ABSTRACT
This chapter proposes a novel hierarchical framework to achieve real-time path planning and following for a gliding underwater robot, including a learning-based path planner and an adaptive following controller. Subsequent to considering higher intelligence in an unknown ocean environment, a novel hierarchical deep Q-network method is presented to separately plan the collision avoidance path and the approach path and design different continuous states under the kinematic constraints. Next, an improved line-of-sight method is adopted to obtain the desired points from planned path. More important, a nonlinear control law based on the backstepping technique is derived, especially avoiding singularities in the law derivation using a 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.
