Learning When to Transfer among Agents: An Efficient Multiagent Transfer Learning Framework

Transfer Learning has shown great potential to enhance the single-agent Reinforcement Learning (RL) efficiency, by sharing previously learned policies. Inspired by this, the team learning performance in multiagent settings can be potentially promoted with agents reusing knowledge between each other when all agents interact with the environment and learn simultaneously. However, how each independent agent selectively learns from other agents' knowledge is still a problem. In this paper, we propose a novel multi-agent transfer learning framework to improve the learning efficiency of multiagent systems. Our framework learns when and what advice to give to each agent and when to terminate it by modeling multi-agent transfer as the option learning problem. We also propose a novel option learning algorithm, named as the Successor Representation Option (SRO) learning that decouples the dynamics of the environment from the rewards to learn the option-value function under each agent's preference. The proposed framework can be easily combined with existing deep RL approaches. Experimental results show it significantly accelerates the learning process and surpasses state-of-the-art deep RL methods in terms of learning efficiency and final performance in both discrete and continuous action spaces.