Just a short distance away from me, two astronauts are practicing for a spacewalk. I’m drifting weightlessly, in a silence broken only by my own breathing and the occasional update from Mission Control in my headset. But this isn’t the dark void of space. I’m in Houston, scuba diving in a massive swimming pool that NASA uses to train astronauts for zero-gravity environments. And though it’s a thrill to watch the space-suited figures at work, I didn’t come to see them. I’m here for a peek at Aquanaut, the bright orange robot that we’re sharing the pool with.
Aquanaut glides smoothly through the water like a miniature submarine. At first, it doesn’t seem all that different from other unmanned underwater vehicles, or UUVs, equipped with sensors for gathering data and thrusters for propulsion. Then, in what could be a scene from the movie Transformers, the top part of the robot’s hull rises up from the base, exposing two massive arms that unfold from either side. A wedge-shaped head packed full of sensors rotates into place, and in a matter of seconds, the transformation is complete. The sleek submarine is now a half-humanoid robot, ready to get to work.
Aquanaut represents a radical new design that its creators, at a startup called Houston Mechatronics Inc. (HMI), hope will completely change subsea robotics. Conventional UUVs typically fit into two categories: torpedo-like free-swimming submersibles, which are used for long-distance survey missions, and boxy remotely operated machines, which are tethered to support vessels and used for underwater manipulation. HMI wants to combine both of these modes into a single robot. It’s a bold approach that no one has attempted before.
The HMI engineers, who often joke that building a Transformer has been one of their long-term career objectives, are convinced that it can be done. Aquanaut has been designed primarily for servicing subsea oil and gas installations. The companies that own and operate this infrastructure spend vast sums of money to inspect and maintain it. They rely on robotic technologies that haven’t fundamentally changed in decades, largely because of the challenge of working in such an extreme environment. For HMI, however, that’s not a problem: Of its 75 employees, over two dozen used to work for NASA. Extreme environments are what they’re best at.
HMI cofounder and chief technology officer Nic Radford spent 14 years working on advanced robotics projects at NASA’s Johnson Space Center, in Houston. “I’ll grant you that getting into space is harder than getting underwater,” he says. “But space is a pristine environment. Underwater, things are extraordinarily dynamic. I haven’t decided yet whether it’s 10 times harder or 50 times harder for robots working underwater than it is in space.”
Radford and fellow cofounders Matt Ondler and Reg Berka have raised more than US $23 million in venture capital since starting HMI in 2014. Now, after countless design iterations, Aquanaut is finally coming together. Before taking to the open ocean, though, the robot needs to prove itself in more controlled conditions, and that means a swim in NASA’s pool.
Holding 23.5 million liters of water and with a maximum depth of 12 meters, NASA’s Neutral Buoyancy Laboratory, or NBL, is large enough to contain a full-scale mock-up of most of the International Space Station, with room to spare. Astronauts train for spacewalks at the NBL, coming just about as close to weightlessness as you can get here on the ground. On this late-March morning, HMI has taken over the north end of the facility to test Aquanaut.
Ten meters down and with two tanks of nitrox on my back, I try to keep myself steady as I track the robot moving through the water. Aquanaut has been in one piece for only about eight days now, but the testing is going well. The only hiccup is a communication glitch with the arms, but the HMI team is unfazed; they know there’s still a lot of work to do, and the robot will be back here early tomorrow.
Radford tells me he enjoys the frenetic routine of running a startup, a sharp contrast with the typical pace of a huge government agency like NASA. Before HMI, he spent five years as chief engineer of NASA’s Robonaut program, developing a humanoid robot that flew to the International Space Station, and he later led the development of Valkyrie, an even more sophisticated humanoid platform. In his office at HMI, small 3D-printed models of Aquanaut prototypes fit right in with wall art featuring Valkyrie and Marvel’s Iron Man.
“The type of skills that we have at NASA,” he says, “putting robots in remote locations, and getting them to do useful work in austere data environments, best matched this big problem: working offshore.”
Most of what we see and hear about the offshore oil and gas industry involves work done from platforms, where people conduct underwater drilling operations from the surface. The platforms are the most visible part of the process, but there’s an enormous amount of complex infrastructure on the seabed as well.
Wellheads on the ocean floor are capped by metal assemblies used to control the flow of hydrocarbons to the surface. These structures, covered with pipes, valves, manifolds, and gauges, are so intricate they are commonly known as Christmas trees. Some are the size of a four-story building.
To perform routine maintenance on a wellhead, or to change the output of the well, some of the valves on the tree have to be turned, and with wells in deep water—below 300 meters, where divers normally don’t operate—the only way to do that is with a robotic vehicle.
For decades, the established procedure for working on deepwater wells has been to send out a remotely operated underwater vehicle, or ROV, to the well site. But you can’t just send the ROV by itself—you also have to send a large support vessel packed with highly trained people to serve as a base of operations for the ROV, which has little or no autonomy and is tethered to the surface for power and control. This gets very expensive very quickly, with typical jobs costing tens to hundreds of thousands of dollars per day.
HMI’s plan is to cut the cord—cutting out most of the need for people along with it. Aquanaut will not require a tether or a support ship. It will travel in submarine mode to its deepwater destination, where it’ll transform into its humanoid form, unfolding its powerful arms. Each arm is equipped with force-torque sensors and has eight axes of motion, similar to that of a human arm. The arms on Aquanaut also have grippers capable of turning valves on the subsea “trees” and even operating specialized maintenance tools that the robot carries with it in an internal payload bay.
Aquanaut will carry out tasks with human operators supervising but not directly controlling it. And once the job is finished, the robot will autonomously return home. Radford says the approach will make Aquanaut both faster to deploy and cheaper to operate than today’s ROVs. He estimates that costs could be well below half the market rate of a traditional operation.
The timing seems right. According to Chuck Richards , a subsea technology pioneer who is currently chair of the ROV Committee of the Marine Technology Society, based in Washington, D.C., the low price of oil over the past several years has cut profits and led to increased competition among oil companies, driving the adoption of new technologies. Richards, whose firm, C.A. Richards & Associates, in Houston, supplies equipment to dozens of subsea companies—HMI among them—explains that while the industry will likely be cautious about an innovation like Aquanaut, it will also be excited to see what the robot can do.
Richards explains that when the benefits of commercial ROV technology became evident after its introduction in the 1970s, the industry was eager to embrace it, even though things were a little rough in the beginning. “The oil companies were very helpful and patient with the ROV industry as it matured,” he says, “and I think they’ll be the same way with a more autonomous vehicle.”
Aquanaut’s main advantage over conventional ROVs depends on its untethered operation, and HMI had to solve several key problems to enable that capability. The first is simply getting the robot to the offshore work site without a large support vessel. While Aquanaut could be deployed from a relatively small boat, or even dropped out of a helicopter, the robot can travel more than 200 kilometers in submarine mode. Once it arrives, the robot transforms into ROV mode, with additional thrusters hidden inside the hull folding out to make it more maneuverable.
The transformation itself was another major challenge—and a source of much internal debate. “We fought ourselves the whole way trying to prove that we didn’t need to do it,” says Sandeep Yayathi, Aquanaut’s chief engineer, who prior to HMI was the power lead on the Lunar Prospector rover at NASA. But the group eventually decided that the benefits outweighed the added complexity: They were going to build their underwater Transformer.
To enable Aquanaut to alter its shape so drastically, the robot is equipped with four custom linear actuators that separate the top and bottom halves of its body. Additional motors, also highly customized and housed in waterproof cases, drive the arms and the head. For power, Aquanaut uses a lithium-ion battery similar to those found in electric cars. The full transformation currently takes just 30 seconds.
But perhaps none of these challenges is as significant as designing Aquanaut’s control system. Traditional ROVs have multiple live camera feeds, and human operators maneuver these vehicles with joysticks in real time. Without a tether, the only way to communicate with Aquanaut is through an acoustic modem. This well-established technology has a range of a few tens of kilometers underwater, at the price of high latency and very low bandwidth, in the neighborhood of a few kilobytes per second, at best. HMI plans to rely on small unmanned surface vessels to act as relays between the robot and communication satellites, and from there, Aquanaut can be controlled from anywhere in the world. However, these constraints make direct human control impractical, so Aquanaut will need to do as much as it can on its own.
“There’s a lot of autonomy that has to be built in,” Yayathi explains. “You trust the robot to do a lot.”
HMI is planning on maintaining high-level supervisory control over Aquanaut, while delegating most of the low-level decisions to the robot’s powerful onboard computers, which run the Robot Operating System, or ROS, a popular software platform for research and commercial robots. Using the sensor suite in the head, which includes stereo cameras, a structured light sensor, and a sonar system, the robot constructs a detailed 3D rendering of its surroundings. But instead of trying to send the entire 3D map back to the operator, only very small and highly compressed subsections are transmitted, and the operator can then match them to an existing model of the structure that Aquanaut is looking at.
The operator then sends simple commands, such as “Turn the valve at these coordinates 90 degrees clockwise.” The robot will autonomously decide how to grasp the valve and how much force to apply while turning, and it will send back a confirmation when the task is complete. The operator is still directing the robot’s actions, but in a way that doesn’t require steering the robot by hand, or a bandwidth-intensive live video feed.
HMI’s long-term plan is to sell Aquanaut interventions as a service. Using small fleets of robots distributed across areas like the North Sea or off the coast of California, oil and gas companies would simply need to request that a given task be completed, and HMI would then schedule the nearest robot to take care of it. Radford says it takes about seven people to operate a single traditional ROV. “We think we can invert that,” he says. “We think one operator could run seven Aquanauts.”
With a low-bandwidth connection and an operator only intermittently in the loop, there may be a greater risk of something going wrong, says Matthew A. Franchek , a professor of mechanical engineering at the University of Houston and director of the International Subsea Engineering Research Institute . “The uncertainty is there,” he says. “I’m worried about a malfunction during an operation, which could have both financial and environmental consequences. Although the technology is exciting, they’re going to need to prove that it’ll work.”
After three exhausting days testing Aquanaut at the NBL, the team celebrates with a crawfish boil in the parking lot behind the HMI office, accompanied by improbable cans of Robot Fish IPA, which came all the way from a brewery in Brooklyn, N.Y. Stories about robotics at NASA flow as quickly as the beer, while I learn how to play cornhole and suck the juice out of crawfish heads.
A sense of relief that the testing went well transitions easily into excitement about the future. Radford explains that the current version of Aquanaut is primarily a demonstration and testing platform, designed for relatively shallow water with a maximum operational depth of 300 meters. While this version could perform commercial operations in many parts of the world, HMI is already in the process of designing a scaled-up version that will be able to travel several hundred kilometers and reach depths of 3,000 meters, necessary for servicing areas like the Gulf of Mexico.
And of course, commercial operations aren’t the only things that HMI is exploring for Aquanaut. Radford couldn’t talk to me on the record about any potential work with the U.S. Department of Defense, but in late 2018 the Defense Advanced Research Projects Agency announced a program called Angler seeking proposals to “develop an undersea autonomous system that can navigate to and physically manipulate objects on the sea floor.” DARPA’s illustration accompanying the announcement features a streamlined robotic submarine with two arms, a concept that bodes well for a certain Houston company.
The party continues outside, but folks are already starting to trickle back to their desks, intent on getting Aquanaut ready for its next NBL test. Its first open-water demonstration will likely take place at a naval technology exercise in Rhode Island in August. For a robot that was in pieces in March, it’s an aggressive timeline, but Radford is confident that his team can handle it.
“It’s fun to come to work on something that’s audacious,” he says. “We think there’s a better, more cost-effective way to do work underwater. And we’re going to prove it.”
This article appears in the August 2019 print issue as “The Underwater Transformer.”