EPIROME - A novel framework to investigate high-level episodic robot memory

Abstract

Episodic memory has been examined in different disciplines such as psychology and neuroscience for more than 30 years. Now, engineering and computer science are developing an increasing interest in episodic memory for artificial systems. In this paper, we propose a novel framework referred to as EPIROME to develop and investigate high-level episodic memory mechanisms which can be used to model and compare episodic memories of high-level events for technical systems. We applied the framework in the domain of service robotics to enable our service robot TASER to collect autobiographical memories to improve action planning based on past experiences. The framework provides the robot with a life-long memory since past experiences can be stored and reloaded. In practise, one main advantage of our episodic memory is that it provides oneshot learning capabilities to our robot. This reduces the demerit of other learning strategies where learning takes too long when used with a real robot system in natural environments and therefore is not feasible.