Avec l'apparition des paradigmes des systèmes ubiquitaires ou omniprésents et de l'intelligence ambiante, on assiste à l'émergence d'un nouveau domaine de recherche visant à créer des environnements ou écosystèmes intelligents pouvant offrir une multitude de services permettant d'améliorer la qualité de vie, l'état physique et mental, et le bien-être social des usagers. Dans cette thèse, nous nous focalisons sur la problématique de la représentation sémantique des connaissances et du raisonnement dans le cadre des systèmes à intelligence ambiante et des robots ubiquitaires. Nous proposons deux modèles sémantiques permettant d'améliorer les fonctions cognitives de ces systèmes en termes de gestion du contexte. Au premier modèle, de type ontologique, sont associés un langage de règles et un raisonnement réactif pour la sensibilité au contexte. Pour prendre en compte le caractère dynamique du contexte et assurer une prise de décision cohérente, le mode de raisonnement retenu garantit deux propriétés essentielles : la décidabilité et la non-monotonie. Le deuxième modèle, également de type ontologique, complète le modèle précédent en termes d'expressivité pour la représentation de contextes non-triviaux et/ou liés au temps. Il s'appuie sur des relations n-aires et une représentation narrative des événements pour inférer des causalités entre événements et reconnaitre des contextes complexes non-observables à partir d'événements passés et courants. Les modèles proposés ont été mis en oeuvre et validés sur la plateforme ubiquitaire d'expérimentation du LISSI à partir de trois scenarii d'assistance cognitive et de reconnaissance de contexte / With the appearance of the paradigms of ubiquitous systems and ambient intelligence, a new domain of research is emerging with the aim of creating intelligent environments and ecosystems, that can provide multiple services that can improve quality of life, the physical and mental status and the social wellness of the users. In this thesis, we address the problem of semantic knowledge representation and reasoning, in the context of ambient intelligent systems and ubiquitous robots. We propose two semantic models that improve the cognitive functions of these systems, in terms of context recognition, and context adaptation. The first one is an ontology-based model, which is associated with a rule language to model reactive reasoning process on contextual knowledge. To take into account the dynamicity of context and insure coherent decision-making, this process guarantees two essential properties: decidability and non-monotonic reasoning. The second model is also an ontology-based model that completes the previous model in terms of expressiveness for semantic representation of non-trivial contexts with temporal dimension It is based on n-ary relations and a narrative representation of events for inferring causalities between events, and therefore to build the chronological context of a situation as from past and current events. The proposed models have been implemented on the ubiquitous experimental platform of LISSI, and validated through three scenarios for cognitive assistance and context recognition

Nous proposons donc un modèle original d'apprentissage constructiviste adapté pour un système d'AmI. Ce modèle repose sur une approche décentralisée, permettant de multiples implémentations convenant à un environnement hétérogène. Dans les environnements réels continus sans modélisation à priori, se pose la question de la modélisation des structures élémentaires de représentation et particulièrement le problème d'amorçage de l'apprentissage sensorimoteur (comme décrit par [Kuipers06]). Dans le cadre du modèle général proposé, nous explicitons ce problème particulier et proposons de le traiter comme une forme d'auto-organisation modélisée par un système multi-agent. Cette approche permet de construire des motifs d'interaction élémentaires à partir des seules données brutes, sur lesquels peut reposer la construction d'une représentation plus élaborée (voir [Mazac14]). Nous présentons enfin une série d'expérimentations illustrant la résolution de ce problème d'amorçage : tout d'abord grâce à un environnement simulé, qui permet de maitriser les régularités de l'environnement et autorise des expérimentations rapides ; ensuite en implémentant ce système d'apprentissage au sein d'un environnement d'AmI réel. Pour cela le modèle est intégré dans le système d'AmI développé par l'entreprise partenaire de cette thèse CIFRE. Puis nous présentons une possible application industrielle des résultats de cette première étape implémentée d'amorçage de l'apprentissage sensorimoteur. Nous concluons par l'analyse des résultats et des perspectives de ce type d'approche pour l'AmI et l'application en général de l'IA aux systèmes réels en environnements continus / The theory of cognitive development from Jean Piaget (1923) is a constructivist perspective of learning that has substantially influenced cognitive science domain. Within AI, lots of works have tried to take inspiration from this paradigm since the beginning of the discipline. Indeed it seems that constructivism is a possible trail in order to overcome the limitations of classical techniques stemming from cognitivism or connectionism and create autonomous agents, fitted with strong adaptation ability within their environment, modelled on biological organisms. Potential applications concern intelligent agents in interaction with a complex environment, with objectives that cannot be predefined. Like robotics, Ambient Intelligence (AmI) is a rich and ambitious paradigm that represents a high complexity challenge for AI. In particular, as a part of constructivist theory, the agent has to build a representation of the world that relies on the learning of sensori-motor patterns starting from its own experience only. This step is difficult to set up for systems in continuous environments, using raw data from sensors without a priori modelling.With the use of multi-agent systems, we investigate the development of new techniques in order to adapt constructivist approach of learning on actual cases. Therefore, we use ambient intelligence as a reference domain for the application of our approach

The future Internet will need to cater for an increasing number of powerful devices and entire groups of networks to roam in heterogeneous access networks. The current approach towards meeting such requirements, which is to retrofit mobility solutions to different layers of the protocol stack, has given rise to an increasingly fragmented network control layer. Furthermore, retrofitting solutions in an ad-hoc manner to the protocol stack does not provide consistent support from the network to different applications. This lack of a common control layer for facilitating roaming in heterogeneous networking environments represents a crucial challenge both technically and from a user perspective. To this end, a novel mobility architecture forms the basis and the first part of this dissertation. The work on investigating current network mobility solutions and improving these solutions if deemed necessary, in order to reuse within the novel mobility architecture constitutes the second part of this dissertation. The IETF standard protocol for network mobility was implemented and its performance was analysed on a real networking environment. This enabled to identify problems in the standard which affect the handover and routing performance. To address the identified routing and protocol header overheads of the standard network mobility protocol a novel optimal routing framework, OptiNets was proposed. To address the handover latency issues, optimizations to IPv6 network attachment were incorporated and also an access technology independent multiple interface Make-Before-Break handover mechanism was proposed. The viability of the OptiNets framework and the handover optimizations were demonstrated by analysis and by implementation. A more general external factor that affects the performance of mobile networks which is bandwidth scarcity of Wireless Wide Area Networks was addressed, by proposing a bandwidth fuelling architecture for on-board mobile networks. The feasibility of the bandwidth fuelling architecture was analysed by implementing a prototype and evaluating its performance.

In the vision of Ambient Intelligence, people are assisted in their everyday activities through the proactive, opportunistic support of non-intrusive computing devices offering intuitive interaction modalities. The usefulness and quality of delivered services can be improved considerably if the devices are able to adapt their behaviour according to sensed changes in their surrounding environment, both at the physical and logical levels. This interplay between context-awareness and dynamic software adaptability is key to the construction of applications that are smart with respect to user needs. Unfortunately, most current applications do not reach this level of adaptability, due to a lack of appropriate programming technology. Most applications exhibit fixed functionality and seldom do they sense their environment and adapt their services in a context-aware fashion. Many chances of delivering improved services to users and network peers are thus missed. This dissertation presents a programming model to ease the construction of applications that can react to changes in their execution context by adapting their behaviour dynamically. The starting point of our research is the development of novel language abstractions and the adaptation of existing abstractions to render context-aware, self-adaptable applications easier to develop. We demonstrate that a simple yet powerful computation model readily provides the needed support, leading to straightforward application code that is not concerned with context adaptation, behaviour that can be adapted dynamically to different contexts in a non-intrusive fashion, and context-aware applications with software architectures that are not biased towards context adaptation ---rather, they can be designed freely according to their domain. The proposed computation model is realised through the Ambience programming language, and its underlying open implementation, the Ambient Object System. A small-step operational semantics describes it formally. Much in the vein of prototype-based programming, the model has been designed with simplicity and concreteness in mind. It is highly dynamic, featuring dynamic (multiple) dispatch, dynamic inheritance, dynamic typing, and dynamic method scoping. Application logic adaptation is enabled by means of an intuitive, first-class reification of context that is straightforwardly connected to dynamic behaviour selection. We describe needed management techniques for such context, and a few programming guidelines on how to develop context-aware applications using our approach. The approach is validated by showing its application in a number of scenarios inspired on Ambient Intelligence.

The delivery of quality healthcare to all citizens at reasonable costs is an important challenge. With the increase in the aging population, the costs of managing chronic diseases increase. Today, healthcare services tend to shift from recovery to prevention. Remote healthcare monitoring is crucial for prevention and monitoring of chronic diseases since they require continuous and long-term monitoring. The advances in networking, mobile communications and medical device technologies offer a great potential to realize remote healthcare monitoring. However, seamless integration of multi-modal medical devices to the existing healthcare information systems is necessary for the automated use of medical device observations in related applications. The thesis addresses the automatic acquisition and use of multi-modal medical device observations in healthcare information systems. The interoperability of medical devices with healthcare information systems requires both physical connectivity and application level interoperability. Therefore, the thesis concentrates on both the medical device domain and the interoperability efforts on the existing healthcare information systems. It provides an interoperability solution based on ISO/IEEE 11073 and HL7 standards. This work is also realized the automatic acquisition and use of multi-modal medical device observations in an intelligent healthcare monitoring and decision support system which is developed as a part of the IST-027074 SAPHIRE project funded by the European Commission.

Potentially transformational ideas in several applications of human and computer interaction form the motivation for this work. It is targeted towards a systematic approach to the design of systems with complex, intelligent and interactive exchange of information between a system and the environment it is meant to monitor, and gather knowledge about. The Pahl and Beitz systematic design method is modified with the inclusion of generic sub-systems from Living Systems Theory, modeling and simulation tools and other adaptations within the context of the validation square to synthesize a design method for the design of systems with intelligent and interactive information flow. The validation of the proposed design method is carried out with the aid of an example wherein a motion capture system is designed based on the Nintendo Wii Remote(TM). Results include an evaluation of the performance of a prototype as well as the design method itself in the context of the requirements that the method must fulfill.

The aim of the thesis is to design a web services architecture that shall support the automatic determination of a personalised sensor-system, which is embedded in smart garments, and which shall be used in self-care in order to allow a profane user to record himself a personal electrocardiogram (ECG), at anytime and anywhere. The main challenge consists in the intelligent and dynamic orchestration of context-aware business services that supply the user with an optimal personalised solution, while mastering the system's complexity: context dependent user and system interactions, knowledge extraction from subject-specific vital signs, services reconfiguration automation. The solution to this challenge is to create an Ambient Intelligence which goes beyond Ubiquitous Computing and is capable to replace an expert by proposing an Intelligent Assistance to any citizen. We propose a methodology expressed in terms of Data and Model driven Service Oriented Architecture (DM-SOA), which provides a framework for the production of context-aware intelligent business services. This architecture supports the automation of sophisticated and personalised expert activities, which apply professional knowledge to process an individual case. The proposed solution is based on a new dynamic business process modelling approach and in its implementation via automatically reconfigurable services. It consists in setting-up an intelligent environment based on a business process ontology of context-aware concepts and on related context handling rules for services orchestration. A core-ontology has been designed to support the automation of expert activities related to decision-making. The soundness of the method and of the underlying decision-making techniques has been demonstrated in the cardiology domain for selecting the most adequate subject-specific sensor-system, characterised by its ability to yield for an ECG signal of similar diagnostic content than a standard 12-lead ECG. We propose the design of a new three modalities sensor-system prototype as a response to the need of sensor-systems used on demand in self-care situations and ensuring diagnostic quality signals recordings. This prototype has been tested on a set of healthy volunteers presenting various characteristics in age, sex and morphology. In this context, Ambient Intelligence is demonstrated as an ensemble of professional quality services ready to use by profane users. These services are accessible pervasively, trough the objects people work with / wear on / use each day, and provide an appropriate guidance to the non-competent users. Such an Ambient Intelligence approach strongly supports the vision of an Information Society which corresponds to the long-term goal of the EU Information Society Technologies Research Programme that aims to ease knowledge access. The future healthcare will benefit from this approach and will significantly improve patient care.

Leveraging our experience with the traditional approach to ambient assisted living (AAL) which relies on a large spread of heterogeneous technologies in deployments, this thesis studies the possibility of a more "stripped down" and complementary approach, where only a reduced hardware subset is deployed, probing a transfer of complexity towards the software side, and enhancing the large scale deployability of the solution. Focused on the reasoning aspects in AAL systems, this work has allowed the finding of a suitable semantic inference engine for the peculiar use in these systems, responding to a need in this scientific community. Considering the coarse granularity of situational data available, dedicated rule-sets with adapted inference strategies are proposed, implemented, and validated using this engine. A novel semantic reasoning mechanism is proposed based on a cognitively inspired reasoning architecture. Finally, the whole reasoning system is integrated in a fully featured context-aware service framework, powering its context awareness by performing live event processing through complex ontological manipulation. the overall system is validated through in-situ deployments in a nursing home as well as private homes over a few months period, which itself is noticeable in a mainly laboratory-bound research domain

This thesis takes place in the field of Ambient Intelligence (AmI). AmI Systems are interactive systems composed of many heterogeneous components. From a hardware perspective these components can be divided into two main classes: sensors, using which the system observes its surroundings, and actuators, through which the system acts upon its surroundings in order to execute specific tasks.From a functional point of view, the goal of AmI Systems is to activate some actuators, based on data provided by some sensors. However, sensors and actuators may suffer failures. Our motivation in this thesis is to equip ambient systems with self fault detection capabilities. One of the particularities of AmI systems is that instances of physical resources (mainly sensors and actuators) are not necessarily known at design time; instead they are dynamically discovered at run-time. In consequence, one could not apply classical control theory to pre-determine closed control loops using the available sensors. We propose an approach in which the fault detection and diagnosis in AmI systems is dynamically done at run-time, while decoupling actuators and sensors at design time. We introduce a Fault Detection and Diagnosis framework modeling the generic characteristics of actuators and sensors, and the physical effects that are expected on the physical environment when a given action is performed by the system's actuators. These effects are then used at run-time to link actuators (that produce them) with the corresponding sensors (that detect them). Most importantly the mathematical model describing each effect allows the calculation of the expected readings of sensors. Comparing the predicted values with the actual values provided by sensors allows us to achieve fault-detection.

Almost 20 years ago, Marc Weiser envisioned the prospect of computer in 21st century, and proposed the pioneering notion of ubiquitous computing. One of Weiser's primary ideas has recently evolved to a more general paradigm known as context awareness, becoming a central research theme in many other ubiquitous computing programs. From Active Badge considered as the first context-aware application, there are numerous attempts to build effective context-aware systems. However, how to acquire context, how to process context and how to create context-aware applications is still faced with enormous challenges in the both of research and practice. This dissertation investigates deeply some chosen key issues in context awareness and develops a context-aware middleware. The main research contributions are presented in three categories: a spatialtemporal context represent model, a context-aware middleware and an intelligence context inference engine. The spatial-temporal context representation model is proposed to organize context and relations for context-aware system. Ontology-based method is adopted to construct our model, supporting both knowledge sharing and reuse as well as logic inference. This model adopts two-layer hierarchy structure for different situation. The higher layer comes up with the generic common context, while the lower layer focuses on various specific situations. Differing from existing models, besides taking locational factors into account, it supports different historical context service depending on different context resource. These context histories may be used to predict and infer the context. A context-aware middleware is designed as a platform associated with context retrieval and context processing. It is organized in two layers: the low layer provides a solution to integrate sensors and actuators with a standardized data representation; the high layer: versatile context interpreter focuses on context processing, which is made up of four parts: Context Aggregator, Inference Engine, Context Knowledge Base, and Query Engine in charge of context inferences, expressive query, and persistent storage. This middleware provides an environment for rapid prototyping of context aware services in ambient intelligent. The intelligent inference engine is the central and intellectual component of context-aware middleware. We review all the methods on activity context recognition published in three premier conferences in past decade and conclude that activity context recognition is divided into three facets: basic activity inference, dynamic activity analysis and future activity recommendation. Then we propose an intelligent inference engine based on our context-aware middleware. Beside satisfying requirements of checking the context consistency, our inference engine integrates the three most popular methods on activity context recognition: Rules, Decision Tree, and Hide Markov Model. It provides a solution for all facets of activity context recognition based on our context-aware middleware. The individuals' information collecting from their social networks under permission are leveraged to train intelligent inference engine. We finally use two scenarios (applications) to explain the generic process to develop application via our middleware, and compare and analyze the main aspects of our middleware with other five representative context-aware applications. Our middleware profits good features from existing context-aware systems and improve intelligence via supporting activity context recognition. It provides an efficient platform for a rapid developing of new context-aware applications in ambient intelligence.