Intensional Couplings in Variable-Structure Models

Abstract

In modular, hierarchical modeling, couplings (connections) describe and constrain the communication, and thus interaction, between model components. Defining couplings between a large set of components in an extensional manner—listing all existing couplings individually—often proves to be rather tedious. Moreover, if models change their structure, that is, composition and interaction patterns and, in some cases, even their interfaces during simulation, questions about the consistency of the couplings arise. For instance, an extensionally defined coupling may refer to a model that no longer exists. Instead, an intensional coupling definition, based on attributes of the components to couple and dynamically translated into concrete couplings during simulation, promises to alleviate these problems. We propose a concept that combines a flexible, yet expressive, definition of couplings that rests on component interfaces announcing attributes of interest. However, intensional couplings come at a price, as they need to be translated during simulation; in variable-structure models, this translation has to happen frequently. We illuminate our concept based on a revision of the modeling formalism Multilevel Discrete Event System Specification ( ML - DEVS ). Developed for multilevel modeling and simulation, ML - DEVS exhibits another alternative to intensional couplings, that is, sharing parts of model states for up- and downward causation. The intricate interplay between these different types of couplings is revealed in the abstract simulator of ML - DEVS .