Terrain-adaptive optimal guidance for near-bottom survey by an autonomous underwater vehicle

An optimal guidance strategy for an autonomous underwater vehicle dive specified to a near-bottom survey over an uneven seafloor topography is presented. Major objective of the optimal guidance is letting the longitudinal vehicle attitude follow the along-track gradient of the seafloor throughout the near-bottom survey dive. The proposed method derives depth sequence corresponding to waypoints predefined in a horizontal plane, combined set of which defines the 3-dimentional trackline for the optimal near-bottom survey. A supervised learning approach based on gradient descent search algorithm is employed in deriving the optimal depth sequence. Iterative simulations of vehicle's dives driven by the learning algorithm subject to vehicle dynamics and interacting with given seabed topography derive the optimal depth sequence. The bottom gradient-following is an important issue in achieving higher quality sonograph and precise echo-sounding, as well as keeping the reliability of altitudes obtained by an acoustic altimeter. Proposed strategy is applied to near-bottom survey dives performed with a cruising AUV Aqua-Explorer 2000a. Based on the body structure of Aqua-Explorer 2000 originally developed by KDDI for inspecting and monitoring the undersea cables, Aqua-Explorer 2000a was born via full renovation conducted by Institute of Industrial Science, The University of Tokyo.