Enhancing intelligent agents with episodic memory

  • Authors:
  • John E. Laird;Andrew M. Nuxoll

  • Affiliations:
  • University of Michigan;University of Michigan

  • Venue:
  • Enhancing intelligent agents with episodic memory
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this dissertation, we explore the effects of adding an episodic memory to an intelligent agent. First, we define the design space for episodic memory systems and the properties that any implementation must have in order to be integrated into a cognitive architecture. We then describe our exploration of this space including two major implementations of an architectural episodic memory as well as several refinements to those implementations and their impact on agent performance. We also present a series of cognitive capabilities that are facilitated by virtue of an agent possessing an episodic memory. We hypothesize that these capabilities improve an agent's ability to effectively sense its environment, reason and learn. We then demonstrate five of these cognitive capabilities using a specific task in one of two different virtual environments.