The use of autonomous marine vehicles, and especially autonomous underwater vehicles, is rapidly increasing within several fields of study. In particular, such vehicles can be applied for sea floor mapping, oceanography, environmental monitoring, inspection and maintenance of underwater structures (for instance within the oil and gas industry) and military purposes. They are also highly suitable for operations below ice-covered areas in the Arctic. However, there are still many challenges related to making such underwater vehicles autonomous. A fundamental task of an autonomous underwater vehicle vessel is to follow a general path in the presence of unknown ocean currents. There exist several results for underwater vehicles to follow a general path when no ocean currents are present [1] and to follow a geometrically simple path such as a straight line when ocean currents affect the vehicle [2, 3], but the problem of general path following in the presence of unknown ocean currents has not been solved yet. This paper presents a method to achieve this. The results are an extension of the results in [1], and introduce a virtual Serret-Frenet reference frame that is anchored in and propagates along the desired path. The closed-loop system consists of an ocean current observer, a guidance law, a controller and an update law to drive the Serret-Frenet frame along the path, and is shown to be asymptotically stable given that certain assumptions are fulfilled. This guarantees that the autonomous underwater vehicle will converge to the desired path and move along it with the desired velocity. Simulation results are presented to verify and illustrate the theoretical results.
Deep Interactive Reinforcement Learning for Path Following of Autonomous Underwater Vehicle
