On Improving Deep Reinforcement Learning for POMDPs

TL;DR

This paper introduces ADRQN, a novel deep RL architecture that combines action-observation pairs with LSTM to improve learning in partially observable environments, demonstrated on flickering Atari games.

Contribution

The paper proposes ADRQN, integrating action-specific encoding with recurrent neural networks to enhance deep RL performance in POMDPs, addressing a gap in existing methods.

Findings

ADRQN outperforms traditional DQNs in partially observable Atari games.

Action-observation pairing with LSTM improves latent state learning.

Enhanced deep RL performance in POMDPs demonstrated.

Abstract

Deep Reinforcement Learning (RL) recently emerged as one of the most competitive approaches for learning in sequential decision making problems with fully observable environments, e.g., computer Go. However, very little work has been done in deep RL to handle partially observable environments. We propose a new architecture called Action-specific Deep Recurrent Q-Network (ADRQN) to enhance learning performance in partially observable domains. Actions are encoded by a fully connected layer and coupled with a convolutional observation to form an action-observation pair. The time series of action-observation pairs are then integrated by an LSTM layer that learns latent states based on which a fully connected layer computes Q-values as in conventional Deep Q-Networks (DQNs). We demonstrate the effectiveness of our new architecture in several partially observable domains, including flickering Atari games.