Episodic-Like Memory for Cognitive Robots

Abstract

The article presents an approach to providing a cognitive robot with a long-term memory of experiences-a memory, inspired by the concept of episodic memory (in humans) or episodic-like memory (in animals), respectively. The memory provides means to store experiences, integrate them into more abstract constructs, and recall such content. The paper presents an analysis of key characteristics of natural episodic memory systems. Based on this analysis, conceptual and technical requirements for an episodic-like memory for cognitive robots are specified. The paper provides a formal design that meets these requirements, and discusses its full implementation in a cognitive architecture for mobile robots. It reports results of simulation experiments which show that the approach can run efficiently in robot applications involving several hours of experience.