Pervasive (ubiquitous) computing is a new paradigm where the computers are submerged into the background of the everyday life. One important aspect of pervasive systems is context-awareness. Context-aware systems are those that can adapt their behaviours according to the current context. Context-aware applications are being integrated into our everyday activity aspects such as: health care, smart homes and transportations. There exist a wide range of context-aware applications such as: mobile phones, learning systems, smart vehicles. Some context-aware systems are critical since the consequence of failing to identify a given context may be catastrophic. For example, an auto-pilot system is a critical context-aware system; it senses the humidity, clouds, wind speed and accordingly adjusts the altitude, throttle and other parameters. Being a critical context-aware system has to be provably correct. Policy-based approaches has been used in many applications but not in context-aware systems. In this research, we want to discover the anatomy (i.e. architecture, structure and operational behaviour) of policy-based management as applied to context-aware systems, and how policies are managed within such a dynamic system. We propose a novel computational model and its formalisation is presented using the Calculus of Context-aware Ambients (CCA). CCA has been proposed as a suitable mathematical notation to model mobile and context-aware systems. We decided to use CCA due to three reasons: (i) in CCA, mobility and context-awareness are primitive constructs and are treated as first-class citizens; (ii) properties of a system can be formally analysed; (iii) CCA specifications are executable, and thus, leading to rapid prototyping and early validation of the system properties. We, then show how policies can be expressed in CCA. For illustration, the specification of the event-condition-action (ECA) conceptual policy model is modelled in CCA in a natural fashion. We also propose a policy-based architecture for context-aware systems, showing its different components, and how they interact. Furthermore, we give the specification of the policy enforcement mechanism used in our proposed architecture in CCA. To evaluate our approach, a real-world case study of an infostation-based mobile learning (mLearning) system is chosen. This mLearning system is deployed across a university campus to enable mobile users to access mobile services (mServices) represented by course materials (lectures, tests and tutorials) and communication services (intelligent message notification and VoIP). Users can access the mServices through their mobile devices (Hand-set phones, PDAs and laptops) regardless of their device type or location within a university campus. We have specified the mLearning system in CCA (i.e. specification based on policies of the mServices), afterwards, the specification is simulated using the CCA interpreter tool. We have developed an animation tool specially designed for the mLearning system. The animation tool provides graphical representation of the CCA processes. In terms of safety and liveness, some important properties of the mLearning system have been validated as a proof of concept.

Stochastic models, in particular Continuous Time Markov Chains, are a commonly employed mathematical abstraction for describing natural or engineered dynamical systems. While the theory behind them is well-studied, their specification can be problematic in a number of ways. Firstly, the size and complexity of the model can make its description difficult without using a high-level language. Secondly, knowledge of the system is usually incomplete, leaving one or more parameters with unknown values, thus impeding further analysis. Sophisticated machine learning algorithms have been proposed for the statistically rigorous estimation and handling of this uncertainty; however, their applicability is often limited to systems with finite state-space, and there has not been any consideration for their use on high-level descriptions. Similarly, high-level formal languages have been long used for describing and reasoning about stochastic systems, but require a full specification; efforts to estimate parameters for such formal models have been limited to simple inference algorithms. This thesis explores how these two approaches can be brought together, drawing ideas from the probabilistic programming paradigm. We introduce ProPPA, a process algebra for the specification of stochastic systems with uncertain parameters. The language is equipped with a semantics, allowing a formal interpretation of models written in it. This is the first time that uncertainty has been incorporated into the syntax and semantics of a formal language, and we describe a new mathematical object capable of capturing this information. We provide a series of algorithms for inference which can be automatically applied to ProPPA models without the need to write extra code. As part of these, we develop a novel inference scheme for infinite-state systems, based on random truncations of the state-space. The expressive power and inference capabilities of the framework are demonstrated in a series of small examples as well as a larger-scale case study. We also present a review of the state-of-the-art in both machine learning and formal modelling with respect to stochastic systems. We close with a discussion of potential extensions of this work, and thoughts about different ways in which the fields of statistical machine learning and formal modelling can be further integrated.

Les neurones biologiques communiquent par le biais d’impulsions électriques, appelées spikes, et les fonctions cérébrales émergent notamment de la coordination entre les réceptions et émissions de ces spikes. Par ailleurs, il est largement admis que la fonction de chaque neurone dépend de sa morphologie. Les dendrites conditionnent l’intégration spatio-temporelle des spikes reçus et influent sur les temps d’occurrence des spikes émis. Elles sont donc fondamentales pour l’étude in silico des mécanismes de coordination, et en particulier pour l’étude des assemblées de neurones. Les modèles de neurones existants prenant en compte les dendrites, sont généralement des modèles mathématiques détaillés, souvent à base d’équations différentielles, dont la simulation nécessite des ressources de calculs importantes. De plus, leur complexité intrinsèque rend difficile l’analyse et les preuves sur ces modèles. Dans cette thèse, nous proposons un modèle de neurone intégrant des dendrites d’une manière abstraite. Dans l’objectif d’ouvrir la porte aux méthodes formelles, nous établissons une définition rigoureuse du cadre de modélisation et mettons en évidence des propriétés algébriques remarquables de l’intégration dendritique. Nous avons notamment démontré qu’il est possible de réduire la structure d’un neurone en préservant sa fonction d’entrée/sortie. Nous avons ainsi révélé des classes d’équivalence dont nous savons déterminer un représentant canonique. En s’appuyant sur la théorie des catégories et par des morphismes de neurones judicieusement définis, nous avons ensuite analysé plus finement ces classes d’équivalence. Un résultat surprenant découle de ces propriétés : un simple ajout de délais dans les modèles informatiques de neurones permet de prendre en compte une intégration dendritique abstraite, sans représenter explicitement la structure arborescente des dendrites. À la racine de l’arborescence dendritique, la modélisation du soma contient inévitablement une équation différentielle lorsque l’on souhaite préserver l’essence du fonctionnement biologique. Ceci impose de combiner une vision analytique avec la vision algébrique. Néanmoins, grâce à une étape préalable de discrétisation temporelle, nous avons également implémenté un neurone complet en Lustre qui est un langage formel autorisant des preuves par model checking. Globalement, nous apportons dans cette thèse un premier pas encourageant vers une formalisation complète des neurones, avec des propriétés remarquables sur l’intégration dendritique. / Biological neurons communicate by means of electrical impulses, called spikes. Brain functions emerge notably from reception and emission coordination between those spikes. Furthermore, it is widely accepted that the function of each neuron depends on its morphology. In particular, dendrites perform the spatio-temporal integration of received spikes and affect the occurrence of emitted spikes. Dendrites are therefore fundamental for in silico studies of coordination mechanisms, and especially for the study of so-called neuron assemblies. Most of existing neuron models taking into account dendrites are detailed mathematical models, usually based on differential equations, whose simulations require significant computing resources. Moreover, their intrinsic complexity makes difficult the analysis and proofs on such models. In this thesis, we propose an abstract neuron model integrating dendrites. In order to pave the way to formal methods, we establish a rigorous definition of the modeling framework and highlight remarkable algebraic properties of dendritic integration. In particular, we have demonstrated that it is possible to reduce a neuron structure while preserving its input/output function. We have thus revealed equivalence classes with a canonical representative. Based on category theory and thanks to properly defined neuron morphisms, we then analyzed these equivalence classes in more details. A surprising result derives from these properties: simply adding delays in neuron computational models is sufficient to represent an abstract dendritic integration, without explicit tree structure representation of dendrites. At the root of the dendritic tree, soma modeling inevitably contains a differential equation in order to preserve the biological functioning essence. This requires combining an analytical vision with the algebraic vision. Nevertheless, thanks to a preliminary step of temporal discretization, we have also implemented a complete neuron in Lustre which is a formal language allowing proofs by model checking. All in all, we bring in this thesis an encouraging first step towards a complete neuron formalization, with remarkable properties on dendritic integration.

L’utilisation des communications multi-canaux au niveau des protocoles MAC (MAC multi-canaux) dans les réseaux sans fil peut améliorer significativement le débit du réseau par rapport aux protocoles MAC mono-canal en permettant à plusieurs transmissions concurrentes d’avoir lieu simultanément sans interférence sur différents canaux non recouverts. Cependant, l’environnement multicanaux ajoute une nouvelle dimension aux protocoles MAC qui nécessite un nouveau mécanisme appelé "le mécanisme d’allocation des canaux" pour coordonner l’utilisation des canaux. Le rôle de ce mécanisme est de permettre aux nœuds de s’accorder sur le canal qui doit être utilisé et à quel moment pour la transmission de données dans un contexte distribué. L’absence de ce mécanisme ou une mauvaise coordination de l’utilisation des canaux peut produire l’effet inverse de celui recherché. Ainsi, l’efficacité d’un protocole MAC multi-canaux dépend strictement de l’efficacité du mécanisme d’allocation des canaux, qui par ailleurs, doit offrir les garanties nécessaires au fonctionnement de la communication multi-canaux. Dans cette thèse, nous nous intéressons à la modélisation, l’analyse et la conception de nouveaux protocoles MAC multi-canaux dans les réseaux sans fil. Dans un premier temps, nous nous focalisons sur la spécification formelle et la vérification des propriétés fondamentales inhérentes aux activités de l’allocation des canaux et son impact sur le transfert de données des principales approches MAC multi-canaux proposées dans la littérature. Dans un second temps, nous visons à proposer, modéliser et à analyser formellement deux nouveaux protocoles MAC multicanaux résultants de l’analyse qualitative réalisée. Enfin, une étude quantitative et comparative est menée afin d’analyser les performances des protocoles proposés en terme de débit et d’impact du temps de commutation entre les canaux. / The use of multi-channel communications in MAC protocols (multi-channel MAC) for wireless networks can significantly improve the network throughput compared with single channel MAC protocols since concurrent transmissions can take place simultaneously on distinct non-overlapping channels without interfering. However, due to the multi-channel environment, a novel issue must be addressed in MAC protocols, which is called "the channel assignment mechanism" in order to coordinate the channels access. The role of this mechanism is to enable nodes to agree on which channel should be used and when for data transmission in a distributed way. The absence of this mechanism or a bad coordination of the channel usage can incur the opposite effect to that intended. Thus, the efficiency of a multi-channel MAC protocol depends strictly on the efficiency of the channel assignment mechanism, which also must provide the guarantees necessary to the good behavior of the multi-channel communications. In this thesis, we focus on the modeling, analysis, and design of new multi-channel MAC protocols in wireless networks. Firstly, we focus on the formal specification and verification of fundamental properties inherent to the activities of the channel assignment and its impact on the data transfer for the main multi-channel MAC approaches proposed in the literature. Secondly, we aim to propose, model and analyze formally two new multichannel MAC protocols, resulting from the qualitative analysis conducted. Finally, a quantitative and comparative study is conducted to analyze the performance of the proposed protocols in terms of the throughput and the impact of the switching time between channels.

no / Formal modelling is essential for precisely defining, understanding and reasoning when designing complex systems, such as cyberphysical systems. In this paper we present a formal specification using Event-B and Rodin platform for a case study of a cruise control system for a hybrid propulsion vehicle and electric bicycle (e-Bike). Our work uses the EventB method, a formal approach for reliable systems specification and verification, being supported by the Rodin platform, based on theorem proving, allowing a stepwise specification process based on refinement. We also use, from the same platform, the ProB model checker for the verification of the B-Machine and iUML plug-in to visualize our model. This approach shows the benefits of using a formal modelling platform, in the context of cyberphysical systems, which provides multiple ways of analysing a system. / Romanian National Authority for Scientific Research, CNCS-UEFISCDI, project number PN-III-P4-ID-PCE-20160210.

CoordenaÃÃo de AperfeÃoamento de Pessoal de NÃvel Superior / This work presents a method to generate games with multilevel and multiple users using Hierarchical Coloured Petri Nets. The design of a multiuser game containing multiple navigation environments (multilevel) is made from the model of a Hierarchical Coloured Petri Net, in which are specified all of the rules, properties and structures of the game. The created model for the game can be formally analysed, verifying the existence of deadlocks and invalid paths, for example, and others possible conception problems, that can be done with the tools available on CPN Tools. To validate this method, a tool has been developed, called CPN Games, which allows fast and dynamic development of simple conception games only using Hierarchical Coloured Petri Nets. The XML codes of the designed models in CPN Tools are interpreted by CPN Games following a set of predefined rules to instantiate different games, without additional programing. It is demonstrated the mechanisms of the design, analysis and validation of the models and finally it is presented several examples of games developed by this tool. / Este trabalho apresenta um mÃtodo para geraÃÃo de jogos multinÃveis e com mÃltiplos usuÃrios por meio de modelagem em Redes de Petri Coloridas HierÃrquicas. A concepÃÃo de um jogo multiusuÃrio, contendo mÃltiplos ambientes de navegaÃÃo (multinÃveis), à feita a partir da criaÃÃo de um modelo hierÃrquico em Rede de Petri Colorida, no qual se especificam todas as regras, propriedades e estruturas do jogo. O modelo criado para o jogo pode ser analisado formalmente, verificando-se, por exemplo, a existÃncia de bloqueios e de transiÃÃes mortas (caminhos invÃlidos), entre outros possÃveis problemas de concepÃÃo, o que pode ser feito com a utilizaÃÃo de ferramentas disponÃveis no CPN Tools. Para validar esse mÃtodo, foi concebida uma ferramenta, denominada CPN Games, que permite o desenvolvimento rÃpido e dinÃmico de jogos de concepÃÃo simples exclusivamente por Rede de Petri Colorida HierÃrquica. Os cÃdigos em XML de modelos constituÃdos com o CPN Tools sÃo interpretadas pelo CPN Games seguindo um conjunto de regras prà estabelecidas para instanciar diferentes jogos, sem a necessidade de programaÃÃo adicional. SÃo demonstrados os mecanismos de criaÃÃo, anÃlise e validaÃÃo dos modelos e diferentes exemplos de jogos construÃdos com a ferramenta.