A. L. Thomaz, G. Hoffman, C. Breazeal

Abstract

It is our goal to understand the role real-time human interaction can play in machine learning algorithms for robots. In this paper we present Interactive Reinforcement Learning (IRL) as a plausible approach for training human-centric assistive robots by natural interaction. We describe an experimental platform to study IRL, pose questions arising from IRL, and discuss initial observations obtained during the development of our system.