On the Importance of Exploration for Generalization in Reinforcement Learning

TL;DR

This paper emphasizes the critical role of exploration strategies in enhancing generalization in deep reinforcement learning, introducing a novel method called EDE that leverages distributional ensembles to improve performance on high-dimensional benchmarks.

Contribution

The paper introduces EDE, a value-based exploration method using distributional ensembles, demonstrating state-of-the-art results in RL generalization benchmarks.

Findings

Exploration improves generalization to unseen environments.

EDE outperforms existing methods on Procgen and Crafter benchmarks.

Ensemble-based exploration effectively captures epistemic uncertainty.

Abstract

Existing approaches for improving generalization in deep reinforcement learning (RL) have mostly focused on representation learning, neglecting RL-specific aspects such as exploration. We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments. Through a series of experiments in a tabular contextual MDP, we show that exploration is helpful not only for efficiently finding the optimal policy for the training environments but also for acquiring knowledge that helps decision making in unseen environments. Based on these observations, we propose EDE: Exploration via Distributional Ensemble, a method that encourages exploration of states with high epistemic uncertainty through an ensemble of Q-value distributions. Our algorithm is the first value-based approach to achieve state-of-the-art on both Procgen and Crafter, two benchmarks for generalization in RL with high-dimensional observations. The open-sourced implementation can be found at https://github.com/facebookresearch/ede .