The Design and Evaluation of Distributed Graph-Theoretic Heuristics for Computing Pairwise Interactions of Moving Entities
In recent years the focus of training simulations, particularly those in the defense domain, has been on increasing the number of simulated entities. This focus, coupled with the requirement for geographically distributed training sites, has led to increased interest in distributed simulation of the battlefield environment. The large efforts involved in the development of the DIS (Distributed Interactive Simulation) protocol and the HLA (High Level Architecture) have attempted to address these DoD (Department of Defense) requirements in several ways. The cost of con1puting the interactions between pairs of simulated entities frequently dominates system execution time. The difficulty of evaluating these interactions efficiently has been a major impediment to the development of the causality-preserving large-scale human-in-the-loop simulations which the defense training community desires. Many of the problems which arise in the computation of entity interactions are intractable, and admit no reasonably-efficient optimal solution. This intractability suggests that it is reasonable to seek heuristic solutions to these problems which provide good efficiency for most problem instances. The research described in this report attempted to find distributed graph-theoretic heuristics which would offer iinprove1nents in efficiency by simply making judicious mappings of simulated entities to simulation nodes.