When AI goes over the horizon: mapping the future of marine surveys

As with all disruptive technologies which challenge tried and tested orthodoxy, the true extent and potential of autonomy in the survey industry is still being revealed. This will continue as new systems, along with AI and machine learning, enable more human-dependent processes to be completed by machines. And we are already quite a long way down that track, not least around the UK and the Continental Shelf, which provide immediate opportunities to dispatch and operate USVs which are fully autonomous, independent of onshore controls.

Similarly, the barriers to successfully operating over-the-horizon USVs for hydrographic and geophysical surveys are rapidly being overcome. As practical concerns are being mitigated, regulatory codes of practice are being adapted and developed and forward-thinking manufacturers are providing technical solutions that enable ever-greater autonomy and efficiency.

But a vessel operated by remote control from the shore is not the same thing as a vessel which is independently capable of maintaining all of its operational functionality in the absence of any form of connectivity, at the same time avoiding shipping, marine mammals, and all other dynamically positioned obstacles.

Significant advancements in graphic processor unit (GPU) technology over the past few years have enabled AI software to be deployed ‘on the edge,’ which means the AI models can be deployed on physically small GPUs, which in turn are mounted on the boat. These are relatively small, roughly the size of an A5 box with low electrical consumption, but are extremely powerful machines. This is the significant step change in chip technology led by NVIDIA, which three years ago would still have required a server rack level installation. Edge computer devices with the AI models installed on board ensure enough intelligence is built into the vessel to be able to carry out missions without having to rely or constantly revert to an ‘always on’ satellite connection. 

Without delving into specifics, the electric-hybrid powertrain we have developed solves the historic issues of reliability and power optimisation.

While it may still be argued by a few that USVs and marine autonomy is restricted to areas away from structural or geophysical hazards, I would argue this isn’t the case, with advances both in AI and in connectivity technology, particularly in providing robust communications over the horizon. 

With the ushering in of low earth orbit satellites such as Starlink, and hybrid connectivity solutions hitting the market which bond multiple bearers (4G, LTE, satellite, etc), autonomous navigation and edge collation of survey data are becoming much less of an issue, and certainly this will continue to develop.

When the Mayflower Autonomous Ship project, sponsored by Promare, IBM and other partners, completed its fully autonomous Atlantic crossing in 2022, it was the first autonomous USV to do so. At the heart of the vessel’s onboard AI systems was UK-based software engineering company Marine AI, who together with the parent company MSubs (responsible for the design and marine engineering), jointly developed the vessel’s capability to traverse the North Atlantic Ocean autonomously, devoid of any human intervention. 

It is these same two companies who are providing the ‘under the hood’ systems and technology that are crucial for the deployment of Zero USV’s Oceanus 12 class with the AI and advanced sensor systems, enabling the vessels to navigate and make decisions in real-time. 

Beyond its autonomous capabilities, Oceanus 12 is equipped with an array of sensors and scientific instruments aimed at collecting data on oceanographic and environmental variables. 

Extensive R&D and innovation, aligned with robust engineering and planning, will help to overcome the physical challenges associated with many of the perceived challenges with autonomy in marine surveying, including towing. Hull-mounted multi-beam echosounders are part of the solution, but it’s AI that is emerging as the real gamechanger.

With AI, survey companies can automate and optimise speed, in addition to immediate-course and whole voyage routeing for fuel optimisation and vessel safety. This yields cost-saving benefits for vessel owners, fleet managers, insurers and cargo customers. As recently noted elsewhere, “When used properly within an optimised workflow, from streaming the seafloor to producing the final geophysical product, AI will be responsible for massively improved operational efficiency.”

Given this, commercial performance and resultant job security will follow through quickly. In sum, AI promises immediate improvements to marine data acquisition and comes at a time when scarcity and the cost of human skills and experience are increasingly significant industry factors. But AI will not replace skilled surveyors – it will help to streamline operations and the collation of accurate data and ultimately reduce human error and risk.

Furthermore, with AI, survey companies are ingesting data the entire time and whether it is object data from the computer vision system or the data from the radars and AIS, all of it can be used to train the AI data sets and curate these for improvements in the field.

As for USVs, they need to deploy the sensors that are fit for purpose, such as high-definition radar that can help with near-range situational awareness and obstacle detection and avoidance.

By dint of their much lighter payload – with no crew with all their attendant food, water and accommodation – USVs help reduce carbon emissions, so they are greener, which means they can travel further, and longer, without shifts and other human dynamics.

Without a full understanding of the power and reliability of AI, it’s easy to see why some still fear the deployment of USVs in shipping lanes, and yet as the Mayflower voyage proved, the reality is very different and it is only going to get better. 

USVs can also work in pairs or even fleets, looking out for each other, or in tandem with crewed vessels if need be. Whatever the future, get ready for more computers to take the helm of survey vessels.

*Matthew Ratsey is managing director of Zero USV