Acquisition of optical flow considering the pitching motion to estimate the self-position of a manta robot

The development of underwater exploration robots is expected for the biological research of aquatic life. In particular, biomimetic robots are possible to investigate aquatic organisms without affecting them seriously. A Manta robot has been already developed by mimicking the pectoral fin of the manta ray. Note however that it is difficult to obtain the self-position data, because small AUVs like a Manta robot can't communicate underwater with radio waves. As one approach to obtain the self-position data, a method of using optical flow is constructed in this paper. The amount of movement of the robot is estimated by using the optical flow obtained from a monocular camera, when the robot moves around an underwater structure or the bottom of the sea. Furthermore, after evaluating the influence of pitch motion of a Manta robot on the detection of an optical flow, an optical flow acquisition method is proposed using a noise removal filter and a gimbal mechanism to reduce such an influence on the detection. Finally, the effectiveness of the proposed method is confirmed by actual experiments.