RL-Glue: Language-Independent Software for Reinforcement-Learning Experiments. RL-Glue is a standard, language-independent software package for reinforcement-learning experiments. The standardization provided by RL-Glue facilitates code sharing and collaboration. Code sharing reduces the need to re-engineer tasks and experimental apparatus, both common barriers to comparatively evaluating new ideas in the context of the literature. Our software features a minimalist interface and works with several languages and computing platforms. RL-Glue compatibility can be extended to any programming language that supports network socket communication. RL-Glue has been used to teach classes, to run international competitions, and is currently used by several other open-source software and hardware projects.

This software is also peer reviewed by journal TOMS.