RL-X: A Deep Reinforcement Learning Library (not only) for RoboCup
This work provides a faster and more extensible tool for researchers and practitioners in reinforcement learning, though it is incremental as it builds on existing methods with performance optimizations.
The authors introduced RL-X, a deep reinforcement learning library that offers flexible, JAX-based implementations achieving up to 4.5x speed improvements over existing frameworks like Stable-Baselines3, and demonstrated its application in RoboCup Soccer Simulation and standard benchmarks.
This paper presents the new Deep Reinforcement Learning (DRL) library RL-X and its application to the RoboCup Soccer Simulation 3D League and classic DRL benchmarks. RL-X provides a flexible and easy-to-extend codebase with self-contained single directory algorithms. Through the fast JAX-based implementations, RL-X can reach up to 4.5x speedups compared to well-known frameworks like Stable-Baselines3.