A NOVEL STRUCTURE FOR REALIZING GOAL-DIRECTED BEHAVIOR


Yucelgen C., Kuyumcu Y., Şengör N. S.

4th International Conference on Agents and Artificial Intelligence, Portugal, 6 - 08 February 2012, pp.249-255 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.5220/0003719902490255
  • Country: Portugal
  • Page Numbers: pp.249-255
  • Istanbul Technical University Affiliated: Yes

Abstract

Intelligent organisms complete goal-directed behaviour by accomplishing a series of cognitive process. Inspired from these cognitive processes, M this work, a novel structure composed of Adaptive Resonance Theory and an Action Selection module is introduced. This novel structure is capable of recognizing task relevant patterns and choosing task relevant actions to complete goal-directed behavior. In order to construct these task relevant choices the parameters of the system are modified by Reinforcement Learning. Thus the proposed structure is capable of modifying its choices and evaluates the outcome of these choices. In order to show the efficiency of the proposed structure word hunting task is solved.