Distributed truth maintenance

Abstract

Distributed AI systems are intended to fill the gap between classical AI and distributed computer science. Such networks of different problem solvers are required for naturally distributed problems, and for tasks which exhaust the resource of an individual node. To guarantee a certain degree of consistency in a distributed AI system, it is necessary to inspect the beliefs of both single nodes and the whole net. This task is performed by Distributed Truth Maintenance Systems. Based on classical TMS theories, distributed truth maintenance extends the conventional case to incorporate maintenance in DAI scenarios.