Image-based navigation of Small-size Autonomous Underwater Vehicle "Kyubic" in International Underwater Robot Competition

Abstract

An International underwater robot competition, "RoboSub" is held in USA to demonstrate robot's autonomy by completing underwater tasks, with a new theme each year. Student project team, "Kyutech Underwater robotics" developed autonomous underwater vehicle (AUV) "KYUBIC" with built-in image processing module for the RoboSub. Sensors on the AUV are connected via ethernet and its automatic navigation program is built using ROS. The AUV moves through gate to the target objects based on self-localization using doppler velocity log and IMU. When approaching the object, the AUV aligns its direction to the object based on the red line detection by image processing as well as the type, position, and altitude of the target objects are detected by deep learning on the image processing module.