Abstract

This paper reports lessons learned from using guided policy search to train human-robot handover behaviors with a collaborative robot, reflecting on the practical challenges of applying this reinforcement learning method to physical handover tasks.