Using Fuzzy Logic as a Reasoning Model for BDI Agents

To model human reasoning with uncertainty and imprecision, fuzzy logic has been employed to reason for BDI agents in this paper. Due to the important role of beliefs in reasoning of human or BDI Agents, formal semantics of fuzzy-based belief representation are defined including basic semantics of beliefs whose members are quantitative or quantitative variables. And two different means are applied to calculate the membership functions of the new beliefs which are derived by one rule or multiple rules arose from the process of reasoning, respectively. For further deep deliberation, OR and NOT relations among different kinds of beliefs are converted to AND relations. More specifically, by assigning weights for different beliefs which have different importance with respect to a given goal, an effective way is provided to combine different beliefs with AND relations. To the end, fuzzy-based reasoning is applied in a simple scenario. The result demonstrates that the fuzzy-based reasoning is more flexible and human-like compared to non-fuzzy based BDI agents reasoning in the real world applications. Particularly, it provides a more rational action selection mechanism for a given goal. The work in this paper causes a distinct improvement in reasoning capability of the previous BDI Agents.