The University of Sussex

Goal detection: representation, mechanisms, architecture

Remedios de Dios Bulos

Today there appears to be considerable research enthusiasm towards endowing an autonomous intelligent agent with the capability of forumulating its own goals. Goal Formulation refers to the intelligent behavior that an agent exhibits when reasoning (and deciding) what goals to pursue and when to pursue them. It may be described as the integration of several processes, namely Goal Detection, Goal Assessment, Goal Prioritizing, Goal Evaluation and Goal Modification. However this paper concentrates only on goal detection. It seeks to identify the ways by which an autonomous intelligent agent can detect its goals. Goal detection is the process of identifying the goals that an agent needs to accomplish to affect the external world. It is a process of continuous reasoning about the agent's changing beliefs about the world, motives, desires and intentions. This paper identifies four goal detection mechanisms, namely: current data-driven (bottom-up), goal-driven (top-down), communicated (prescribed by an external agent), and forecasted data-driven. These machanisms are parallel and are continuously and simultaneously executed by the agent. To prove the applicability of goal detection concepts elaborated in this work, a simulated household robot that detects its goals is presented at the end.

Download compressed postscript file