Developing an efficient landmark for autonomous docking tasks of underwater robots

This paper proposes an unobtrusive and flexible artificial landmark for the underwater robot homing problem. We designed modified Self-Similar Landmark (SSL) which reveals two different kinds of features: 1) a self-similar feature at long distances and 2) corner features at short distances from the target. That is, the self similarity of the landmark attracts a robot until it approaches close to the landmark. When the robot approaches to the dock and the corner features of the target board become conspicous, the proposed framework starts estimating 3D poses of the robot with respect to the target. Thus, proposed framework extracts currently available information from the target board and adaptively uses it in order to find an optimal docking trajectory. Our method has been tested in the underwater environment, and we have observed a greate possibility of the proposed landmark as a passive docking target for underwater robots.