Learning Generalizable Tool Use with Non-rigid Grasp-pose Registration

TL;DR

This paper introduces a novel reinforcement learning approach that enables robots to learn and generalize tool use behaviors from a single demonstration, effectively handling complex contacts and high-dimensional actions.

Contribution

It presents a new method for generalizing grasp configurations to unseen objects, improving the scalability and adaptability of robotic tool use learning.

Findings

Policies solve complex tool use tasks

Generalizes to unseen tools

Uses only a single demonstration

Abstract

Tool use, a hallmark feature of human intelligence, remains a challenging problem in robotics due the complex contacts and high-dimensional action space. In this work, we present a novel method to enable reinforcement learning of tool use behaviors. Our approach provides a scalable way to learn the operation of tools in a new category using only a single demonstration. To this end, we propose a new method for generalizing grasping configurations of multi-fingered robotic hands to novel objects. This is used to guide the policy search via favorable initializations and a shaped reward signal. The learned policies solve complex tool use tasks and generalize to unseen tools at test time. Visualizations and videos of the trained policies are available at https://maltemosbach.github.io/generalizable_tool_use.