Steiniger, Alexander (2018) Toward Composing Variable Structure Models and Their Interfaces: A Case of Intensional Coupling Definitions. PhD thesis, Institute of Computer Science, University of Rostock.
Full text not available from this repository.Abstract
Modeling and simulation are well established tools to study intriguing systems, both real and imaginary. For this, systems of interest need to be represented by formal models capturing the systems' essential behavior, while abstracting from irrelevant aspects. Many systems of interest are inherently complex, i. e., they consists of numerous homogeneous or heterogeneous components, where each component can be viewed as a system of its own. The behavior of such a complex system emerges from the interaction of its components and can often not be understood by studying the components in isolation. In addition, the structure of complex systems is often variable, i. e., can change over time. Systems with a time-variant structure are, e. g., socio-technical systems or biological systems. Component-based modeling and simulation takes the structural complexity of systems under study as well as the correctness and consistency of model compositions representing these systems into account. Creating a model of a complex system by pursuing a component-based approach (i. e., by composition) allows the modeler to reduce (i) the complexity of individual model units, and (ii) the development costs by reusing already existing components. Variable structure modeling, on the other hand, deals with the structural variability of systems, by allowing the modeler to explicitly reflect structure changes in the models representing those systems (i. e., system specifications). Variable structure modeling, similar to component-based modeling, enables the modeler to reduce the complexity of system specifications. Both component-based modeling and variable structure modeling describe and focus on a similar aspect of a model, that is its structure. However, traditional component-based modeling assumes a static model structure, whereas variable structure modeling often does not provide means as sophisticated as those provided by component-based modeling to specify complex models as an assembly of reusable, self-contained, replaceable, retrievable, customizable, and interoperable components, which can be used in different contexts or by third parties. In this thesis, we investigate a combination of both kinds of modeling approaches and discuss its implications and limitations. The focus is on a structural consistent specification of couplings in modular, hierarchical models with a variable structure. For this, we exploit intensional definitions, as known from logic, and introduce a novel intensional coupling definition, which allows a concise yet expressive specification of complex communication and interaction patterns in static as well as variable structure models, without the need to worry about structural consistency. The intensional coupling definition is frequently translated into a concrete coupling scheme by the respective simulation algorithm, which takes the current state and structure of the model into account and guarantees the correctness of the derived concrete couplings. Furthermore, we emphasize model (component) interfaces based upon which couplings are defined intensionally. As a proof of concept, the introduced intensional coupling mechanism is realized as a part of ML-DEVS, a modular-hierarchical, systemtheoretic modeling formalism, which allows variable structure and multi-level modeling. The abstract simulator, i. e., simulation algorithm, of ML-DEVS illuminates how an intensional coupling definition can be translated into a concrete coupling scheme during simulation while ensuring structural consistency. At the end of this thesis, we briefly discuss how intensional definitions can help modelers to streamline their models further.
| Item Type: | Thesis (PhD) |
|---|
