Centralized control for multi-agent RL in a complex Real-Time-Strategy game
This is an incremental application of existing MARL methods to a new, challenging domain (real-time strategy games) for AI research and competition participants.
The paper tackled the challenge of applying multi-agent reinforcement learning (MARL) to the complex Lux AI v2 real-time strategy game, using a centralized training approach and reporting design decisions, but did not include concrete performance numbers or results.
Multi-agent Reinforcement learning (MARL) studies the behaviour of multiple learning agents that coexist in a shared environment. MARL is more challenging than single-agent RL because it involves more complex learning dynamics: the observations and rewards of each agent are functions of all other agents. In the context of MARL, Real-Time Strategy (RTS) games represent very challenging environments where multiple players interact simultaneously and control many units of different natures all at once. In fact, RTS games are so challenging for the current RL methods, that just being able to tackle them with RL is interesting. This project provides the end-to-end experience of applying RL in the Lux AI v2 Kaggle competition, where competitors design agents to control variable-sized fleets of units and tackle a multi-variable optimization, resource gathering, and allocation problem in a 1v1 scenario against other competitors. We use a centralized approach for training the RL agents, and report multiple design decisions along the process. We provide the source code of the project: https://github.com/roger-creus/centralized-control-lux.