Vertigo: autonomous glider for coral reef surveys

October 13th, 2020

Since late 2018, a team of  Scientists from CSIRO’s Data61 Cyber-Physical Systems Research Program, and from CSIRO’s Oceans and Atmosphere business units are collaborating with BABEL-sbf and the Queensland University of Technology to develop monitoring and communication technologies for the collection of underwater video data in coastal ecosystems.

Our goal is to build machine-learning models and algorithms that recognise important environmental and physiological characteristics of these ecosystems with human-level accuracy, at superior speeds, depths, and under adverse environmental conditions.

An autonomous underwater glider robot equipped with smart vision.

The CSIRO research is led and sponsored through CSIRO’s Active Integrated Matter Future Science Platform (AIM FSP).


Coral reefs are home to a large number of marine animal species, but being unique and fragile, they are in danger from global warming, pollution, overfishing and boat traffic.

‘Manta Tow’ is the current technique used for data collection to understand the current and future of the reefs. It requires a snorkel diver (observer) to be towed at a constant speed behind a boat. The observer makes a visual assessment of specific variables during each two-minute session, manually recording this data on plastic paper attached to the manta board.

The technique relies on trained experts and is limited by tow-speed, depth, favourable weather conditions, and the amount of visual information that a human is capable of processing in real-time. We believe that technology can greatly enhance its accuracy and efficiency.


The team is developing a towed underwater glider platform for broad-scale coral reef surveys. Equipped with computer vision, accurate underwater geo-location, near real-time communications, and marine vessel autonomy, the Vertigo glider platform can improve the scale of broad-scale surveys by orders of magnitude, while achieving human-level accuracy and latency.

The platform can translate huge video datasets collected (half a million calibrated images per day) into higher-level actionable information, such as species population estimates and spatial coverage. Automated visual classification machine learning algorithms are applied to bypass the time-consuming process of archiving and uploading large video datasets.

The initial glider design relies on a human-operated surface tow-vessel, however, the process can be fully automated when deployed with a long-range autonomous surface vessel such as SailDrone.


This new reef survey capability will provide societal and environmental impacts through better, more accurate, data-driven management of coral reef environments and timely mitigation of their major threats. In addition, the platform will enable exploration of deeper and more remote environments, which will lead to new science and better understanding of coral reef habitats.

Annual GBRMPA-funded diver-based manta operations on the Great Barrier Reef (GBR) average 1.5 hours per day and 73 days per year, covering 50 reefs in a year, or about two per cent of the GBR. Each glider, surveying at 4kts and operating six hours in a day, would increase the spatial coverage of the GBR 6-8 times over the 73 day-per-year-period. Three such gliders operating 200 days-per-year could survey all reefs in the GBR, every year.

In addition to the dramatic coverage improvement, we estimate 10 to 100-fold improvements in target localisation accuracy and 10 to 100-fold improvements in latency of information flows, leading to a faster and more precise response. As each glider captures almost 500,000 images per day, each single day of operations would capture 1.5 million calibrated reef images, or approximately the same number of images captured in the last 30 years by the teams that have conducted long term monitoring of the GBR.

Project benefits include allowing the exploration of deeper and more remote environments whilst mitigating major threats, increasing the rate of marine data gathering by 1000-fold, and creating permanent archives of imagery that can be reviewed to address management issues that emerge in the future.

For more information, contact us.

Related content


Subscribe to our News via Email

Enter your email address to subscribe and receive notifications of new posts by email.