Pervasive computing is an intuitive evolution of computing paradigms driven by the wide adoption of mobile devices and wireless networks. It introduces a novel way to support users in their everyday life based on open and dynamic environments populated with unobtrusive services able to perform user tasks on the fly. Nevertheless, supporting user tasks from a functional point of view is not enough to gain the user's satisfaction. Users instead require that their tasks meet a certain Quality of Service (QoS) level. QoS is indeed an inherent and primary requisite of users going along with their required tasks. In the context of pervasive environments, fulfilling user tasks while delivering satisfactory QoS brings about several challenges that are mainly due to the openness, dynamics, and limited underlying resources of these environments. These challenges are mainly about (i) the lack of common QoS understanding among users and service providers, (ii) determining and integra- ting, on the fly, the services available in the environment and able to fulfill the functional and QoS requirements of users, and (iii) adapting the provided services at run-time to cope with QoS fluctuations and ensure meeting user requirements. To cope with the aforementioned issues, we opt for a middleware-based solution. Middle- ware represents indeed the appropriate software system to deal with common concerns of user applications such as QoS. In particular, we opt for a specific kind of middleware, viz., Ser- vice Oriented Middleware (SOM). SOM can leverage middleware technologies and the Service Oriented Computing (SOC) paradigm to enable pervasive environments as dynamic service en- vironments. Particularly, SOM can provide middleware services that allow for supporting QoS of user applications offered by pervasive environments. This thesis presents a QoS-aware service-oriented middleware for pervasive environments. The main contributions of this middleware are : (1) a semantic end-to-end QoS model that enables shared understanding of QoS in pervasive environments, (2) an efficient QoS-aware service composition approach allowing to build service compositions able to fulfill the user functional and QoS requirements, and (3) a QoS-driven adaptation approach to cope with QoS fluctuations during the execution of service compositions. The proposed contributions are implemented within a middleware platform called QASOM and their efficiency is validated based on experimental results.

Natural disasters such as floods, droughts and storms cause many deaths and a great deal of damage worldwide. Recently, several countries have suffered from an the increased number of floods. This has led government agencies to seek to improve flood risk management by providing historical data obtained from stationary sensor networks to help communities that live in hazardous areas. However, the sensor networks can only help to check specific features (e.g. temperature and pressure), and are unable to contribute significantly to supplying the missing information that is required. In addition to stationary sensors, mobile sensors have also been used to monitor floods since they can provide images and reach distances that are not within the coverage of stationary sensors. By combining these heterogeneous sensors, an initiative called Sensor Web Enablement (SWE) seeks to free these applications from the idiosyncrasies that affect the implementation of these heterogeneous sensors. However, SWE cannot always be applied effectively in a context where sensors are embedded and removed dynamically. This dynamic context makes it a complex task to handle, control, access and discover sensors. In view of this, the aim of this work is to dynamically manage heterogeneous sensors involved in flood risk management in near real-time, by enabling interoperable access to their data and using open and reusable components. To achieve this goal, a service-oriented middleware was designed that contains a common protocol message, a dynamic sensor management component and a repository. This approach was evaluated performed by employing an application that prioritizes geographically social media messages based on sensor data. / Os desastres naturais, como inundações, secas e tempestades causam muitas mortes e danos em todo o mundo. Mais recentemente, alguns países sofreram com o aumento das inundações, comparado com outros tipos de desastres. Para melhor gerenciá-las, agências governamentais têm fornecido dados históricos de redes de sensores estáticas para ajudar comunidades que vivem em áreas de risco. No entanto, tais redes de sensores apenas ajudam a verificar propriedades específicas (por exemplo, temperatura e pressão) e pouco contribuem com a falta de informações presente nesse contexto. Além dos sensores estáticos, sensores móveis também têm sido utilizados para monitorar inundações, uma vez que podem fornecer imagens e alcançar distâncias onde sensores estáticos não funcionam adequadamente. Para combinar esses sensores, deve ser utilizado uma iniciativa chamada Sensor Web Enablement (SWE) que isola as aplicações das idiossíncrasias da implementação desses sensores heterogêneos. Entretanto, a SWE não gerencia completamente contextos em que sensores são inseridos e removidos dinamicamente. Este contexto dinâmico torna complexo o controle, o acesso e a descoberta de novos sensores. Logo, o objetivo deste trabalho é gerenciar dinamicamente e próximo do tempo-real sensores heterogêneos envolvidos na gestão de inundações, permitindo um acesso interoperável para seus dados usando componentes abertos e de re-uso. Para alcançar esse objetivo, um middleware orientado a serviços contendo um protocolo de mensagens comum, um componente de gerenciamento dinâmico de sensores e um repositório foi desenvolvido. A avaliação dessa abordagem foi feita considerando uma aplicação que prioriza geograficamente dados de mídias sociais baseados em dados de sensores.

Natural disasters such as floods, droughts and storms cause many deaths and a great deal of damage worldwide. Recently, several countries have suffered from an the increased number of floods. This has led government agencies to seek to improve flood risk management by providing historical data obtained from stationary sensor networks to help communities that live in hazardous areas. However, the sensor networks can only help to check specific features (e.g. temperature and pressure), and are unable to contribute significantly to supplying the missing information that is required. In addition to stationary sensors, mobile sensors have also been used to monitor floods since they can provide images and reach distances that are not within the coverage of stationary sensors. By combining these heterogeneous sensors, an initiative called Sensor Web Enablement (SWE) seeks to free these applications from the idiosyncrasies that affect the implementation of these heterogeneous sensors. However, SWE cannot always be applied effectively in a context where sensors are embedded and removed dynamically. This dynamic context makes it a complex task to handle, control, access and discover sensors. In view of this, the aim of this work is to dynamically manage heterogeneous sensors involved in flood risk management in near real-time, by enabling interoperable access to their data and using open and reusable components. To achieve this goal, a service-oriented middleware was designed that contains a common protocol message, a dynamic sensor management component and a repository. This approach was evaluated performed by employing an application that prioritizes geographically social media messages based on sensor data. / Os desastres naturais, como inundações, secas e tempestades causam muitas mortes e danos em todo o mundo. Mais recentemente, alguns países sofreram com o aumento das inundações, comparado com outros tipos de desastres. Para melhor gerenciá-las, agências governamentais têm fornecido dados históricos de redes de sensores estáticas para ajudar comunidades que vivem em áreas de risco. No entanto, tais redes de sensores apenas ajudam a verificar propriedades específicas (por exemplo, temperatura e pressão) e pouco contribuem com a falta de informações presente nesse contexto. Além dos sensores estáticos, sensores móveis também têm sido utilizados para monitorar inundações, uma vez que podem fornecer imagens e alcançar distâncias onde sensores estáticos não funcionam adequadamente. Para combinar esses sensores, deve ser utilizado uma iniciativa chamada Sensor Web Enablement (SWE) que isola as aplicações das idiossíncrasias da implementação desses sensores heterogêneos. Entretanto, a SWE não gerencia completamente contextos em que sensores são inseridos e removidos dinamicamente. Este contexto dinâmico torna complexo o controle, o acesso e a descoberta de novos sensores. Logo, o objetivo deste trabalho é gerenciar dinamicamente e próximo do tempo-real sensores heterogêneos envolvidos na gestão de inundações, permitindo um acesso interoperável para seus dados usando componentes abertos e de re-uso. Para alcançar esse objetivo, um middleware orientado a serviços contendo um protocolo de mensagens comum, um componente de gerenciamento dinâmico de sensores e um repositório foi desenvolvido. A avaliação dessa abordagem foi feita considerando uma aplicação que prioriza geograficamente dados de mídias sociais baseados em dados de sensores.

L'Internet of Things (IoT) est caractérisé par l'introduction, auprès des utilisateurs, d'un nombre grandissant d'objets (ou things) capables d'acquérir des données depuis leur environnement et d'agir sur celui-ci, et dotés de capacités de calcul et de communication sophistiquées. Une grande partie de ces objets ont pour avantage d'être mobiles, mais cette particularitéprovoque aussi l'émergence de problèmes nouveaux. Les plus critiques d'entre eux découlent directement de l'Internet actuel, sous une forme amplifiée, et portent sur la gestion du grand nombre d'utilisateurs et d'objets connectés, l'interopérabilité entre des objets aux technologies hétérogènes et les changements d'environnement dus à la mobilité d'un très grand nombre d'objets. Cette thèse se propose d'étudier et de résoudre les problèmes susmentionnés en adaptant l'Architecture Orientée Service (SOA) pour que les capteurs et les actionneurs intégrés aux objets puissent être présentés comme des services et, de fait, réduire le couplage entre ces services et leurs hôtes de façon à abstraire leur nature hétérogène. Toutefois, en dépit de ses avantages, SOA n'a pas été conçue pour gérer une aussi grande échelle que celle de l'IoT mobile. En conséquence, la contribution principale de cette thèse porte sur la conception d'une Thing-based Service-Oriented Architecture repensant les fonctionnalités de SOA, et tout particulièrement les mécanismes de découverte et de composition de services. Cette nouvelle architecture a été mise en oeuvre au sein de MobIoT, un middleware spécifiquement conçu pour gérer et contrôler le très grand nombre d'objets mobiles impliqués dans les opérations propres à l'IoT. Dans le but d'évaluer cette nouvelle architecture, nous avons implémenté un prototype et analysé ses performances au travers de nombreuses expériences qui démontrent que les solutions que nous proposons sont viables et pertinentes, notamment en ce qui concerne le passage à l'échelle.