Accelerating Deep Reinforcement Learning via Phase-Level Parallelism for Robotics Applications | IEEE Journals & Magazine | IEEE Xplore