Grâce à sa promesse de réduire les efforts de développement et maintenance du logiciel, l’Ingénierie Dirigée par les Modèles (IDM) attire de plus en plus les acteurs industriels. En effet, elle a été adoptée avec succès dans plusieurs domaines tels que le génie civil, l’industrie automobile et la modernisation de logiciels.Toutefois, la taille croissante des modèles utilisés nécessite de concevoir des solutions passant à l’échelle afin de les traiter (transformer), et stocker (persister) de manière efficace. Une façon de pallier cette problématique est d’utiliser les systèmes et les bases de données répartis. D’une part, les paradigmes de programmation distribuée tels que MapReduce et Pregel peuvent simplifier la distribution de transformations des modèles (TM). Et d’autre part, l’avènement des base de données NoSQL permet le stockage efficace des modèles d’une manière distribuée. Dans le cadre de cette thèse, nous proposons une approche pour la transformation ainsi que pour la persistance de grands modèles.Nous nous basons d’un côté, sur le haut niveau d’abstraction fourni par les langages déclaratifs (relationnels) de transformation et d’un autre côté, sur la sémantique bien définie des paradigmes existants de programmation distribués, afin de livrer un moteur distribué de TM. La distribution est implicite et la syntaxe du langage n’est pas modifiée (aucune primitive de parallélisation n’est ajoutée). Nous étendons cette solution avec un algorithme efficace de distribution de modèles qui se base sur l’analyse statique des transformations et sur résultats récents sur le partitionnement équilibré des graphes. Nous avons appliqué notre approche à ATL, un langage relationnel de TM et MapReduce, un paradigme de programmation distribué. Finalement, nous proposons une solution pour stocker des modèles à l’aide de bases de données NoSQL, en particulier au travers d’un cadre d’applications de persistance répartie. / Model-Driven Engineering (MDE) is gaining ground in industrial environments, thanks to its promise of lowering software development and maintenance effort. It has been adopted with success in producing software for several domains like civil engineering, car manufacturing and modernization of legacy software systems. As the models that need to be handled in model-driven engineering grow in scale, it became necessary to design scalable algorithms for model transformation (MT) as well as well-suitable persistence frameworks. One way to cope with these issues is to exploit the wide availability of distributed clusters in the Cloud for the distributed execution of model transformations and their persistence. On one hand, programming models such as MapReduce and Pregel may simplify the development of distributed model transformations. On the other hand, the availability of different categories of NoSQL databases may help to store efficiently the models. However, because of the dense interconnectivity of models and the complexity of transformation logics, scalability in distributed model processing is challenging. In this thesis, we propose our approach for scalable model transformation and persistence. We exploit the high-level of abstraction of relational MT languages and the well-defined semantics of existing distributed programming models to provide a relational model transformation engine with implicit distributed execution. The syntax of the MT language is not modified and no primitive for distribution is added. Hence developers are not required to have any acquaintance with distributed programming.We extend this approach with an efficient model distribution algorithm, based on the analysis of relational model transformation and recent results on balanced partitioning of streaming graphs. We applied our approach to a popular MT language, ATL, on top of a well-known distributed programming model, MapReduce. Finally, we propose a multi-persistence backend for manipulating and storing models in NoSQL databases according to the modeling scenario. Especially, we focus on decentralized model persistence for distributed model transformations.

[pt] Práticas de engenharia de requisitos ágeis estão se tornando mais comuns em equipes de desenvolvimento de software. Contudo, as práticas relacionadas ao controle de qualidade ainda dependem fortemente do conhecimento, da experiência e do trabalho manual de testadores, em adição as especificações de requisitos produzidas são frequentemente imprecisas e difíceis de verificar estaticamente por interessados ou por algum computador. Essa tese ataca conjuntamente o problema de verificar estaticamente especificações de requisitos ágeis e de gerar casos de teste e scripts de teste automatizados completos a partir delas. Suas contribuições principais incluem: (1) uma nova metalinguagem, chamada Concordia, que permite escrever especificações de requisitos ágeis que podem ser usadas para atividades de verificação e validação (V e V); (2) uma nova abordagem para gerar casos de teste e scripts de teste automatizado completos, a partir de requisitos especificados com a metalinguagem; (3) a medição, em contexto industrial, da capacidade da abordagem em reduzir o risco de defeitos e custos de V e V. / [en] Agile requirements engineering practices are being used more commonly by software development teams. However, practices related to quality control still depend heavily on testers expertise and manual labor, whilst produced require-ments specifications are often imprecise and hard to verify statically by both stake-holders and computers. This thesis jointly tackles the problem of verifying statically agile requirements specifications and generating full-featured test cases and auto-mated test scripts from them. Its main contributions include: (1) a new metalan-guage, called Concordia, for writing agile requirement specifications that can be used for both verification and validation (V and V) activities involving stakeholders; (2) a novel approach to generate full-featured ready to use test cases and automated test scripts from the requirements specified with the metalanguage; (3) the assess-ment in industrial context of the approaches ability to reduce risk of remaining defects and the costs of V and V.

The need for modern enterprises and Web users to simply and rapidly develop and deliver platform-independent services to be accessed over the Web by the global community is growing. This is self-evident, when one considers the omnipresence of electronic services (e-services) on the Web. Accordingly, the Service-Oriented Architecture (SOA) is commonly considered as one of the de facto standards for the provisioning of heterogeneous business functionalities on the Web. As the basis for SOA, Web Services (WS) are commonly preferred, particularly because of their ability to facilitate the integration of heterogeneous systems. However, WS only focus on syntactic descriptions when describing the functional and behavioural aspects of services. This makes it a challenge for services to be automatically discovered, selected, composed, invoked, and executed – without any human intervention. Consequently, Semantic Web Services (SWS) are emerging to deal with such a challenge. SWS represent the convergence of Semantic Web (SW) and WS concepts, in order to enable Web services that can be automatically processed and understood by machines operating with limited or no user intervention. At present, research efforts within the SWS domain are mainly concentrated on semantic services automation aspects, such as discovery, matching, selection, composition, invocation, and execution. Moreover, extensive research has been conducted on the conceptual models and formal languages used in constructing semantic services. However, in terms of the engineering of semantic services, a number of challenges are still prevalent, as demonstrated by the lack of development and use of semantic services in real-world settings. The lack of development and use could be attributed to a number of challenges, such as complex semantic services enabling technologies, leading to a steep learning curve for service developers; lack of unified service platforms for guiding and supporting simple and rapid engineering of semantic services, and the limited integration of semantic technologies with mature service-oriented technologies. vi In addition, a combination of isolated software tools is normally used to engineer semantic services. This could, however, lead to undesirable consequences, such as prolonged service development times, high service development costs, lack of services re-use, and the lack of semantics interoperability, reliability, and re-usability. Furthermore, available software platforms do not support the creation of semantic services that are intelligent beyond the application of semantic descriptions, as envisaged for the next generation of services, where the connection of knowledge is of core importance. In addressing some of the challenges highlighted, this research study adopted a qualitative research approach with the main focus on conceptual modelling. The main contribution of this study is thus a framework called iSemServ to simplify and accelerate the process of engineering intelligent semantic services. The framework has been modelled and developed, based on the principles of simplicity, rapidity, and intelligence. The key contributions of the proposed framework are: (1) An end-to-end and unified approach of engineering intelligent semantic services, thereby enabling service engineers to use one platform to realize all the modules comprising such services; (2) proposal of a model-driven approach that enables the average and expert service engineers to focus on developing intelligent semantic services in a structured, extensible, and platform-independent manner. Thereby increasing developers’ productivity and minimizing development and maintenance costs; (3) complexity hiding through the exploitation of template and rule-based automatic code generators, supporting different service architectural styles and semantic models; and (4) intelligence wrapping of services at message and knowledge levels, for the purposes of automatically processing semantic service requests, responses and reasoning over domain ontologies and semantic descriptions by keeping user intervention at a minimum. The framework was designed by following a model-driven approach and implemented using the Eclipse platform. It was evaluated using practical use case scenarios, comparative analysis, and performance and scalability experiments. In conclusion, the iSemServ framework is considered appropriate for dealing with the complexities and restrictions involved in engineering intelligent semantic services, especially because the amount of time required to generate intelligent semantic vii services using the proposed framework is smaller compared with the time that the service engineer would need to manually generate all the different artefacts comprising an intelligent semantic service. Keywords: Intelligent semantic services, Web services, Ontologies, Intelligent agents, Service engineering, Model-driven techniques, iSemServ framework. / Computing / D. Phil. (Computer science)

L’avènement de l’IDM (Ingénierie Dirigée par les Modèles) a suscité beaucoup d’intérêt de la part des organisations qui de fait commencent à transformer leur processus de développement traditionnel en un processus de développement dirigé par les modèles, appelé aussi processus IDM.Au moment où ces processus commencent à émerger, nous notons l’absence d’un langage dédié pour les modéliser et les mettre en œuvre. Le standard SPEM 2.0 propose des concepts génériques qui sont supposés être capables de décrire tout type de processus logiciel. Cependant, les concepts de SPEM ne capturent pas la nature exacte des processus IDM. D’autre part, une autre insuffisance majeure de SPEM réside dans le fait qu’il n’intègre pas les concepts relatifs à la mise en œuvre des processus.L’objectif de cette thèse est triple : (1) proposer une extension de SPEM dans laquelle les concepts centraux des processus IDM sont réifiés ; (2) proposer un langage dédié à la modélisation comportementale des processus IDM ; (3) proposer une architecture conceptuelle d’un environnement logiciel d’aide à la modélisation et à la mise en œuvre des processus IDM.Pour valider notre approche, un prototype a été développé sous l’environnement TOPCASED. Ce prototype fournit d’une part un éditeur graphique pour la modélisation structurelle et comportementale des processus IDM et d’autre part un environnement de mise en œuvre s’appuyant sur les modèles comportementaux des processus. Nous avons également appliqué notre approche à une étude de cas significatif: le processus UWE (UML-based Web Engineering), qui est un processus IDM dédié au développement d’applications web. / With the emergence of MDE, many organizations have been starting to transform their traditional software development processes into model-driven processes. Kleppe and al. define a model-driven software development as “a process of developing software using different models on different levels of abstraction with (automated) transformations between these models”.While model-driven development processes – called MDE processes – have started to appear, a tool-supported Process Modeling Language (PML) for describing and enacting such processes is still lacking. The concepts of SPEM 2.0 are quite generic since they are supposed to allow describing any kind of software. However, SPEM 2.0 concepts do not succeed in capturing the exact nature of most activities and artifacts of model-driven development. In addition, another major weakness of SPEM 2.0 is the lack of concepts for process enactment.The objective of this thesis is threefold: (1) provide an extension of SPEM that reifies the MDE concepts; (2) provide a language dedicated to behavioral modeling of MDE processes; (3) provide a conceptual architecture of a PSEE (Process-centered Software Engineering Environment) that guides process designer at modeling phase and developers at enactment time.To validate our approach, a prototype of this PSEE is developed under the TOPCASED environment. This prototype provides a graphical editor for structural and behavioral modeling of MDE processes, and a process enactment engine based on process behavior models. We have also applied our approach to a significant case study: the UWE (UML-based Web Engineering) process, which is a MDE process dedicated to web applications development.

Les activités de validation et vérification de modèles sont devenues essentielles dans le développement de systèmes complexes. Les efforts de formalisation de ces activités se sont multipliés récemment étant donné leur importance pour les systèmes embarqués critiques. Notre travail s’inscrit principalement dans cette voie. Nous abordons deux visions complémentaires pour traiter cette problématique. La première est une description syntaxique implicite macroscopique basée sur une ontologie pour aider les concepteurs dans le choix des outils selon leurs exigences. La seconde est une description sémantique explicite microscopique pour faciliter la construction de techniques de vérification compositionnelles. Nous proposons dans la première partie de cette thèse une ontologie pour expliquer et expliciter les éléments fondateurs du domaine que nous appelons VVO. Cette ontologie pourra avoir plusieurs autres utilisations : une base de connaissance, un outil de formation ou aussi un support pour le choix de la méthode à appliquer et l’inférence de correspondance entre outils. Nous nous intéressons dans la seconde partie de cette thèse à une formalisation dans un assistant à la preuve de l’introduction de composants dans un langage de modélisation et des liens avec les activités de validation et vérification. Le but est d’étudier la préservation des propriétés par composition : les activités de vérification sont généralement coûteuses en terme de temps et d’effort, les faire d’une façon compositionnelle est très avantageux. Nous partons de l’atelier formel pour l’Ingénierie Dirigée par les Modèles Coq4MDE. Nous suivons la même ligne directrice de développement prouvé pour formaliser des opérateurs de composition et étudier la conservation des propriétés par assemblage. Nous nous intéressons au typage puis à la conformité de modèles par rapport au métamodèle et nous vérifions que les opérateurs définis permettent de conserver ces propriétés. Nous nous focalisons sur l’étude d’opérateurs élémentaires que nous exploitons pour spécifier des opérateurs de plus haut niveau. Les préconditions des opérateurs représentent les activités de vérification non compositionnelles qui doivent être effectuées en plus de la vérification des composants pour assurer la postcondition des opérateurs qui est la propriété souhaitée. Nous concluons en présentant des perspectives pour une formalisation algébrique en théorie des catégories. / The validation and verification of models have become essential in the development of complex systems. The formalisation efforts for these activities have increased recently being given their importance for critical embedded systems. We discuss two complementary visions for addressing these issues. The first is a syntactic implicit macroscopic description based on an ontology to help designers in the choice of tools depending on their requirements. The second is a microscopic explicit semantics description aiming to facilitate the construction of compositional verification techniques. We propose in the first part of this thesis an ontology to explain and clarify the basic elements of the domain of Verification and Validation that we call VVO. This ontology may have several other uses: a knowledge base, a training tool or a support for the choice of the method to be applied and to infer correspondence between tools. We are interested in the second part of this thesis in a formalisation using a proof assistant for the introduction of components in a modelling language and their links with verification and validation activities. The aim is to study the preservation of properties by the composition activities. The verification are generally expensive in terms of time and efforts, making theme in a compositional way is very advantageous. Starting from the formal framework for Model Driven Engineering COQ4MDE, we follow the same line of though to formalize the composition operators and to study the conservation of properties by composition. We are interested in typing and conformity of models in relation with metamodels and we verify that the defined operators allow to preserve these properties. We focus on the study of elementary operators that we use to specify hight level operators. The preconditions for the operators represent the non-compositional verification activities that should be performed in addition to verification of components to ensure the desired postcondition of the operator. We conclude by studying algebraic formalisation using concepts from category theory.

Les fonctions de commande et de contrôle sont parmi les plus importantes des systèmes embarqués critiques utilisés dans des activités telles les transports, la santé ou la gestion de l’énergie. Leur impact potentiel sur la sûreté de fonctionnement fait de la vérification de leur correction l’un des points les plus critiques de leur développement. Cette vérification est usuellement effectuée en accord avec les normes de certification décrivant un ensemble d’objectifs à atteindre afin d’assurer un haut niveau de qualité du système et donc de prévenir l’apparition de défauts. Cette vérification du logiciel est traditionnellement basée sur de nombreux tests et des activitiés de relectures de code, toutefois les versions les plus récentes des standards de certification permettent l’utilisation de nouvelles approches de développement telles que l’ingénierie dirigée par les modèles et les méthodes formelles ainsi que l’utilisation d’outil pour assister les processus de développement. Les outils de génération automatique de code sont exploités dans la plupart des processus de développement de systèmes embarqués critiques afin d’éviter des erreurs de programmation liées à l’humain et pour assurer le respect des règles de production de code. Ces outils ayant pour vocation de remplacer les humains pour la production de code, des erreurs dans leur conception peuvent causer l’apparition d’erreurs dans le code généré. Il est donc nécessaire de vérifier que le niveau de qualité de l’outil est le même que celui du code produit en s’assurant que les objectifs spécifiées dans les normes de qualification sont couverts. Nos travaux visent à exploiter l’ingénierie dirigée par les modèles et les méthodes formelles pour développer ces outils et ainsi atteindre un niveau de qualité plus élevé que les approches traditionnelles. Les fonctions critiques de commande et de contrôle sont en grande partie conçues à l’aide de langages graphiques à flot de données. Ces langages sont utilisés pour modéliser des systèmes complexes à l’aide de blocs élémentaires groupés dans des librairies de blocs. Un bloc peut être un objet logiciel sophistiqué exposant une haute variabilité tant structurelle que sémantique. Cette variabilité est à la fois liée aux valeurs des paramètres du bloc ainsi qu’à son contexte d’utilisation. Dans notre travail, nous concentrons notre attention en premier lieu sur la spécification formelle de ces blocs ainsi que sur la vérification de ces spécifications. Nous avons évalué plusieurs approches et techniques dans le but d’assurer une spécification formelle, structurellement cohérente, vérifiable et réutilisable des blocs. Nous avons finalement conçu un langage basé sur l’ingénierie dirigées par les modèles dédié à cette tâche. Ce langage s’inspire des approches des lignes de produit logiciel afin d’assurer une gestion de la variabilité des blocs à la fois correcte et assurant un passage à l’échelle. Nous avons appliqué cette approche et la vérification associée sur quelques exemples choisis de blocs issus d’applications industrielles et l’avons validé sur des prototypes logiciels que nous avons développé. Les blocs sont les principaux éléments des langages d’entrée utilisés pour la génération automatique de logiciels de commande et de contrôle. Nous montrons comment les spécifications formelles de blocs peuvent être transformées en des annotations de code afin de simplifier et d’automatiser la vérification du code généré. Les annotations de code sont vérifiées par la suite à l’aide d’outils spécialisés d’analyse statique de code. En utilisant des observateur synchrones pour exprimer des exigences de haut niveau sur les modèles en entrée du générateur, nous montrons comment la spécification formelle de blocs peut être utilisée pour la génération d’annotations de code et par la suite pour la vérification automatique des exigences. / Control and command softwares play a key role in safety-critical embedded systems used for human related activities such as transportation, healthcare or energy. Their impact on safety makes the assessment of their correctness the central point in their development activities. Such systems verification activities are usually conducted according to normative certification guidelines providing objectives to be reached in order to ensure development process reliability and thus prevent flaws. Verification activities usually relies on tests and proof reading of the software but recent versions of certification guidelines are taking into account the deployment of new development paradigms such as model-based development, and formal methods; or the use of tools in assistance of the development processes. Automatic code generators are used in most safety-critical embedded systems development in order to avoid human related software production errors and to ensure the respect of development quality standards. As these tools are supposed to replace humans in the software code production activities, errors in these tools may result in embedded software flaws. It is thus in turn mandatory to ensure the same level of correctness for the tool itself than for the expected produced code. Tools verification shall be done according to qualification guidelines. We advocate in our work the use of model-based development and formal methods for the development of these tools in order to reach a higher quality level. Critical control and command software are mostly designed using graphical dataflow languages. These languages are used to express complex systems relying on atomic operations embedded in blocks that are gathered in block libraries. Blocks may be sophisticated pieces of software with highly variable structure and semantics. This variability is dependent on the values of the block parameters and of the block's context of use. In our work, we focus on the formal specification and verification of such block based languages. We experimented various techniques in order to ensure a formal, sound, verifiable and usable specification for blocks. We developed a domain specific formal model-based language specifically tailored for the specification of structure and semantics of blocks. This specification language is inspired from software product line concepts in order to ensure a correct and scalable management of the blocks variability. We have applied this specification and verification approach on chosen block examples from common industrial use cases and we have validated it on tool prototypes. Blocks are the core elements of the input language of automatic code generators used for control and command systems development. We show how our blocks formal specification can be translated as code annotations in order to ease and automate the generated code verification. Code annotations are verified using specialised static code analysis tools. Relying on synchronous observers to express high level requirements at the input model level, we show how formal block specification can also be used for the translation of high level requirements as verifiable code annotations discharged using the same specialised tooling. We finally target the assistance of code generation tools qualification activities by arguing on the ability to automatically generate qualification data such as requirements, tests or simulation results for the verification and development of automatic code generators from the formal block specification.

Lors d’un développement logiciel, et plus particulièrement d’un développement d’applications embarquées avioniques, les activités de vérification représentent un coût élevé. Une des pistes prometteuses pour la réduction de ces coûts est l’utilisation de méthodes formelles. Ces méthodes s’appuient sur des fondements mathématiques et permettent d’effectuer des tâches de vérification à forte valeur ajoutée au cours du développement. Les méthodes formelles sont déjà utilisées dans l’industrie. Cependant, leur difficulté d’appréhension et la nécessité d’expertise pour leur mise en pratique sont un frein à leur utilisation massive. Parallèlement au problème des coûts liés à la vérification logicielle, vient se greffer la complexification des logiciels et du contexte de développement. L’Ingénierie Dirigée par les Modèles (IDM) permet de faire face à ces difficultés en proposant des modèles, ainsi que des activités pour en tirer profit.Le but des travaux présentés dans cette thèse est d’établir un lien entre les méthodes formelles et l’IDM afin de proposer à des utilisateurs non experts une approche de vérification formelle et automatique de programmes susceptible d’améliorer les processus de vérification actuels. Nous proposons de générer automatiquement sur le code source des annotations correspondant aux propriétés comportementales attendues du logiciel, et ce, à partir de son modèle de conception. Ces annotations peuvent ensuite être vérifiées par des outils de preuve déductive, afin de s’assurer que le comportement du code est conforme au modèle. Cette thèse CIFRE s’inscrit dans le cadre industriel d’Atos. Il est donc nécessaire de prendre en compte le contexte technique qui s’y rattache. Ainsi, nous utilisons le standard UML pour la modélisation,le langage C pour l’implémentation et l’outil Frama-C pour la preuve du code. Nous tenons également compte des contraintes du domaine du logiciel avionique dans lequel Atos est impliqué et notamment les contraintes liées à la certification.Les contributions de cette thèse sont la définition d’un sous-ensemble des machines à états UML dédié à la conception comportementale de logiciel avionique et conforme aux pratiques industrielles existantes, la définition d’un patron d’implémentation C, la définition de patrons de génération des propriétés comportementales sur le code à partir du modèle et enfin l’implémentation de l’approche dans un prototype compatible avec l’environnement de travail des utilisateurs potentiels en lien avec Atos. L’approche proposée est finalement évaluée par rapport à l’objectif de départ, par rapport aux attentes de la communauté du génie logiciel et par rapport aux travaux connexes. / During software development, and more specifically embedded avionics applications development, verification is very expensive. A promising lead to reduce its costs is the use of formal methods. Formal methods are mathematical techniques which allow performing rigorous and high-valued verification tasks during software development. They are already applied in industry. However, the high level of expertise required for their use is a major obstacle for their massive use. In addition to the verification costs issue, today software and their development are subject to an increase in complexity. Model Driven Engineering (MDE) allows dealing with these difficulties by offering models, and tasks to capitalize on these models all along the development lifecycle. The goal of this PhD thesis is to establish a link between formal methods and MDE in order to propose to non-expert users a formal and automatic software verification approach which helps to improve software verification processes. We propose to automatically generate annotations, corresponding to the expected behavioural properties of the software, from the design model to the source code. Then, these annotations can be verified using deductive proof tools in order to ensure that the behaviour of the code conforms to the design model. This PhD thesis takes place in the industrial context of Atos. So, it is necessary to take into account its technical specificities. We use UML for the design modeling, the C language for the software implementation and the Frama-C tool for the proof of this implementation. We also take into account the constraints of the avionics field in which Atos intervenes, and specifically the certification constraints. The contributions of this PhD thesis are the definition of a subset of UML state machine dedicated to the behavioural design of embedded avionics software and in line with current industrial practices, the definition of a C implementation pattern, the definition of generation patterns for the behavioural properties from the design model to the source code and the implementation of the whole approach in a prototype in accordance with the working environment of the potential users associated with Atos. The proposed approach is then assessed with respect to the starting goal of the thesis, to the expectation of the software engineering community and to related work.

Heutzutage sind wir von Geräten umgeben, die sowohl Hardware wie auch Software- Komponenten beinhalten. Diese Geräte unterstützen ein breites Spektrum an verschiedenen Domänen, so zum Beispiel Telekommunikation, Luftfahrt, Automobil und andere. Derartige Systeme sind überall aufzufinden und werden als Eingebettete Systeme bezeichnet, da sie zur Informationsverarbeitung in andere Produkte eingebettet werden, wobei die Informationsverarbeitung des eingebetteten Systems jedoch nicht die bezeichnende Funktion des Produkts ist. Die ständig zunehmende Komplexität moderner eingebettete Systeme erfordert die Verwendung von mehreren Komponenten um die Funktionen von einem einzelnen System zu implementieren. Eine solche Steigerung der Funktionalität führt jedoch ebenfalls zu einem Wachstum in der Entwurfs-Komplexität, die korrekt und effizient beherrscht werden muss. Neben hohen Anforderungen bezüglich Leistungsaufnahme, Performanz und Kosten hat auch Time-to-Market-Anforderungen großen Einfluss auf den Entwurf von Eingebetteten Systemen. Design Space Exploration (DSE) beschreibt die systematische Erzeugung und Auswertung von Entwurfs-Alternativen, um die Systemleistung zu optimieren und den gestellten Anforderungen an das System zu genügen. Bei der Entwicklung von Eingebetteten Systemen, speziell beim Platform-Based Design (PBD) führt die zunehmende Anzahl von Design-Entscheidungen auf mehreren Abstraktionsebenen zu einer Explosion der möglichen Kombinationen von Alternativen, was auch für aktuelle DSE Methoden eine Herausforderung darstellt. Jedoch vermag üblicherweise nur eine begrenzte Anzahl von Entwurfs-Alternativen die zusätzlich formulierten nicht-funktionalen Anforderungen zu erfüllen. Darüber hinaus beeinflusst jede Entwurfs- Entscheidung weitere Entscheidungen und damit die resultierenden Systemeigenschaften. Somit existieren Abhängigkeiten zwischen Entwurfs-Entscheidungen und deren Reihenfolge auf dem Weg zur Implementierung des Systems. Zudem gilt es zwischen einer spezifischen Heuristik für eine bestimmte DSE, welche zu verbesserten Optimierungsresultaten führt, sowie globalen Verfahren, welche ihrerseits zur Flexibilität hinsichtlich der Anwendbarkeit bei verschiedenen DSE Szenarien beitragen, abzuwägen. Um die genannten Herausforderungen zu lösen wird eine Modellgetriebene Entwicklung (englisch Model-Driven Engineering, kurz MDE) Methodik für DSE vorgeschlagen. Für diese Methodik wird ein DSE-Domain-Metamodell eingeführt um relevante DSEKonzepte wie Entwurfsraum, Entwurfs-Alternativen, Auswertungs- und Bewertungsverfahren, Einschränkungen und andere abzubilden. Darüber hinaus modelliert das Metamodell verschiedenen DSE-Frage- stellungen, was zur Verbesserung der Flexibilität der vorgeschlagenen Methodik beiträgt. Zur Umsetzung von DSE-Regeln, welche zur Steuerung, Einschränkung und Generierung der Ent- wurfs-Alternativen genutzt werden, finden Modell-zu-Modell-Transformationen Anwendung. Durch die Fokussierung auf die Zuordnung zwischen den Schichten in einem PBDAnsatz wird eine neuartige Entwurfsraumabstraktion eingeführt, um multiple Entwurfsentscheidungen als singuläres DSE Problem zu repräsentieren. Diese auf dem Categorial Graph Product aufbauende Abstraktion entkoppelt den Explorations-Algorithmus vom Entwurfsraum und ist für Umsetzung in automatisierte Werkzeugketten gut geeignet. Basierend auf dieser Abstraktion profitiert die DSE-Methode durch die eingeführte MDEMethodik als solche und ermöglicht nunmehr neue Optimierungsmöglichkeiten sowie die Verbesserung der Integration von DSE in Entwicklungsprozesse und die Spezifikation von DSE-Szenarien. / Atualmente dispositivos contendo hardware e software são encontrados em todos os lugares. Estes dispositivos prestam suporte a uma varieadade de domínios, como telecomunicações, automotivo e outros. Eles são chamados “sistemas embarcados”, pois são sistemas de processamento montados dentro de produtos, cujo sistema de processamento não faz parte da funcionalidade principal do produto. O acréscimo de funções nestes sistemas implica no aumento da complexidade de seu projeto, o qual deve ser adequadamente gerenciado, pois além de requisitos rigorosos em relação à dissipação de potência, desempenho e custos, a pressão sobre o prazo para introdução de um produto no mercado também dificulta seu projeto. Exploração do espaço de projeto (DSE) é a atividade sistemática de gerar e avaliar alternativas de projetos, com o objetivo de otimizar suas propriedades. No desenvolvimento de sistemas embarcados, especialmente em Projeto Baseado em Plataformas (PBD), metodologias de DSE atuais são desafiadas pelo crescimento do número de decisões de projeto, o qual implica na explosão da combinação de alternativas. Porém, somente algumas destas resultam em projetos que atedem os requisitos nãofuncionais. Além disso, as decisões influenciam umas às outras, de forma que a ordem em que estas são tomadas alteram a implementação final do sistema. Outro desafio é o balanço entre flexibilidade da metodologia e seu desempenho, pois métodos globais de otimização são flexíveis, mas apresentam baixo desempenho. Já heurísticas especialmente desenvolvidas para o cenário de DSE em questão apresentam melhor desempenho, porém dificilmente são aplicáveis a diferentes cenários. Com o intuito de superar os desafios é proposta uma metodologia de projeto dirigido por modelos (MDE) adquada para DSE. Um metamodelo do domínio de DSE é definido para representar conceitos como espaço de projeto, métodos de avaliação e restrições. O metamodelo também representa diferentes problemas de DSE aprimorando a flexibilidade da metodologia. Regras de transformações de modelos implementam as regras de DSE, as quais são utilizadas para restringir e guiar a geração de projetos alternativos. Restringindo-se ao mapeamento entre camadas no PBD é proposta uma abstração para representar o espaço de projeto. Ela representa múltiplas decisões de projeto envolvidas no mapeamento como um único problema de DSE. Esta representação é adequada para a implementação em ferramentas automática de DSE e pode beneficiar o processo de DSE com uma abordagem de MDE, aprimorando a especificação de cenários de DSE e sua integração no processo de desenvolvimento. / Nowadays we are surrounded by devices containing hardware and software components. These devices support a wide spectrum of different domains, such as telecommunication, avionics, automobile, and others. They are found anywhere, and so they are called Embedded Systems, as they are information processing systems embedded into enclosing products, where the processing system is not the main functionality of the product. The ever growing complexity in modern embedded systems requires the utilization of more components to implement the functions of a single system. Such an increasing functionality leads to a growth in the design complexity, which must be managed properly, because besides stringent requirements regarding power, performance and cost, also time-to-market hinders the design of embedded systems. Design Space Exploration (DSE) is the systematic generation and evaluation of design alternatives, in order to optimize system properties and fulfill requirements. In embedded system development, specifically in Platform-Based Design (PBD), current DSE methodologies are challenged by the increasing number of design decisions at multiple abstraction levels, which leads to an explosion of combination of alternatives. However, only a reduced number of these alternatives leads to feasible designs, which fulfill non-functional requirements. Moreover, each design decision influences subsequent decisions and system properties, hence there are inter-dependencies between design decisions, so that the order decisions are made matters to the final system implementation. Furthermore, there is a trade-off between heuristics for specific DSE, which improves the optimization results, and global optimizers, which improve the flexibility to be applied in different DSE scenarios. In order to overcome the identified challenges an MDE methodology for DSE is proposed. For this methodology a DSE Domain metamodel is proposed to represent relevant DSE concepts such as design space, design alternatives, evaluation method, constraints and others. Moreover, this metamodel represents different DSE problems, improving the flexibility of the proposed framework. Model transformations are used to implement DSE rules, which are used to constrain, guide, and generate design candidates. Focusing on the mapping between layers in a PBD approach, a novel design space abstraction is provided to represent multiple design decisions involved in the mapping as a single DSE problem. This abstraction is based on Categorical Graph Product, decoupling the exploration algorithm from the design space and being well suited to be implemented in automatic exploration tools. Upon this abstraction, the DSE method can benefit from the MDE methodology, opening new optimization opportunities, and improving the DSE integration into the development process and specification of DSE scenarios.

Atualmente, o projeto de sistemas tempo-real embarcados e distribuídos está crescendo em complexidade devido à sua natureza heterogênea e ao crescente número e diversidade de funções que um único sistema desempenha. Sistemas de automação industrial, sistemas eletrônicos em automóveis e veículos aéreos, equipamentos médicos, entre outros, são exemplos de tais sistemas. Tais sistemas são compostos por componentes distintos (blocos de hardware e software), os quais geralmente são projetados concorrentemente utilizando modelos, ferramentas e linguagens de especificação e implementação diferentes. Além disso, estes sistemas tem requisitos específicos e importantes, os quais não representam (por si só) as funcionalidades esperadas do sistema, mas podem afetar a forma como o sistema executa suas funcionalidades e são muito importantes para a realização do projeto com sucesso. Os chamados requisitos não-funcionais são difíceis de tratar durante todo o ciclo de projeto porque normalmente um único requisito não-funcional afeta vários componentes diferentes. A presente tese de doutorado propõe a integração automatizada das fases de projeto de sistemas tempo-real embarcados e distribuídos focando em aplicações na área da automação. A abordagem proposta usa técnicas de engenharia guiada por modelos (do inglês Model Driven Engineering ou MDE) e projeto orientado a aspectos (do inglês Aspect-Oriented Design ou AOD) juntamente com o uso de plataformas previamente desenvolvidas (ou desenvolvida por terceiros) para projetar os componentes de sistemas tempo-real embarcados e distribuídos. Adicionalmente, os conceitos de AOD permitem a separação no tratamento dos requisitos de naturezas diferentes (i.e. requisitos funcionais e não-funcionais), melhorando a modularização dos artefatos produzidos (e.g. modelos de especificação, código fonte, etc.). Além disso, esta tese propõe uma ferramenta de geração de código, que suporta a transição automática das fases iniciais de especificação para as fases seguintes de implementação. Esta ferramenta usa um conjunto de regras de mapeamento, que descrevem como elementos nos níveis mais altos de abstração são mapeados (ou transformados) em elementos dos níveis mais baixos de abstração. Em outras palavras, tais regras de mapeamento permitem a transformação automática da especificação inicial, as quais estão mais próximo do domínio da aplicação, em código fonte para os componentes de hardware e software, os quais podem ser compilados e sintetizados por outras ferramentas para se obter a realização/implementação do sistema tempo-real embarcado e distribuído. / Currently, the design of distributed embedded real-time systems is growing in complexity due to the increasing amount of distinct functionalities that a single system must perform, and also to concerns related to designing different kinds of components. Industrial automation systems, embedded electronics systems in automobiles or aerial vehicles, medical equipments and others are examples of such systems, which includes distinct components (e.g. hardware and software ones) that are usually designed concurrently using distinct models, tools, specification, and implementation languages. Moreover, these systems have domain specific and important requirements, which do not represent by themselves the expected functionalities, but can affect both the way that the system performs its functionalities as well as the overall design success. The so-called nonfunctional requirements are difficult to deal with during the whole design because usually a single non-functional requirement affects several distinct components. This thesis proposes an automated integration of distributed embedded real-time systems design phases focusing on automation systems. The proposed approach uses Model- Driven Engineering (MDE) techniques together with Aspect-Oriented Design (AOD) and previously developed (or third party) hardware and software platforms to design the components of distributed embedded real-time systems. Additionally, AOD concepts allow a separate handling of requirement with distinct natures (i.e. functional and non-functional requirements), improving the produced artifacts modularization (e.g. specification model, source code, etc.). In addition, this thesis proposes a code generation tool, which supports an automatic transition from the initial specification phases to the following implementation phases. This tool uses a set of mapping rules, describing how elements at higher abstraction levels are mapped (or transformed) into lower abstraction level elements. In other words, suchmapping rules allow an automatic transformation of the initial specification, which is closer to the application domain, in source code for software and hardware components that can be compiled or synthesized by other tools, obtaining the realization/ implementation of the distributed embedded real-time system.

Le système PLM est l’un des outils stratégiques de l’entreprise. Ces systèmes sont sujets à des changements récurrents dans l’entreprise. Les évolutions organisationnelles, le changement de l’offre produit ou encore le remplacement de logiciels PLM peuvent déclencher l’évolution du système d’information PLM. Une des structures les plus importantes dans les systèmes PLM est le modèle du produit, autour duquel s’articule les informations et processus. C’est autour du modèle produit que se concentrent nos recherches. Les causes d’évolution des modèles produits sont des éléments signifiants qui différencient les étapes de la démarche à suivre. Les méthodes d’IDM sont utilisées afin de formaliser la transformation des modèles. En plus, cette démarche bénéfice d’un cadre de similarité spécialement développé pour la configuration de produit. Un exemple industriel est illustré et résolu en appliquant cette démarche. Il s’agit de l’évolution d’un système gérant les modèles spécifiques de produit vers un système qui est capable de construire et d’utiliser les modèles génériques de produit. Un outil informatique support à nos travaux est développé dans le cadre d'Eclipse. / PLM systems are among the strategic components of enterprise’s information system architecture. These systems undergo frequent evolutions of enterprise. Organizational evolution or product offer variation as well as PLM application replacement may launch PLM systems’ evolution.One of the important structures in PLM systems is the product configuration, which organize and structure all product’s information and processes. Our research activities concern product model evolution. Reasons of product model evolution specify the appropriate methodology and necessary steps in order to handle it. MDE methods are used to formalize the model transformation process.Moreover, our methodology contains a specific similarity framework dedicated to product configuration. An industrial example was illustrated and resolved by this methodology. The problematic of this example is the migration of a system which manage only specific product configuration to a new system that is capable to construct and use generic models of product.