The design and development of large and complex systems, such as System of Systems (SoS) with autonomous constituent systems is gaining increasing importance in many application domains (e.g., medical, aerospace, military and transportation). The wide scale and huge number of interactions involved in SoS makes it difficult to model and analyse the SoS architecture through state-of-the-art techniques for modelling, architecture development and analysis of monolithic systems. SoS are large-scale concurrent and distributed systems that are comprised of autonomous constituent systems with operational and managerial independence. The characteristics of SoS result in various challenges during the system design. It is necessary to consider emergent behaviour, an evolutionary development process, and an increased state space. In particular, these challenges are unsolved when moving form directed SoS, where the system has central control, towards virtual SoS with no central management and no common purpose ascribed to the constituent systems.
In addition to meeting functional requirements, a major challenge is the design space exploration and optimisation of non-functional properties among heterogeneous constituent systems such as reliability, timeliness, safety and security. Many SoS are real-time systems where timing requirements are central to the development process. The failure in meeting these requirements in the late stages of the SoS development can be avoided by introducing them already in the architecture development. Likewise, many SoS represent critical infrastructures that provide safety-relevant services to the users and the environment. Therefore, the specified reliability of the SoS must be guaranteed despite failures of individual constituent systems using fault-tolerance mechanisms.
The state-of-the-art modelling approaches describe the SoS using architecture frameworks such as the US Department of Defence Architecture Framework (DoDAF), British Ministry of Defence Architecture Framework (MODAF) and NATO Architecture Framework (NAF). However, these frameworks do not offer a precise connection between the different views of the SoS, which is needed to enable design exploration technologies, reusability and design automation using integrated tool chains. In addition, significant non-functional system properties (e.g., real time, reliability) are not addressed in the state-of-the-art modelling frameworks.
This thesis addresses the research gap by establishing a modelling framework that extends the current SoS frameworks to link the different views, and to satisfy real-time and reliability requirements. We provide a methodology that uses design exploration techniques and tools for SoS architecture optimization. The use of architecture patterns is introduced to enhance the model re-usability and facilitate the model evolution.
Scientific contributions beyond the state-of-the-art include a Model Based System Engineering (MBSE) methodology for SoS architecting based on the Unified Profile for DoDAF and MODAF (UPDM) with extensions to support significant SoS properties such as timing requirements and reliability. The methodology supports an architecture optimisation process based on Mixed Integer Linear Programming (MILP) connected to architecture patterns that enhance the model reusability. The developed modelling language and design methods are evaluated using simulations and example scenarios.