EFFECTIVE GRAPH REPRESENTATION SUPPORTING MULTI-AGENT DISTRIBUTED COMPUTING

The parallel processing is an effective approach to solving those high complexity problems which may be represented as a set of independent or loosely coupled subproblems. In the latter case, however, the critical factor for a computation time is an overhead generated by communication among particular subtasks. The decomposition of a graph-based computational problem allows transforming it into a set of subproblems to be processed in parallel. A decomposition method should guarantee a good performance of Parallel computations with respect to communication and synchronization among agents managing a distributed representation of a considered system. In this paper we present the novel method of a decomposition, reducing coupling among subproblems and thus minimizing a required cooperation among agents. Comparison and performance tests are also included.