Autonomous Underwater Vehicle for Pipeline Following Using YOLOv8 Instance Segmentation and Robotic Operating System

Abstract

The autonomous movement of an Autonomous Underwater Vehicle (AUV) is essential for underwater pipeline inspection to ease the operator's task. However, current technology does not yet allow AUVs to autonomously achieve precise navigation for tracking underwater pipelines. In this paper, we develop an AUV capable of following an underwater pipeline using YOLOv8 Instance Segmentation and Robot Operating System (ROS). The video captured by the camera is transmitted to a ground station and processed using YOLOv8 Instance Segmentation. A mini-PC is employed for this process. The AUV's communication and movement are managed through ROS as the main library. The AUV was successfully developed and tested, demonstrating the ability to follow an underwater pipeline with a Mean Absolute Error (MAE) of 30.4% at the upper point detection with a standard deviation of 125 pixels, and an MAE of 12.5% at the lower point detection with a standard deviation of 79 pixels.