A General-Purpose Method for Decision-Making in Autonomous Robots

In this paper, it is argued that the standard taxonomy of behavior selection is incomplete. In order to overcome the limitations of standard behavior selection, a novel method for decision-making, the extended utility function (EUF) method, has been developed. Based oil the concept of utility as a common currency for decision-making, the method handles decision-making involving both cognitive processes and (motor) behaviors, and is applicable as a general-purpose framework for decision-making in autonomous robots (as well as software agents). The EUF method is introduced and described, and it is then illustrated by means of an example. Preliminary tests indicate that the method performs well, allowing users rapidly to set up a decision-making system.