An Autonomous Sail-Powered System for Non-Invasive Multimodal Marine Monitoring

Background

Despite the UN global consensus to protect 30% of the ocean by 2030 (United Nations, 2022), just 8.4% of the ocean is designated as Marine Protected Areas (MPAs), of which only 2.8% are effectively enforced (Marine Conservation Institute, 2023). As a result, the majority of marine key biodiversity areas remain vulnerable to threats such as illegal fishing. Conventional patrol vessels are fuel-intensive, costly to operate at scale, and can disturb marine mammals, limiting their effectiveness and accessibility to resource-limited regions (Kelly et al., 2004).

In addition to holding vessels accountable, improved marine monitoring technology could be critical to reducing vessel collisions with marine mammals. Approximately 20,000 whales are killed by vessel strikes annually (Gavigan, 2025), which is acutely felt by small populations like the North Atlantic Right whale (Moore et al., 2021). Improved detection and documentation of both whale locations and vessel behaviour is essential to reducing the frequency of these encounters.

Lastly, accessible ocean surveillance methods could be applied to broader research and conservation purposes, as marine species are difficult and expensive to study in situ. Globally, more than 40% of of marine mammal species are threatened with extinction (Vazquez et al., 2024; IUCN, 2023), and as conventional research vessels can disrupt their hunting behaviour (Mitson, 2022), there is a need for non-invasive studies to understand and manage their threats (Vazquez et al., 2024). Low-cost monitoring technology could also improve long-term environmental data collection, with applications including noise pollution monitoring and biodiversity observations.

Approach

This project aims to provide a fleet of low-cost autonomous robots to address the critical data gap in marine research and conservation. To ensure accessibility, the following engineering objectives were conceived:

1. Low-cost

→ The hull is made from laser-cut plywood and aluminum, and it is easily assembled and wrapped in fibreglass.

2. Low-disturbance

→ An airfoil provides propulsion with no disruptive engine noise, and the energy to power the components is obtained through non-polluting solar panels.

3. Effective

→ Monitors above and below the surface with high-quality hydrophones and cameras, and covers larger distances with a fleet than a single patrol vessel of the same cost.

Methodology

The vessel was designed in Onshape, and each component was laser-cut from 0.5 mm birch and 6.35 mm aluminum plates. The tabs at the top were drilled vertically to maintain sharp, precise edges, and M4 threads were hand-tapped into the screw holes. The frame was assembled using carbon fibre rods for strength. The hull was wrapped in multiple layers of fibreglass fabric and two epoxy resin gel coats, and testing was conducted to ensure it was fully waterproof.

Two custom hydrophones were constructed using piezoelectric transducers and amplifier circuits at a cost of approximately $80, significantly lower than commercial alternatives. They were positioned orthogonally, with one detecting sounds in the front and back, and the other detecting sounds from the left and right. The hydrophones were encased in epoxy using a custom mould, and air bubbles were removed from the resin using a vacuum and heat to ensure clear signal transmission. Then, 4 kg of scrap lead was tied to the keel for righting torque in case of capsizing in rough waters, and another coat of epoxy sealed the lead weights and hydrophones inside the bulb keel capsule.

Autonomous Navigation

A Raspberry Pi microcomputer processes inputs from four cameras to generate a 360° field of view using custom stitching software based on OpenCV SIFT feature matching. The resulting panoramic image, annotated with directional bearings, is analyzed using machine learning to identify objects of interest.

The system also integrates cellular communication and Global Navigation Satellite System (GNSS) data via a Walter ESP32-S3 module, enabling each vessel to follow predefined routes and transmit real-time alerts to shore. Navigation is further supported by an ArduPilot flight controller, which receives inputs from an anemometer, compass, and speed sensor to adjust a trailing-edge flap on the airfoil. This allows the sail to optimize its angle of attack for propulsion, enabling travel at speeds up to twice the apparent wind speed.

All onboard systems are powered by solar panels and lithium iron phosphate batteries, allowing for sustainable autonomous operation.

Surface Monitoring: Object Detection

Camera data is processed using a YOLO-based object detection model trained on a combined dataset of trimaran imagery and external marine datasets. The model detects vessels, debris, and marine mammals in real time. YOLO’s lightweight architecture was selected over more complex models to ensure compatibility with onboard processing constraints, and allows image analysis every three seconds in open-ocean conditions.

Underwater Sensing: Audio Classification

Two perpendicular hydrophones mounted in the keel detect marine mammal vocalizations and vessel noise. Acoustic data is processed using machine learning classifiers trained on open-source datasets, including whale calls and engine noise recordings. Audio clips were augmented with background noise and converted into spectrograms for classification.

Differences in signal intensity between the hydrophones are used to estimate the direction of incoming sounds, enabling basic acoustic localization beyond the visual range of the cameras.

In-situ testing showed that the autonomous trimaran can carry out sustained marine monitoring while avoiding many of the limitations of traditional patrol vessels. Although each unit moves at lower speeds, deploying them as a coordinated fleet allows for wider coverage over time at a much lower cost. Because the system runs on wind and solar power, it reduces both operating expenses and environmental disturbance, and its modular design makes it easier to transport and assemble in resource-limited regions. Autonomous operation also removes risks to onboard personnel and enables continuous, long-term data collection.

A key strength of the system is its ability to combine surface imaging with underwater acoustic sensing on a single platform. This allows it to detect both vessel activity and marine wildlife at the same time, rather than relying on a single data source. However, performance still depends on environmental conditions like wind variability, and the accuracy of audio classification is limited by the availability of high-quality labelled data. Longer-term testing is needed to assess durability and reliability in open-ocean conditions.

Instead of relying on a few large, resource-intensive vessels, this approach uses a distributed fleet of smaller autonomous systems. This shift makes large-scale, low-impact ocean monitoring more realistic and opens the door to more accessible and continuous data collection in support of conservation and enforcement efforts.

Future

Piloting the system in collaboration with marine conservation organizations will be essential to assessing its reliability in different environments. Currently, only one trimaran has been built; ideally, a coordinated fleet would be tested for large-scale marine monitoring. It may also be necessary to increase hull size for more pelagic and open-ocean applications, which will require refinements to the design to reinforce structure and improve manufacturability. Additional developments could include a dedicated app for remote monitoring and control, improved machine learning models for better detection of vessels and wildlife, as well as more advanced hydrophone arrays to improve audio localization.

I would like to thank my parents, grandparents, and sister for supporting me through the many hours it took to assemble this project.

I would like to acknowledge Amanda Barney, the CEO of Teem Fish, for her generosity in taking the time to advise me on the marine monitoring aspects of my project.

Many thanks to Mr. Chan, Ms. Yip, and Mr. Stevens for sponsoring Eric Hamber’s science fair team.

I would also like to thank the Greater Vancouver Regional Science Fair, the Vancouver District Science Fair, and the Youth Innovation Showcase for the incredible opportunities they offer and for providing me with feedback on my work.

References

United Nations. (2022). Kunming-Montreal global biodiversity framework. https://www.unep.org/resources/kunming-montreal-global-biodiversity-framework

Marine Conservation Institute. (2023). MPA atlas. https://mpatlas.org/

Kelly, C., Glegg, G., & Speedie, C. (2004). Management of marine wildlife disturbance. Ocean & Coastal Management, 47(1–2), 1–19. https://www.sciencedirect.com/science/article/abs/pii/S0964569104000183

Gavigan, E. (2025). 20,000 whales are killed by vessel strikes each year. International Marine Mammal Project. https://savedolphins.eii.org/news/20-000-whales-are-killed-by-ship-strikes-each-year

Moore, M., Rowles, T., Fauquier, D., Baker, J.., Biedron, I., Durban, J., Hamilton, P., Henry, A., Knowlton, A., McLellan, W., Miller, C., Pace, R., Pettis, H., Raverty, S., Rolland, R., Schick, R., Sharp, S., Smith, C., Thomas, L., van der Hoop, J., & Ziccardi, M. (2021). Review: Assessing North Atlantic right whale health: Threats and development of tools critical for conservation of the species. Diseases of Aquatic Organisms, 143. https://www.int-res.com/journals/dao/articles/dao03578

Vazquez, J., Khudyakov, J., Madelaire, C., Godard-Codding, C., Routti, H., Lam, E., Piotrowski, E., Merrill, G., Wisse, J., Allen, K., Conner, J., Blévin, P., Spyropoulos, D., Goksøyr, A., … Vázquez-Medina, J. (2024). Ex vivo and in vitro methods as a platform for studying anthropogenic effects on marine mammals: Four challenges and how to meet them. Frontiers in Marine Science, 11. https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2024.1466968/full

International Union for Conservation of Nature. (2025). The IUCN red list of threatened species. https://www.iucnredlist.org

Mitson, R. (2022). Underwater noise of research vessels: Review and recommendations. ICES Cooperative Research Reports. https://ices-library.figshare.com/articles/report/Underwater_noise_of_research_vessels_review_and_recommendations/18624479

Nelms, S., Alfaro-Shigueto, J., Arnould, J., Avila, I., Bengtson Nash, S., Campbell, E., Carter, M., Collins, T., Currey, R., Domit, C., Franco-Trecu, V., Fuentes, M., Gilman, E., Harcourt, R., Hines, E., Hoelzel, A., Hooker, S., Johnston, D., Kelkar, N., … Godley, B. (2021). Marine mammal conservation: Over the horizon. Endangered Species Research, 44, 291–325. https://www.int-res.com/journals/esr/articles/esr01115

Images

Marine Conservation Institute. (2023). Marine protection Atlas. https://mpatlas.org/