Consistency Maintenance in Concurrent Representations
Natrajan, Anand, Department of Computer Science, University of Virginia
Reynolds, Paul, Department of Computer Science, University of Virginia
Multi-Representation Modeling (MRM) involves executing multiple models of the same phenomenon jointly. MRM is a technique in modeling and simulation for capturing the combined semantics of multiple models. Previous MRM approaches, such as selective viewing and aggregation-disaggregation, have encountered problems such as chain disaggregation, temporal inconsistency and mapping inconsistency. Eliminating these problems has been a difficult task for MRM designers. We eliminate these problems by showing how to achieve MRM effectively, i.e., correctly, consistently and inexpensively. Our thesis is that MRM can be effective. Maintaining consistency among the concurrent representations of jointly-executing models is our approach for effective MRM.
We developed a framework, UNIFY, to achieve effective MRM. UNIFY satisfies three MRM requirements: multi-representation interaction, multi-representation consistency and cost-effectiveness. It enables designers to construct solutions for application-specific multiple models. UNIFY is based on four fundamental observations that reduce the problem of joint execution to the problem of maintaining consistency among the representations of multiple models when dependent concurrent interactions occur. UNIFY consists of processes and techniques such as Multiple Representation Entities (MREs), Attribute Dependency Graphs (ADGs) and a taxonomy of interactions. An MRE maintains concurrent representations. An ADG captures relationships among attributes in concurrent representations. An ADG and application-specific mapping functions that translate attributes across representations constitute a Consistency Enforcer that maintains internal consistency within an MRE. Our taxonomy of interactions provides a way to classify interactions based on their semantic characteristics. This classification presents policies that can be encoded in an Interaction Resolver for resolving the effects of dependent concurrent interactions on an MRE.
UNIFY contributes to the practice of modeling and simulation. We show how designers can apply techniques in UNIFY. We present guidelines for maintaining consistency among concurrent representations. UNIFY is the first known general framework for achieving effective MRM.
PHD (Doctor of Philosophy)
Our thesis is that MRM can be effective. Maintaining consistency among the concurrent representations of jointly-executing models is our approach for effective MRM.
All rights reserved (no additional license for public reuse)