Test method for measuring station-keeping with unmanned marine vehicles using sonar or optical sensors

Abstract

This paper proposes a test method for measuring the ability of a USV or ROV to fixate on an underwater object, i.e., station-keeping. Station-keeping is needed to permit an operator or domain expert to stay focused on an area in an image long enough to identify objects, such as submerged cars and debris, or a condition, such as scour eroding the underwater footing of a bridge. This problem is different from traditional robot control, as the point is not to measure the positions of the robot and sensor payload but rather how well the system maintains the position of the object in the image. The test method uses the Lucas-Kanade optical flow algorithm in OpenCV to track an inexpensive raised plywood and wire fiducial. The rotational, translational, and root mean square (RMS) error is measured over a 3 minute period as well as number of image frames in which the fiducial was not visible. The method was demonstrated using a DIDSON acoustic camera, but is generalizable to other types of sonars and underwater video cameras.