Utilizing theory of mind for action selection applied in the domain of fighter pilot training

  • Authors:
  • Mark Hoogendoorn;Robbert-Jan Merk

  • Affiliations:
  • Department of Computer Science, VU University Amsterdam, Amsterdam, The Netherlands 1081 HV;Training, Simulation, and Operator Performance, National Aerospace Laboratory, Amsterdam, The Netherlands

  • Venue:
  • Applied Intelligence
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

When developing intelligent agents, approaches that allow the anticipation of other agents is of utmost importance. For humans, this has also been shown to be crucial to establish good interactions. In this paper, a design for an agent that is equipped with theory of mind based reasoning capabilities is presented. The approach moves beyond the state of the art from several angles: first, it allows for the expression of certainties with respect to the predicted states of the other agents. Second, it allows the prediction during a substantial number of time steps in the future, thereby utilizing the theory of mind model multiple times. The approach has been applied to the domain of fighter pilots whereby intelligent opponents are developed to facilitate dedicated training for F16 fighter pilots.