendo assim, compõem o escopo deste trabalho: i) Hermes Widget: representação ontológica de contexto independente de domínio, baseada no padrão Stimulus-Sensor-Observation; ii) Hermes Aggregator: agregação de contexto representado ontologicamente e conversão do contexto para domínios de aplicação; e iii) Hermes History: histórico de contexto para os diferentes níveis de expressividade produzidos pelos demais componentes. Estes serviços foram validados e avaliados em um cenário de monitoramento de sinais vitais humanos.