One-Shot Visual Imitation Learning via Meta-Learning

TL;DR

This paper introduces a meta-imitation learning approach that allows robots to quickly learn new skills from a single visual demonstration, scaling to raw pixel inputs and requiring fewer prior tasks.

Contribution

The method enables efficient one-shot learning of new skills directly from raw pixel data, improving scalability and data efficiency over previous approaches.

Findings

Successfully learned new tasks from a single demonstration

Operated effectively on both simulated and real robots

Reduced data requirements for effective skill acquisition

Abstract

In order for a robot to be a generalist that can perform a wide range of jobs, it must be able to acquire a wide variety of skills quickly and efficiently in complex unstructured environments. High-capacity models such as deep neural networks can enable a robot to represent complex skills, but learning each skill from scratch then becomes infeasible. In this work, we present a meta-imitation learning method that enables a robot to learn how to learn more efficiently, allowing it to acquire new skills from just a single demonstration. Unlike prior methods for one-shot imitation, our method can scale to raw pixel inputs and requires data from significantly fewer prior tasks for effective learning of new skills. Our experiments on both simulated and real robot platforms demonstrate the ability to learn new tasks, end-to-end, from a single visual demonstration.