The property of autonomy is a crucial requirement for intelligent agents. Several definitions of autonomy, with significantly different meanings, are available in the literature. In particular a detailed study about this concept, featuring the distinction between cognitive and social autonomy, has been carried out in [Castelfranchi 95]. With reference to these works, we discuss in this paper how the satisfaction of autonomy requirements can be achieved by a novel approach to agent modeling, based on the concept of active mental entity. After introducing the main features of the approach, we show how it is able to guarantee both social and cognitive autonomy. An application example concerning action planning of an autonomous mobile robot provides a concrete illustration of the proposed approach.
PDF file