Les systèmes informatiques sont devenus de plus en plus distribués et hétérogènes, ce qui rend leur administration manuelle difficile et source d'erreurs. L'administration autonome a été proposée comme solution à ce problème. Elle consiste à automatiser l'administration des systèmes informatiques à l'aide de boucles de contrôle appelées gestionnaires autonomes. De nombreux travaux de recherche se sont intéressés à l'automatisation des fonctions d'administration de systèmes informatiques et aujourd'hui, beaucoup de gestionnaires autonomes sont disponibles. Toutefois, les gestionnaires autonomes existants sont, la plupart, spécialisés dans la gestion de quelques aspects d'administration. Cela rend nécessaire la coexistence de plusieurs gestionnaires autonomes pour atteindre une gestion globale des systèmes. La coexistence de plusieurs gestionnaires permet la gestion de plusieurs aspects, mais nécessite des mécanismes de coordination afin d'éviter des décisions incohérentes. Nous étudions l'utilisation de techniques de contrôle pour la conception de contrôleurs de coordination, nous utilisons la programmation synchrone qui fournit des méthodes formelles, et la synthèse de contrôleur discret pour automatiser la construction de contrôleur. Nous suivons une approche à base de composants, et utilisons le contrôle discret modulaire qui permet de décomposer la complexité combinatoire inhérente à la technique d'exploration d'espace d'états. Cela améliore le passage à l'échelle de notre approche et permet la construction d'un contrôle hiérarchique. Notre approche permet la réutilisation de gestionnaires complexes dans des contextes différents, sans modifier leurs spécifications de contrôle. Nous construisons une coordination de gestionnaires basée sur le modèle à composants offrant introspection, adaptabilité et reconfiguration. Cette thèse présente notre méthodologie et des études de cas. Nous évaluons et démontrons les avantages de notre approche par la coordination de gestionnaires autonomes dédiés à la gestion de la disponibilité, et à la gestion de la performance et l'optimisation de ressources. / Computing systems have become more and more distributed and heterogeneous, making their manual administration difficult and error-prone. The Autonomic Computing approach has been proposed to overcome this issue, by automating the administration of computing systems with the help of control loops called autonomic managers. Many research works have investigated the automation of the administration functions of computing systems and today many autonomic managers are available. However the existing autonomic managers are mostly specialized in the management of few administration concerns. This makes necessary the coexistence of multiple autonomic managers for achieving a global system management. The coexistence of several managers make possible to address multiple concerns, yet requires coordination mechanisms to avoid incoherent management decisions. We investigate the use of control techniques for the design of coordination controllers, for which we exercise synchronous programming that provide formal semantics, and discrete controller synthesis to automate the construction of the controller. We follow a component-based approach, and explore modular discrete control allowing to break down the combinatorial complexity inherent to the state-space exploration technique. This improves scalability of the approach and allows constructing a hierarchical control. It also allows re-using complex managers in different contexts without modifying their control specifications. We build a component-based coordination of managers, with introspection, adaptivity and reconfiguration. This thesis details our methodology and presents case-studies. We evaluate and demonstrate the benefits of our approach by coordinating autonomic managers which addresse the management of availability, and the management of performance and resources optimization.

Driven by the need for automation and autonomy as well as the need to reduce resources and emissions, the automotive industry is currently undergoing a major transformation. Technologically, this transformation is addressing a wide range of challenges and opportunities. The optimal control of all components is significant for the sustainable development and eco-friendly operation of vehicles. Additionally, robust control of the actuators forms the basis for the development of driver assistance systems and functions for autonomous driving. The actuators of the steering system are particularly important, as they enable safe and comfortable lateral vehicle control. Therefore, the model-based development and virtual simulation of an innovative highly robust control approach for modern Steer-by-Wire systems were conducted in this thesis. The approaches and algorithms described in this thesis allow the design of robust Steer-by-Wire systems and offer the opportunity to conduct many investigations in a computer-aided virtual environment at an early stage in the development process. This reduces time- and cost-intensive testing on prototypes, avoids unnecessary iterations in the design and significantly increases the efficiency and quality of the development. The desired high degree of robustness of the steering control also ensures that the parameterization of the steering feel generator can be freely selected for the individual application. This enables safe and comfortable vehicle lateral control.In summary, the research results described in this thesis accelerate the development of new, modern Steer-by-Wire systems whose robust design forms the basis for the realization of functions for highly automated and autonomous driving.

Die steigende Nachfrage an immer komplexeren Systemen in verschiedensten wirtschaftlichen Bereichen, erfordert Strategien, die Wartbarkeit und Wiederverwendbarkeit unterstützen. An diesem Punkt setzen service-orientierte Architekturen (SOAn) an. Dieses Paradigma fordert die Aufspaltung von Funktionalität in Services, die komponiert werden können, um eine gewünschte, komplexe Funktionalität zu erreichen. Besonders in sicherheitskritischen Bereichen, kann eine fehlerbehaftete Komposition jedoch zu hohen finanziellen Einbußen oder sogar zu lebensbedrohlichen Situationen führen. Um die Korrektheit sicherzustellen, müssen Kompositionsmethoden im Vorfeld definierte Eigenschaften garantieren und die, durch die unabhängige Entwicklung auftretenden, Interface-Inkompatibilitäten behandeln. Existierende Ansätze zur automatisierten Service-Komposition durch Controller-Synthese beinhalten jedoch keine formale Datenbehandlung und können daher nicht mit datenabhängigem Verhalten umgehen. In der vorliegenden Arbeit, löse ich dieses Problem durch die Bereitstellung eines Ansatzes zur automatisierten Synthese datenabhängiger, korrekter Service-Controller. Dabei wird ein Controller direkt aus den spezifizierten Anforderungen und dem Verhalten der Services erzeugt. Basierend auf den Annahmen, dass die Anforderungen in RCTL, einer Untermenge der Computational Tree Logic (CTL), spezifiziert und die Services als Algebraische Petrinetze (APNe) gegeben sind, vereinigt mein neuartiger Ansatz die beiden Formalismen und unterstützt eine zuverlässige Extraktion des Controller-Verhaltens. Durch die Nutzung der APNe, erlaubt der Ansatz eine formale Datenbehandlung und somit eine Betrachtung datenabhängigen Verhaltens. Die Anwendbarkeit meines Ansatzes habe ich an drei Fallstudien aus dem medizinischen Bereich gezeigt, wo Geräte sicher miteinander kommunizieren müssen. / The continuously increasing demand for more complex systems in various economical domains requires a strategy that supports maintainability and reusability. This is addressed by the service-oriented architecture (SOA)}-paradigm that encourages the encapsulation of functionality into services. To achieve a specific functionality, services can be composed. Especially in safety-critical systems, an incorrect composition of various components can lead to high financial losses or even life threatening situations. To ensure the correctness, composition methods must particularly be able to guarantee pre-specified requirements and to overcome interface incompatibilities, which result from the independent development of the single services. However, current approaches for automated service composition via controller synthesis do not support a formal data-treatment and do not cope with data-dependent behavior. In this thesis, we overcome this problem by providing an approach for the automated synthesis of data-dependent service controllers that are correct-by-construction. The core idea is to synthesize such a controller directly from given requirements and the behavior of the services. Based on the assumptions that the requirements are specified using a subset of Computational Tree Logic (CTL), called RCTL, and that the services are given as algebraic Petri Nets (APNs), our novel synthesis process unifies the two formalisms and enables a reliable extraction of the controller behavior. Especially due to the use of APNs, our approach supports a formal data-treatment and enables a consideration of data-dependent behavior. With our synthesis process, which is based on a successive combination of requirements and services, we provide a practical applicable approach that works fully automatically. We show the applicability of our approach using three case studies in which medical devices interact with each other.