MARLlib: A Scalable and Efficient Multi-agent Reinforcement Learning Library
This provides a practical tool for MARL researchers to streamline development and avoid compatibility issues, though it is incremental as it builds on existing MARL concepts.
The paper tackles the challenge of lacking a scalable and efficient library for multi-agent reinforcement learning (MARL) by introducing MARLlib, which uses standardized wrappers, agent-level algorithms, and flexible policy mapping to enable fast and compatible development, with the library publicly available on GitHub.
A significant challenge facing researchers in the area of multi-agent reinforcement learning (MARL) pertains to the identification of a library that can offer fast and compatible development for multi-agent tasks and algorithm combinations, while obviating the need to consider compatibility issues. In this paper, we present MARLlib, a library designed to address the aforementioned challenge by leveraging three key mechanisms: 1) a standardized multi-agent environment wrapper, 2) an agent-level algorithm implementation, and 3) a flexible policy mapping strategy. By utilizing these mechanisms, MARLlib can effectively disentangle the intertwined nature of the multi-agent task and the learning process of the algorithm, with the ability to automatically alter the training strategy based on the current task's attributes. The MARLlib library's source code is publicly accessible on GitHub: \url{https://github.com/Replicable-MARL/MARLlib}.