An Approach for Creating Distributed Intelligent Systems

The paper presents an architecture of a distributed expert system for knowledge sharing across the network, which differs from traditional autonomous agents approach by separating knowledgebase and inference into different components possibly located on different network hosts. By combining inference engine and knowledge provider components in different ways it is possible to combine knowledge on rule-level (on the level of domain knowledge) or on the level of problem state representation. It also allows us to apply several inference strategies to one set of rules, or perform combined inference by using several inference engines in turn. The article describes domain knowledge and problem state representation most suitable for remote inference, as well as some technical aspects of implementation of such a system using CORBA as remote computation paradigm. Some examples demonstrating possible real-life usages of the remote inference are given.