Learning Multi-Robot Decentralized Macro-Action-Based Policies via a Centralized Q-Net
This work addresses the problem of non-stationarity in decentralized multi-agent reinforcement learning for real-world robotics applications, though it appears incremental as it builds on existing Q-net and macro-action techniques.
The paper tackles the challenge of learning decentralized policies for multi-robot systems by proposing a macro-action-based method that uses a centralized Q-net for training, achieving near-centralized performance in simulation and enabling real robots to efficiently complete a warehouse delivery task.
In many real-world multi-robot tasks, high-quality solutions often require a team of robots to perform asynchronous actions under decentralized control. Decentralized multi-agent reinforcement learning methods have difficulty learning decentralized policies because of the environment appearing to be non-stationary due to other agents also learning at the same time. In this paper, we address this challenge by proposing a macro-action-based decentralized multi-agent double deep recurrent Q-net (MacDec-MADDRQN) which trains each decentralized Q-net using a centralized Q-net for action selection. A generalized version of MacDec-MADDRQN with two separate training environments, called Parallel-MacDec-MADDRQN, is also presented to leverage either centralized or decentralized exploration. The advantages and the practical nature of our methods are demonstrated by achieving near-centralized results in simulation and having real robots accomplish a warehouse tool delivery task in an efficient way.