DESIGN AND IMPLEMENTATION OF AUTONOMOUS UNDERWATER VEHICLES FOR SUBSEA PIPELINE INSPECTION

ABSTRACT This study explored the development of an Autonomous Underwater Vehicle (AUV) tailored for subsea pipeline inspection, with a strong focus on affordability, precision, and operational independence. Pipelines remain the backbone of offshore oil and gas transport due to their reliability and costefficiency. Still, they carry inherent vulnerabilities; corrosion, material fatigue, construction flaws, and unintentional interference from third-party operations. Left unchecked, these issues can escalate into leaks that compromise human safety, degrade marine ecosystems, and incur significant financial loss. While several detection techniques have gained traction over the years, this project centered on a leaner, vision-based AUV system that eliminates the need for high-cost hardware such as sonar arrays or Doppler Velocity Logs (DVLs). By anchoring the system in real-time computer vision and machine learning, particularly through the implementation of YOLOv8 for object detection and Visual SLAM for localization, the vehicle demonstrated efficient pipeline segmentation and fault identification without dependence on traditional navigational instruments. Field simulations yielded promising metrics: the deep learning model reached a mean average precision (mAP) of 97.4%, with obstacle avoidance and localization performing reliably through vision alone. The mechanical design of the AUV was rated capable of withstanding pressure levels consistent with offshore Nigerian pipeline depths, an encouraging sign for regional deployment. Tests were conducted in controlled conditions, and while these offered consistency for benchmarking performance, they didn't fully capture the unpredictability of open-sea environments. Visual-only systems may also face difficulty in lowvisibility waters, where suspended particles or algal blooms can impair camera function and disrupt model inference. Despite these constraints, the project introduces a practical and significantly more affordable alternative to conventional inspection technologies. Keywords: AUV, Pipeline Leak Detection, Computer Vision.