Logo image
Back
Stable Wreckage Detection for AUVs Using YOLOv8 and Enhanced RRT Path Planning
Conference proceeding

Stable Wreckage Detection for AUVs Using YOLOv8 and Enhanced RRT Path Planning

Harshith Kumar M B
International Conference on Automation, Robotics and Applications (Online), (2025), pp 412-417
12 Feb 2025
DOI: https://doi.org/10.1109/ICARA64554.2025.10977642

Abstract

AUVs Classification algorithms computer vision (CV2) Marine vehicles Path planning PID Controller ROS RRT Security Service robots Simulation Stability analysis Surveys YOLOv8 Computer Vision Safety
Underwater robotics is rapidly emerging as a prominent research field globally. The underwater environment, characterized by a complex array of forces, necessitates the development of highly stable systems for effective operation. By exploring one of the least accessible terrains on Earth, underwater systems unveil a multitude of applications. This research paper presents a comprehensive system designed for Autonomous Underwater Vehicle (AUV) operations, targeting efficient underwater surveys and the detection of underwater wreckage. The system integrates several advanced technologies to achieve its objectives. A new mathematical model of the proposed AUV is designed and the stability of this system is achieved using the Proportional Integral and Derivative (PID) controller. Path planning is facilitated by the enhanced Smooth Rapidly-exploring Random Tree (RRT) algorithm, enabling the AUV to autonomously navigate complex underwater terrains. The AUV is also equipped with an onboard camera system that utilizes the YOLOv8 computer vision algorithm for classification, allowing it to differentiate between marine life, human divers, and wreckage. Furthermore, the system incorporates localization capabilities for identifying wreckage. This integrated approach aims to enhance the efficiency of underwater surveys, offering significant potential for applications in marine research, environmental monitoring, and search and rescue missions.

Metrics

1 Record Views

Details

Logo image