BDI MULTIAGENT LEARNING BASED ON FIRST-ORDER INDUCTION OF LOGICAL DECISION TREES

摘要

This paper is about learning in the context of Multiagent Systems (MAS) composed by intentional agents, e.g. agents that behave based on their beliefs, desires, and intentions (BDI). We assume that MAS learning differs in subtle ways from the general problem of learning, as defined traditionally in Machine Learning (ML). We explain how BDI agents can deal with these differences and introduce the application of first-order induction of logical decision trees to learn in the BDI framework. We exemplify our approach learning the conditions in which plans can be executed by an agent.