This repository is a curated, continually updated list of research papers and resources on Exploration Methods in Reinforcement Learning (ERL). It aims to provide a comprehensive overview of the field for researchers and practitioners, categorizing ERL techniques and linking to seminal and recent publications.

How It Works

The repository categorizes exploration methods into two main phases: "Augmented Collecting Strategy" (applied during experience collection) and "Augmented Training Strategy" (applied during policy updates). Within these, methods are further classified into sub-categories like Action Selection Guidance, Count-Based, Information Theory Based, and Goal-Based exploration. This taxonomy helps users navigate the diverse landscape of ERL techniques.

Quick Start & Requirements

This is a curated list of papers; there are no installation or execution requirements. Links to papers, code (where available), and relevant environments are provided.

Highlighted Details

• Comprehensive taxonomy of exploration RL methods.
• Categorization of papers by conference year (ICLR, NeurIPS, ICML) from 2021 to 2025.
• Includes classic exploration RL papers and links to seminal works like Go-Explore, NoisyNet, and RND.
• Provides key insights and experimental environments for each listed paper.

Maintenance & Community

The repository is actively maintained and updated, welcoming contributions from the community.

Licensing & Compatibility

Awesome Exploration RL is released under the Apache 2.0 license.

Limitations & Caveats

This resource is a curated list of papers and does not provide implementations or code. The "continually updated" nature means the scope and depth may evolve.