This paper proposes a novel hierarchical framework to achieve real-time path planning and following for a gliding robotic dolphin, including a learning-based path planner and an adaptive following controller. Subsequent to considering higher intelligence in unknown ocean environment, we present a novel hierarchical deep Q-network method to separately plan the collision avoidance path and the approach path, and also design different continuous-states under the kinematic constraints. Next, we adopt an improved line-of-sight method to obtain the desired points from planned path. More importantly, we derive a nonlinear control law based on backstepping technique, and specially avoid singularities in the law derivation using barrier Lyapunov function. In particular, the unknown model parameters are adaptively calculated to make the controller more robust. Finally, extensive simulation and experimental results verify the effectiveness of the proposed methods, providing a new idea to further ocean exploration.