Authors: Nare Karapetyan, James Johnson, Ioannis Rekleitis,
Abstract: This work proposes vision-only navigation strategies for an autonomous underwater robot. Thisapproach is a step towards solving the coverage path planning problem in a 3D environment for surveyingunderwater structures. Given the challenging conditions of the underwater domain it is very complicatedto obtain accurate state estimates reliably. Consequently, it is a great challenge to extend known pathplanning or coverage techniques developed for aerial or ground robot controls. In this work we areinvestigating a navigation strategy utilizing only vision to assist in covering a complex underwaterstructure. We propose to use a navigation strategy akin to what a human diver will execute whencircumnavigating around a region of interest, in particular when collecting data from a shipwreck. Thefocus of this paper is a step towards enabling the autonomous operation of light-weight robots nearunderwater wrecks in order to collect data for creating photo-realistic maps and volumetric 3-D modelswhile at the same time avoiding collisions. The proposed method uses convolutional neural networksto learn the control commands based on the visual input. We have demonstrated feasibility of usinga system based only on vision to learn specific strategies of navigation with 80\% accuracy on theprediction of control commands changes. Experimental results and a detailed overview of the proposedmethod are discussed.