Peg-in-Hole Assembly: A Challenge for Robot Learning

In recent years we have seen rapid advancements in the development of robot learning in hand grasping and even fine manipulation. Intelligent robotic manipulation requires and leverages a robot hand to achieve autonomous and rich physical interactions with the surrounding environment. Yet, robotic manipulation is not just about arm and hand motions. Contact is the true heart of manipulation. Precision peg-in-hole assembly is such a challenging problem in robotic learning of fine manipulation. The problem gets at the very heart of manipulation, providing insights that apply across almost all manipulation tasks, whether by human or robot. This presentation will give an overview of the trend in the development of robotic learning with dexterous and versatile capability for grasping and adaptive manipulation. In this perspective, we argue that the multi-modal manipulation capabilities of human-centered robots are essential for a feasible large-scale deployment of the future robotic technology with a potential of commercial success.