Characterising agents' behaviours: Selecting goal strategies based on attributes

Abstract

The growth in the demand of autonomous agent systems which take decisions on behalf of other agents or human users, increases the necessity to study systems which use affective elements to manage their resources and to take decisions in order to become more efficient and to facilitate human-machine interaction. In this paper we present an architecture that allows an agent to select a sequence of actions based on a previously predefined planning structure, by using a tree of goals and a set of informational beliefs. The affective elements which we call attributes, such as urgency, insistence and intensity, have the capacity to alter the agents' behaviours, modifying their priorities with regard to resource consumption, the implicit costs of action execution and even their capabilities to execute an action. In a preliminary experiment made in a multi-agent system environment, a modified predator-prey workbench, we show how the attributes linked to these beliefs change the agents' behaviour and improve their global performance. © Springer-Verlag Berlin Heidelberg 2006.