Model Composability: Experiences with Military Simulation Systems Dr. Marko A. Hofmann Institute for Technology of Intelligent Systems (ITIS) and Institute for Applied Systems Science and Operations Research (IASFOR) University of the Federal Armed Forces Munich Heisenbergweg 39 Germany - 85577 Neubiberg Tel: ++49 89 6004 3242 Fax: ++49 89 6004 3036 marko@informatik.unibw-muenchen.de Abstract : Model designers necessarily abstract and simplify reality according to their own perception, preferences, purposes and conceptions, thus models - even of the same real system - regularly differ in several aspects. Lessons learned with military simulation system indicate that technical and syntactic aspects of model composition, which have been in the focus of interest for almost a decade, are not sufficient to guarantee successful interaction. The crucial task is to ensure that syntactical structures are attributed with the same meanings in all involved models, that the same actions are triggered by identical orders and reports and last but not least that the abstractions from reality to the models do not contradict themselves. From this point of view composability becomes a much more challenging task than the mere coupling of software components. 1. INTRODUCTION In order to successfully build whole executable models (programs) with model components it is necessary to ensure that the conceptual models underlying the executable models are in line with each other. But since model designers necessarily abstract and simplify reality according to their own perception, preferences, purposes and conceptions, models regularly differ in several aspects. Experiences with military simulation system have shown that there are essential preconditions for coupling such “model based” information systems (see for example [1] and [2]). Our results indicate that technical and syntactic preconditions, which have been in the focus of interest for almost a decade, are not sufficient to guarantee a successful interaction. The crucial tasks are to ensure that syntactical structures are attributed with the same meanings in all involved models and that the same actions are triggered by identical orders and reports. These findings have been confirmed during a study we performed for the German Armed Forces addressing the standardization of command and control components for different Army simulation systems. As a consequence of the importance of meanings and triggered actions we have chosen a “linguistic” approach to understand the problems of interoperability, which is based on the idea of successful communication between models/model users. In linguistics one generally presupposes the existence of a technical communication channel (which is so important in computer science) and concentrates on the three semiotic aspects of language which are syntax, semantics and pragmatics. Linguistics provide a perfect framework for investigations into the meaning of interactions, since the whole point of setting up a theory of semantics and pragmatics is to provide a systematic account of the nature of meaning. Communication is successful if and only if sender and receiver have common knowledge on all semiotic aspects. In addition, composability is restricted to models whose abstractions from reality or, more general, whose conceptual modelling ideas are in line with each other. Thus, interoperability of model components is much more than the coupling of simple technical devices, which has been the predominant but misleading idea in the field of military simulations for almost a decade. Some of the costly consequences of this misconception can be found in [3]. This paper aims at a further clarification of the challenges of coupling models. Since the mere presentation of coupling problems stemming from a limited amount of models would be too anecdotic and idiosyncratic to be convincing, the paper first introduces a general framework (sections 2 – 4) in which the examples presented in chapter 5 fit as illustrations of the basic ideas. 2. MODEL BASED INFORMATION PROCESSING SYSTEMS From a general point of view information processing systems can be distinguished into direct and intermediate control information systems (see Figure 1). Most of the (artificial) information processing systems used today are embedded into real systems in which they operate as control units. Their task is to ensure that the state transition of the real system stays within a given trajectory. Such information processing systems exert direct control .....