An Industrial Perspective on Robot Learning

Challenges and Opportunities

Abstract

Robot learning has been a very active field of research recently with impressive scientific results especially in the application of reinforcement learning to robotic tasks. From an industrial perspective, the rapid scientific developments promise reduced commissioning times, simplified programming, increased productivity, and cost reductions. We attempt to put the potential of this promising but sometimes overhyped technology in perspective by matching it to industrial requirements. Key challenges are presented and possible ways to tackle them are discussed.