The modeling of most large and complex systems, such as embedded, cyber-physical, or distributed systems, necessarily involves many designers. The multiple stakeholders carry their own perspectives of the system under development in order to meet a variety of objectives, and hence they derive their own models for the same system. This practice is known as multiview modeling, where the distinct models of a system are called views. Inevitably, the separate views are related, and possible overlaps may give rise to inconsistencies. Checking for multiview consistency is key to multi-view modeling approaches, especially when a global model for the system is absent, and can only be synthesized from the views. The present thesis provides an overview of the representative related work in multi-view modeling, and contributes to the formal study of multi-view modeling and the multi-view consistency problem for views and systems described as sets of behaviors. In particular, two distinct settings are investigated, namely, infinitary languages, and discrete systems. In the former research, a system and its views are described by mixed automata, which accept both finite and infinite words, and the corresponding infinitary languages. The views are obtained from the system by projections of an alphabet of events (system domain) onto a subalphabet (view domain), while inverse projections are used in the other direction. A systematic study is provided for mixed automata, and their languages are proved to be closed under union, intersection, complementation, projection, and inverse projection. In the sequel, these results are used in order to solve the multi-view consistency problem in the infinitary language setting. The second research introduces the notion of periodic sampling abstraction functions, and investigates the multi-view consistency problem for symbolic discrete systems with respect to these functions. Apart from periodic samplings, inverse periodic samplings are also introduced, and the closure of discrete systems under these operations is investigated. Then, three variations of the multi-view consistency problem are considered, and their relations are discussed. Moreover, an algorithm is provided for detecting view inconsistencies. The algorithm is sound but it may fail to detect all inconsistencies, as it relies on a state-based reachability, and inconsistencies may also involve the transition structure of the system.
|Tila||Julkaistu - 2017|