Human-to-Robot Imitation in the Wild
Shikhar Bahl Abhinav Gupta⋆ Deepak Pathak⋆ Carnegie Mellon University
Abstract
We approach the problem of learning by watching humans in the wild. While traditional approaches in Imitation and Reinforcement Learning are promising for learning in the real world, they are either sample inefficient or are constrained to lab settings. Meanwhile, there has been a lot of success in processing passive, unstructured human data. We propose tackling this problem via an efficient one-shot robot learning algorithm, centered around learning from a third person perspective. We call our method WHIRL: In-the-Wild Human Imitating Robot Learning. WHIRL extracts a prior over the intent of the human demonstrator, using it to initialize our agent’s policy. We introduce an efficient real-world policy learning scheme that improves using interactions. Our key contributions are a simple sampling-based policy optimization approach, a novel objective function for aligning human and robot videos as well as an exploration method to boost sample efficiency. We show one-shot generalization and success in real world settings, including 20 different manipulation tasks in the wild. Videos at https://human2robot.github.io.
中文速览
机器人要在实验室外的真实世界中学会操作物体，却长期面临两大难题：强化学习太费样本、模仿学习又需要大量人工示教。WHIRL（野外人类模仿机器人学习）提出一种新思路：让机器人通过观看普通的人类操作视频（如YouTube上的日常动作）来获取"意图先验"，再利用采样优化和一种新颖的人机视频对齐目标函数，在真实环境中用极少的交互次数迭代改进策略。为了不被先验限制，系统还引入了一种与任务无关的探索机制来提升样本效率。实验在3个真实环境中验证了20项不同的操作任务（包括开关门、折叠衣物、清洁白板等），实现了"单次示教即可泛化"的效果，是迄今为止将机器人操作从实验室推向真实世界最大规模的尝试之一。
原文 arXiv:2207.09450;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2207.09450v1