Ontologies and tools for analysing and composing simulation confederations for the training and testing domains

Abstract

Military training and testing events integrate a diverse set of live and simulated systems, most of which were built independently and weren’t specifically designed to work together. Data interoperability and service-oriented architecture approaches, while essential, do not provide a complete solution to ensuring that systems will be fully compatible in their interactions. We describe a complementary approach that uses Web Ontology Language and Semantic Web Rule Language to capture information about the roles and capabilities required to complete a task, and the detailed attributes of candidate resources. Our toolset applies automated reasoning to determine whether each candidate resource has the requisite capabilities and is compatible with other resources. If there are multiple candidates for a role, the reasoner ranks the relative goodness of each with respect to constraints and metrics that are appropriate for the specific task needs of the exercise or deployment. We include worked examples illustrating the kinds of information we capture about resources and how rules and constraints are applied to provide a nuanced assessment of their compatibility in a specific context.

Keywords:

• interoperability
• ontology
• rule languages
• automated analysis

Acknowledgements

ONISTT is sponsored by ODUSD/R/RTPP (Training Transformation). ANSC is sponsored by USD/AT&L-TRMC (S&T Portfolio). JTEOW is jointly sponsored by USJFCOM JWFC and ODUSD/R/RTPP.

Notes

1 This paper is an expansion and revision of an earlier paper (Ford et al 2009) prepared for Winter Simulation Conference 2009 (WSC, 2009).

2 We intentionally use the term confederation here to emphasize the heterogeneous architectures of the large LVC simulations that our work has focused on; confederation is intended to convey the absence of a central governing body, as is typically found in a federation.

3 Currently, our approach to scoring is a simple ‘penalty points’ system: for each failed constraint, penalty points are assigned, and lower scores are better solutions; solutions with a score of zero indicate that no constraints were violated.