Weakly Supervised Disentangled Representation for Goal-conditioned Reinforcement Learning

TL;DR

This paper introduces DR-GRL, a framework that combines disentangled representation learning with goal-conditioned reinforcement learning to improve sample efficiency and policy generalization, especially in visual environments.

Contribution

It proposes a weakly supervised Spatial Transform AutoEncoder (STAE) for interpretable, controllable representations that facilitate goal generation and enhance reinforcement learning performance.

Findings

DR-GRL outperforms previous methods in sample efficiency.

The approach improves policy generalization in goal-conditioned tasks.

It is adaptable to real robot applications.

Abstract

Goal-conditioned reinforcement learning is a crucial yet challenging algorithm which enables agents to achieve multiple user-specified goals when learning a set of skills in a dynamic environment. However, it typically requires millions of the environmental interactions explored by agents, which is sample-inefficient. In the paper, we propose a skill learning framework DR-GRL that aims to improve the sample efficiency and policy generalization by combining the Disentangled Representation learning and Goal-conditioned visual Reinforcement Learning. In a weakly supervised manner, we propose a Spatial Transform AutoEncoder (STAE) to learn an interpretable and controllable representation in which different parts correspond to different object attributes (shape, color, position). Due to the high controllability of the representations, STAE can simply recombine and recode the representations to generate unseen goals for agents to practice themselves. The manifold structure of the learned representation maintains consistency with the physical position, which is beneficial for reward calculation. We empirically demonstrate that DR-GRL significantly outperforms the previous methods in sample efficiency and policy generalization. In addition, DR-GRL is also easy to expand to the real robot.