Efficient Multi-Task and Transfer Reinforcement Learning with Parameter-Compositional Framework

In this work, we investigate the potential of im- proving multi-task training and also leveraging it for transfer- ring in the reinforcement learning setting. We identify several challenges towards this goal and propose a transferring ap- proach with a parameter-compositional formulation. We inves- tigate ways to improve the training of multi-task reinforcement learning which serves as the foundation for transferring. Then we conduct a number of transferring experiments on various manipulation tasks. Experimental results demonstrate that the proposed approach can have improved performance in the multi-task training stage, and further show effective transfer- ring in terms of both sample efficiency and performance.