Intrinsic Motivation and Automatic Curricula via Asymmetric Self-Play

TL;DR

This paper introduces an unsupervised learning scheme using asymmetric self-play where two agents generate a curriculum of tasks, improving reinforcement learning efficiency and outcomes in reversible or resettable environments.

Contribution

The work presents a novel asymmetric self-play method that automatically creates curricula for unsupervised exploration, enhancing RL training efficiency and effectiveness.

Findings

Reduces supervised episodes needed for RL training.

Achieves higher rewards in certain environments.

Automatically generates exploration curricula.

Abstract

We describe a simple scheme that allows an agent to learn about its environment in an unsupervised manner. Our scheme pits two versions of the same agent, Alice and Bob, against one another. Alice proposes a task for Bob to complete; and then Bob attempts to complete the task. In this work we will focus on two kinds of environments: (nearly) reversible environments and environments that can be reset. Alice will "propose" the task by doing a sequence of actions and then Bob must undo or repeat them, respectively. Via an appropriate reward structure, Alice and Bob automatically generate a curriculum of exploration, enabling unsupervised training of the agent. When Bob is deployed on an RL task within the environment, this unsupervised training reduces the number of supervised episodes needed to learn, and in some cases converges to a higher reward.