In practical applications, intelligent agents are designed to be capable of reasoning, making decision under uncertainty. We shall propose in this paper a model of intelligent agents that are capable to reason with uncertain knowledge which are expressed in the form of interval-valued probabilistic F-rules. Knowledge-based agents are considered as F-rule systems, having a F-rule base together with suitable reasoning operators. We proposed two types of reasoning operators: Global operator that uses the whole F-rule base and Local one, using only a subset of F-rules. It is important to verify whether a F-rule agent is consistent and stationary. Based on features of the F-rule base (characteristics of interval functions used in rules, interaction of rules), we can prove the equivalence of reasoning capacity of global and local operators in monotonic F-rule systems, and that in strongly monotonic systems, the terminate problem is decidable.
展开▼
