Autonomous vehicles are the big game changer at the moment. But while Google’s self-driving cars are racking up the miles on the sunny open roads of Mountain View, CA, roughly eight-thousand miles away in the Antarctic, an autonomous underwater robot called SeaBED has been doing much the same thing—only in far more perilous conditions, and on a much tighter budget.
Deployed in 2010 and 2012, SeaBED’s mission was to produce the world’s first high-resolution 3-D maps of Antarctic sea ice, providing accurate ice thickness measurements in areas that were previously impossible to access, and helping to better understand a mystery of climate science: what’s going on with Antarctica’s sea ice?
While researchers have been mostly successful at understanding huge declines in the thickness of sea ice in the Arctic, much of Antarctica’s sea ice has dramatically increased in recent years, stumping the predictions of climate models. At the same time, certain parts of the continent’s ice sheet are .
By going places sensors had never been, the SeaBED came back with detailed maps of ice that will help scientists better grasp the mystery of Antarctic’s anomalous ice growth, and better understand shifts in the Earth’s climate.
And SeaBED addressed another question too: could an autonomous submarine get data from beneath an ice sheet?
“This partnership between the scientific and engineering fraternities has been a real confidence builder,” says Dr. Jeremy Wilkinson, one of the study’s co-authors, from from the British Antarctic Survey (BAS). “That goes for both funders and the scientific community at large. There’s now a belief that, yes, it is possible to send off a robotic sub under the ice and believe that it will come back to you with the data you’ve asked for.”
Like a giant air conditioner for the Earth, the icy Arctic and Antarctic help regulate the climate, but they also indicate how it behaves. To climate scientists studying global warming, their ice changes can act like a canary in a coal mine.
But monitoring the thickness of sea ice, particularly in the Antarctic, isn’t simple. Space satellites can help us discern the difference between ice and water, but measuring actual thickness remains difficult.
In the Arctic, some of this task has been carried out by nuclear submarines traveling beneath the ice. Since the shortest route from the Pacific to the Atlantic involves an Arctic journey, these subs have inadvertently and intentionally collected detailed reports on ice thickness going back half a century using their built-in sonar. Since 1980, Arctic ice lost about 40% of its thickness.
The Antarctic is different. Due to a 1961 agreement called the Antarctic Treaty, which demilitarized the area to ensure that it would remain free of nuclear tests and radioactive waste disposal, subs can’t travel there. As a result, scientists must rely on alternative techniques for measuring ice thickness.
The most common of these involves drilling through the ice and taking measurements by hand. It’s time-consuming work and can prove imprecise, and focuses on areas where ships can travel, which tend to include thinner ice. The resulting data is full of holes.
“That’s where SeaBED comes in,” Dr. Wilkinson says. “And it’s changing everything.”
Robotic subs (also called Autonomous Underwater Vehicles—or AUVs) have been around for decades now, but they’ve never previously been used under ice like this. This is because of the challenges posed by icy conditions.
“The big difference between ice and open water is that, with open water, if something goes wrong the AUV it will just float to the surface,” says Hanumant Singh, an engineering scientist at Woods Hole Oceanographic Institution, and one of the designers of SeaBED. “In the case of being under ice, there is no way this can happen. Even if you can find the sub, it may be incredibly hard to reach it.”
That’s not the only challenge of operating an AUV under thick ice. While Google’s self-driving cars may have to deal with varying amounts of traffic, they don’t have to deal with the roads themselves shifting on a continuous basis.
Sea ice, on the other hand, is very dynamic and can move tens of kilometers in a single day. A hole drilled to allow an AUV underwater can quickly close or shift location.
Making things even more challenging is the fact that only minuscule amounts of data can be sent to the sub from its human operators. All information is conveyed using sound, via an acoustic modem, which becomes unusable when the sub is a certain distance away, or trapped under enough ice.
When the connection does work, in most cases, data transfer is limited to just 32 bytes per minute. The Rosetta spacecraft, which last month landed on the 67P/Churyumov–Gerasimenko comet, can communicate with Earth about 13 times faster, at around 420 bytes per minute.
“Not only is it very low bandwidth, it’s also incredibly spotty,” Singh continues. “We need to build the sub’s comms so that it can take instruction from us. At the same time, if instruction is not possible it needs to be able to operate autonomously.”
The result is that algorithms are needed that allow the the two meter, 200 kilogram sub to achieve a goal, but also to intelligently adapt to changing conditions during its five or six hour missions. Aiding these routing algorithms are a number of on-board sensors. For instance, an in-built radiometer detects light coming from the sun. Since these rays better penetrate clear water than water covered in sea ice, it is possible to set a threshold which means the sub will only ever surface if light is above a certain threshold.
“With technology constantly getting smaller, you can equip each sub with multiple sensors,” says Jeremy Wilkinson. “Not only can we analyze sea ice thickness, we can also look at the chemical properties of the water, its temperature, and the current. It’s possible to get a much more integrated and holistic view of what’s happening in the world than we ever could previously.”
Possibly the sub’s most important piece of technology is its upward-facing sonar, which allows it to measure ice thickness without having to break it. Unlike previous one-sonar subs, SeaBED emits hundreds of sonar beams in a fan-like shape. Operating at a depth of 20 to 30 meters, the sub moves backwards and forwards as if it’s mowing a lawn. In doing so, it gathers lines of data, which are then merged to create high-resolution 3-D images of the underside of the ice.
The result is faster, more efficient, and more accurate than the previous manual drilling measurements. Although things can still go wrong. The British Antarctic Survey lost its Autosub 2 robot on an initial sub-ice mission in 2005, and on the first outing of a SeaBED precursor project, a robot got stuck under the ice and had to be rescued by another robot.
“A lot of research labs were scared to take on a project like this because they worried about losing subs,” says Singh. “That’s a good rule in general, but it also precludes you from getting involved with the really high risk projects. There are many projects where ensuring the survival of your tools shouldn’t be prioritized above doing exciting science.”
Ultimately, this project was all about gathering the data—and the SeaBED was just a tool by which to get there. Fortunately the sub survived, although at one point it did get stuck under thick ice for a day and a half.
The most recent set of SeaBED data that revealed parts of the Antarctic ice sheet—already thicker than climate models had predicted—is thicker than previously thought. There are various theories as to why. It could be related to rain and snowfall, or to Antarctica’s giant land mass. But the new data helps add vital clues in the effort to understand this particular climate phenomenon. Adding to the confusion, the SeaBED team has shown that there’s still a dramatic amount of melting going on at uneven rates in various places like the Bellingshausen sea.
Some have cited the curious expansion of Antarctic sea ice as evidence that global warming isn’t happening, or that Antarctic increase is compensating for the Arctic’s melt. But Walt Meier, a climate scientist at NASA, points out that Arctic’s ice is melting twice as fast as the Antarctic’s ice is growing.
Where there was formerly no data or very hard-won data, the SeaBED’s sonar readings are producing an embarrassment of riches for climate scientists.
“Like many other aspects of computer science, underwater robotics is going through a data revolution,” Singh continues. “It used to be that we’d go out to sea and if we captured a few hundred images in a week we’d consider ourselves lucky. Now we go out with these robots and we capture a million images in that time. Add to that the number of sensors we’re using and—yes, it’s safe to say we’re going through a turning point.”
Following this groundbreaking (icebreaking?) research, the plan is for similar technology to be used to do large-scale surveys, which can then be compared to large-scale observations from aircraft and satellites.
What Jeremy Wilkinson describes as the “holy grail” is the idea that ice thickness will one day be able to be established from space using satellites. Some preliminary work has been made in this area, and the findings from autonomous robots like SeaBED can be used to compare with the results of algorithms created by scientists.
“We’re never going to have AUVs moving around constantly, but by building up more detailed maps, the data we gather can be used to compare against as satellites get better,” he says. “This is just the beginning.”