COOPERATION UNDER UNCERTAINTY IN DISTRIBUTED EXPERT SYSTEMS

Just as cooperation between human experts is important when solving complex problems, so too is cooperation between computerized expert systems in a distributed expert system. Four kinds of cooperation are classified in this paper according to their inter-dependence relationships which are horizontal cooperation, tree cooperation, recursive cooperation, and hybrid cooperation. If individual expert systems in a distributed expert system use different inexact reasoning models, it will be necessary to transform the uncertainties of propositions from one model to another when the expert systems cooperate. The algebraic structures of inexact reasoning models are recognized to be semigroups with unit elements, and homomorphic transformations of the uncertainties of propositions among different inexact reasoning models are realized. The synthesis of solutions with uncertainties is a key problem for cooperation among expert systems in a distributed expert system. A scheme is presented in this paper to synthesize these different solutions, where the final decision is based on the mean values and the uniformity of the uncertainty values. Cooperation under uncertainty in distributed expert systems is necessary in order to solve the real problems because uncertainty exists everywhere in the real world. This paper has solved the part of this problem dealing with heterogeneous transformation and synthesis of uncertainties.