A Conceptual Model for Reusable Simulations
Within a Model-Simulator-Context Framework

Levent Yilmaz
Tuncer I. Ören
The Auburn Modeling and Simulation
Laboratory (AMSL) of the M&SNet
Computer Science and Engineering
Auburn University
Ottawa Center of the McLeod Institute
of Simulation Sciences (OC-MISS) of
the M&SNet
SITE, University of Ottawa
Auburn, AL, USA
yilmaz@eng.auburn.edu
Ottawa, ON, Canada
oren@site.uottawa.ca

Abstract

Model reuse is a longstanding challenge within
the simulation modeling community. While other
disciplines successfully apply component-based
approach to build systems, it has proven more
difficult to apply in simulation development. A
model is reusable to the extent that the
assumptions and constraints of the experimental
frame it is situated in are satisfied. Hence,
without contextual information a conceptual
model and its simulation may not effectively be
used in a new experiment. While the value of an
experimental frame for reuse is clear and well
documented, one of the least appreciated, but
significant aspect central to reuse is the
characterization of the context of the model
itself. To this end, this paper explores the
extension of the basic model-simulatorexperimental
frame viewpoint to assert the role
of context in reuse. Furthermore, the separation
of model’s concept definition and its realization
is emphasized to facilitate sound and formal
reasoning for reuse. Based on that perspective
model qualification is defined in a formal
manner to conceptualize reusability in terms of
formally specified behavioral dependencies.
1 Introduction
Given the significant benefits of reuse in terms
of gained productivity (Basili et al. 1996) and
the potential for increased quality, model reuse is
widely promoted and encouraged, particularly
within the DoD community (Davis and Anderson
2003). The significance and benefits of reuse
libraries is discussed in (Ören 2002). Discussions
on credibility of composable simulations are also
of significant interest. For instance, (Weisel et al.
2003) elaborates on the validity of compositions
in terms of formal metrics for semantic
composability. A more general overview of the
challenges and difficulties involved in model
reuse are discussed by Overstreet et al. (2002).
To achieve progress in reuse, they suggest a
variety of significant development areas,
including capturing specifications of model
constraints, objectives, and assumptions of a
model. Similarly, Ören and Zeigler (1979) argue
about the importance of high level model
specification formalisms and their role in reuse,
as well as maintenance. Reuse and model
composition has been studied in qualitative
simulation (Falkenheimer and Forbus 1991,
Forbus 1984) and software engineering, where
one important goal is to formalize the modeling
process using an underlying domain model.
Nayak’s 1995 ACM Distinguished Dissertation
showed that the general model selection problem
for application composition is NP-hard (Nayak
1992). Others have shown that deciding whether
an identified collection of submodels meet a
stated set of objectives is an NP-complete
problem (Page and Opper 1999).
Given the impracticality of complete automation
of the composition of models to satisfy a set of
objectives, the reuse community within the
modeling and simulation community focused on
developing infrastructures such as HLA
(Dahmann et al. 1997), and product line
architectural frameworks, such as ONESAF to
facilitate reuse through common standards.
Developing such common standards and

.....