Exploring with Sticky Mittens: Reinforcement Learning with Expert Interventions via Option Templates

TL;DR

This paper introduces a framework that leverages expert interventions via option templates to improve reinforcement learning in long-horizon, sparse reward tasks, significantly outperforming existing methods.

Contribution

It proposes using expert intervention with option templates to guide reinforcement learning, enabling efficient learning of complex tasks with less resource expenditure.

Findings

Outperforms state-of-the-art methods by two orders of magnitude

Effective in three challenging reinforcement learning problems

Demonstrates the utility of expert interventions in long-horizon tasks

Abstract

Long horizon robot learning tasks with sparse rewards pose a significant challenge for current reinforcement learning algorithms. A key feature enabling humans to learn challenging control tasks is that they often receive expert intervention that enables them to understand the high-level structure of the task before mastering low-level control actions. We propose a framework for leveraging expert intervention to solve long-horizon reinforcement learning tasks. We consider \emph{option templates}, which are specifications encoding a potential option that can be trained using reinforcement learning. We formulate expert intervention as allowing the agent to execute option templates before learning an implementation. This enables them to use an option, before committing costly resources to learning it. We evaluate our approach on three challenging reinforcement learning problems, showing that it outperforms state-of-the-art approaches by two orders of magnitude. Videos of trained agents and our code can be found at: https://sites.google.com/view/stickymittens