Third-Person Imitation Learning

TL;DR

This paper introduces a novel unsupervised third-person imitation learning method enabling agents to learn from demonstrations observed from different viewpoints without explicit state correspondence, inspired by human learning.

Contribution

It presents a domain confusion-based approach for third-person imitation learning, addressing the challenge of learning from demonstrations with viewpoint differences without supervision.

Findings

Successful learning in pointmass, reacher, and inverted pendulum environments.

Demonstrates effectiveness of domain-agnostic features in third-person imitation.

Outperforms traditional methods requiring first-person demonstrations.

Abstract

Reinforcement learning (RL) makes it possible to train agents capable of achieving sophisticated goals in complex and uncertain environments. A key difficulty in reinforcement learning is specifying a reward function for the agent to optimize. Traditionally, imitation learning in RL has been used to overcome this problem. Unfortunately, hitherto imitation learning methods tend to require that demonstrations are supplied in the first-person: the agent is provided with a sequence of states and a specification of the actions that it should have taken. While powerful, this kind of imitation learning is limited by the relatively hard problem of collecting first-person demonstrations. Humans address this problem by learning from third-person demonstrations: they observe other humans perform tasks, infer the task, and accomplish the same task themselves. In this paper, we present a method for unsupervised third-person imitation learning. Here third-person refers to training an agent to correctly achieve a simple goal in a simple environment when it is provided a demonstration of a teacher achieving the same goal but from a different viewpoint; and unsupervised refers to the fact that the agent receives only these third-person demonstrations, and is not provided a correspondence between teacher states and student states. Our methods primary insight is that recent advances from domain confusion can be utilized to yield domain agnostic features which are crucial during the training process. To validate our approach, we report successful experiments on learning from third-person demonstrations in a pointmass domain, a reacher domain, and inverted pendulum.