Situated agents can have goals

  • Authors:

  • Pattie Maes

  • Affiliations:

  • AI-Laboratory, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium and AI-Laboratory, Massachusetts Institute of Technology, 545 Technology Square, Cambridge, MA 02139, USA

  • Venue:
  • Robotics and Autonomous Systems
  • Year:
  • 1990

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper discusses the problem of action selection for an autonomous agent. We argue that the so-called ''situated agents'', which have been built in response to the limitations observed with classical planners, do not provide a satisfactory solution to this problem because of their lack of goals and run-time arbitration. We present a novel action selection theory which allows arbitration among goals and actions while producing fast and robust activity in a tight interaction loop with the environment. The theory models action slection as an emergent property of an activation/inhibition dynamics among the actions the agent can select and between the actions and the environment. A handful of global parameters make it possible to smoothly mediate between several action selection criteria. For example, one can balance goal-orientedness against situation-orientedness, bias towards ongoing plans (inertia) against adaptivity, thoughtfulness against speed, and adjust sensitivity to goal conflicts.