Resolving Conflicts in Distributed Diagnosis

P. Fröhlich, and W. Nejdl

Abstract

Current research in the area of diagnosis with multiple models focuses on a single central diagnostic agent using a hierarchy of system models with different abstraction levels. In this approach the system models must have certain strong properties, which can only be assured, if all the models are created by the same group of experts. This assumption is a severe limitation in complex systems containing devices from several companies, as for example in communication networks. The current paper therefore discusses the use of a set of heterogeneous agents in diagnosis, the notion of conflict arising in this environment and possible solutions to resolve these conflicts. We identify two forms of distribution and argue that each has its specific definition of a conflict.

Keywords: Diagnosis, Distributed AI

The full report is available as a postscript file .