Three-Dimensional Path Following Control of an Underactuated Robotic Dolphin Using Deep Reinforcement Learning | IEEE Conference Publication | IEEE Xplore