Marine environments are typically monitored using radar systems or human visual inspections. While effective in some contexts, both approaches are restricted. Human inspections vary with experience and judgement and are limited to mostly surface observations, while radar systems are expensive to install and maintain and cannot detect underwater activity.
Unmanned aerial vehicles (UAVs) have helped extend monitoring coverage, particularly over large areas. However, their limited payload capacity, short battery life, and sensitivity to weather make them unsuitable for long-term or continuous marine surveillance, especially below the water’s surface.
Autonomous underwater vehicles (AUVs) operate directly within the marine environment, allowing them to inspect ship hulls for corrosion or biofouling, assess water quality, and detect underwater hazards. Despite this promise, most previous AUV-based systems have focused primarily on vehicle motion control, paying less attention to integrating sensing, communication, and human oversight under real-world underwater constraints.
The new study addresses this gap by proposing a system-level design that treats sensing, communication, control, and human decision-making as a single coordinated framework.
'Human-In-The-Loop' Monitoring System
The researchers developed an AUV-assisted monitoring system that combines AUVs, onshore cameras, wireless communication networks, and a human operator based in a control center. The operator assigns tasks to both AUVs and shore-based sensors, maintaining oversight during complex missions.
When operating near the surface, cameras mounted on AUVs and onshore structures capture images of ships and buoys. These images are processed using a deep-learning model to estimate ship attitudes, such as draught and tilt.
For underwater targets, including shipwrecks or simulated drowning victims, AUV-mounted sonar systems perform acoustic detection where optical sensors and radar are ineffective.
According to the authors, this is the first system to integrate sensing, communication, and control in a single AUV-based monitoring framework designed specifically for underwater environments, distinguishing it from earlier autonomous vehicle platforms developed primarily for land or surface use.
Communication, Localization, and Control Underwater
Maintaining reliable communication underwater is a central challenge, as electromagnetic signals used in conventional wireless systems do not propagate well through water.
The proposed system, therefore, combines surface radio communication with underwater acoustic transmission, supported by a hierarchical, multi-channel data exchange framework to improve performance.
For localization, the AUVs use GPS when on the surface and switch to an underwater positioning system based on time-difference-of-arrival (TDOA) algorithms when submerged, enabling precise navigation where GPS signals are blocked. A three-dimensional motion-planning controller allows the vehicles to operate safely in the vertical and horizontal dimensions required underwater.
To estimate ship attitudes, the researchers trained a YOLOv8-OBB deep-learning network to recognize ship waterlines from camera images. In port-based experiments, the system achieved an average recognition accuracy of 92.4 % while operating at 28 frames per second.
Using the detected waterlines, the system was able to infer ship draughts, including a measured value of 8.5 m in one test case.
Visual and acoustic data collected by the AUVs were also used to reconstruct a three-dimensional model of the port basin using Unity (U3D) software, supporting inspection planning and long-term infrastructure assessment.
Rather than replacing human control, the system incorporates the open-source AI model DeepSeek as a shore-based decision-support tool. DeepSeek assists operators with mission planning, task prioritization, and interpretation of multi-AUV data streams, while final control authority remains with the human operator.
In controlled pool experiments simulating underwater search scenarios, the AI-assisted framework helped coordinate AUV movements and supported reliable sonar-based target detection, with acoustic detections confirmed visually.
Experimental Validation and Practical Implications
The system was tested through real-world port experiments for surface ship monitoring and controlled pool experiments for underwater target detection. Together, these tests demonstrate that cooperative AUV systems can support reliable marine monitoring under conditions relevant to ports, infrastructure inspection, and search-and-rescue preparation.
While the study does not provide a quantitative cost analysis, the authors argue that automated, cooperative AUV monitoring could reduce reliance on manual inspections and improve operational efficiency over time.
The researchers note that the current system does not address aerial monitoring and suggest that future work could integrate AUVs with UAVs and unmanned surface vessels. Such cross-domain cooperation could further expand monitoring coverage across air, surface, and underwater environments.
For now, the study offers a practical demonstration of how coordinated AUV systems, guided by human oversight and supported by modern AI and sensing technologies ,could play a growing role in safeguarding marine environments and infrastructure.
Journal Reference
Yan, J. et al. (2026). Ubiquitous sensing of marine activities via the cooperation of autonomous underwater vehicles. Scientific Reports, 16, 2430. DOI: 10.1038/s41598-025-29532-y
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.