DEAR: Disentangled Environment and Agent Representations for Reinforcement Learning without Reconstruction

TL;DR

DEAR introduces a novel disentangled representation learning approach for visual reinforcement learning that leverages agent shape information without reconstruction, significantly improving sample efficiency and robustness.

Contribution

The paper proposes DEAR, a method that uses segmentation masks for disentangled environment and agent representations, enhancing visual RL without requiring reconstruction.

Findings

DEAR outperforms state-of-the-art methods in sample efficiency.

DEAR achieves comparable or better performance with fewer parameters.

Disentangled representations improve learning robustness.

Abstract

Reinforcement Learning (RL) algorithms can learn robotic control tasks from visual observations, but they often require a large amount of data, especially when the visual scene is complex and unstructured. In this paper, we explore how the agent's knowledge of its shape can improve the sample efficiency of visual RL methods. We propose a novel method, Disentangled Environment and Agent Representations (DEAR), that uses the segmentation mask of the agent as supervision to learn disentangled representations of the environment and the agent through feature separation constraints. Unlike previous approaches, DEAR does not require reconstruction of visual observations. These representations are then used as an auxiliary loss to the RL objective, encouraging the agent to focus on the relevant features of the environment. We evaluate DEAR on two challenging benchmarks: Distracting DeepMind control suite and Franka Kitchen manipulation tasks. Our findings demonstrate that DEAR surpasses state-of-the-art methods in sample efficiency, achieving comparable or superior performance with reduced parameters. Our results indicate that integrating agent knowledge into visual RL methods has the potential to enhance their learning efficiency and robustness.