UNDERWATER FRONTIERS
How robots are helping discover the most hidden places on Earth.
By Michael W. Richardson
The ocean covers most of our planet, yet much of it remains stubbornly beyond our reach. Beneath the waves, GPS disappears, radio signals fade, visibility can shrink to almost nothing, and even a small mechanical failure can end an expedition. For robots, the ocean is therefore both an extraordinary frontier and an unforgiving engineering test. Building machines that can navigate, observe and work there requires researchers to rethink some of the basic assumptions that have guided robotics on land and in the air.
At NYU Tandon, several researchers are approaching that challenge from very different directions. Christopher Clark is developing autonomous vehicles that can follow sharks, map archaeological sites and eventually perform scientific tasks with less human supervision. Nana Obayashi is looking to biology, designing soft robots whose flexible bodies can exploit currents much as fish do. And Jason Röhr is investigating whether new solar technologies could give underwater vehicles the energy to remain at sea for far longer than batteries alone allow. Together, their work points toward a future in which oceanic robots are not merely remote-controlled machines, but persistent scientific partners capable of exploring places humans rarely see.
The Next Great Frontier for Robots Lies Beneath the Waves
For decades, robotics researchers have looked to land and sea. They have built autonomous drones that inspect bridges, deliver packages, map forests and even prepare for missions to Mars. But Christopher Clark believes the next revolution in autonomous exploration will happen somewhere far less visible: beneath the ocean’s surface.
“The underwater world is one of the last truly unexplored environments,” says Clark, Industry Professor of Mechanical and Aerospace Engineering at NYU Tandon School of Engineering. “From the robotics point of view, it’s also one of the hardest.” That difficulty is precisely what drew him to the field.
Clark has spent more than two decades developing intelligent underwater robots capable of performing scientific work that would otherwise require teams of researchers, divers and research vessels. His autonomous systems have tracked sharks along the California coast, mapped medieval water systems beneath ancient castles in Malta, searched for centuries-old shipwrecks in the Mediterranean and measured the health of fragile marine ecosystems.
The coast of Malta. Inset: Autonomous Underwater Vehicle deployed by Chris Clark.
The coast of Malta. Inset: Autonomous Underwater Vehicle deployed by Chris Clark.
Rather than treating robotics as an end in itself, Clark views robots as scientific collaborators. Marine biologists, archaeologists and oceanographers bring questions about the natural world; his lab develops autonomous systems capable of answering them.
The oceans present challenges that make even the most sophisticated aerial drones seem straightforward by comparison.
Above ground, robots can rely on GPS, wireless communication, high-resolution cameras and laser-based sensors to understand where they are and what surrounds them. Underwater, nearly all of those tools become unreliable or unusable. GPS signals cannot penetrate water. Radio communication is ineffective. Visibility is often poor, particularly away from shallow tropical waters, forcing robots to navigate using sonar and acoustic communication, both of which are easily distorted.
“The ocean is a much less forgiving environment,” Clark says. “Communication is terrible, perception is difficult and small mistakes can quickly become catastrophic failures.”
A tiny leak can flood an expensive robot. Unexpected currents can push vehicles off course. Marine life, including curious sharks, have occasionally collided with his machines during deployments. Every expedition demands extensive logistical planning, specialized equipment and considerable patience.
Those obstacles have limited the number of researchers willing to work in underwater robotics. For Clark, they provide an opportunity.
One of his longest-running collaborations demonstrates how robotics can transform marine biology. Working alongside shark researcher Chris Lowe, Clark developed autonomous underwater vehicles (AUVs) capable of autonomously following acoustically tagged sharks without requiring scientists to spend countless hours piloting boats.
“The ocean is a much less forgiving environment. Communication is terrible, perception is difficult and small mistakes can quickly become catastrophic failures.”
That strategy could dramatically reduce the time and cost required to develop capable underwater robots while opening new possibilities for exploration in environments where direct human access is dangerous or impossible.
Clark’s newest projects already point toward that future. One autonomous surface vehicle under development will carry hydrophones to detect tagged sharks while simultaneously collecting water samples, allowing researchers to correlate animal behavior with changing ocean conditions. Eventually, he envisions fleets of cooperating surface and underwater vehicles continuously monitoring coastal ecosystems.
The same technologies could support archaeological surveys, environmental monitoring, offshore infrastructure inspection and scientific expeditions in some of Earth’s least explored places.
Despite remarkable advances in autonomous systems over the past decade, more than 80 percent of the ocean remains poorly mapped and largely unexplored. For Clark, that statistic is less a limitation than an invitation.
Each improvement in robotic intelligence allows scientists to spend less time wrestling with the mechanics of exploration and more time asking deeper questions about the hidden world beneath the waves. In that sense, underwater robots are becoming more than machines. They are evolving into scientific partners, extending human curiosity into an environment that has remained beyond our reach for centuries.The robots continuously monitor individual animals, recording detailed information about where they travel, how quickly they swim and how they move through different parts of the water column. Those fine-scale behavioral measurements can then be compared with environmental variables such as water temperature, salinity, habitat and seasonal changes.
The resulting data help scientists answer questions that extend well beyond shark biology. Understanding where marine predators spend their time can inform fisheries management, conservation planning and decisions about marine protected areas.
Clark’s robots have also become unlikely tools for archaeology. In collaboration with researchers at the University of Malta, his team has spent years exploring submerged cultural heritage sites throughout the Mediterranean. Early expeditions focused on flooded tunnels, cisterns, and underground chambers beneath castles and ancient fortifications. Small robotic vehicles collected sonar scans and thousands of photographs that were later combined into detailed three-dimensional reconstructions.
More recently, those efforts have expanded offshore. Autonomous vehicles equipped with side-scan sonar survey the seafloor in systematic search patterns, generating enormous volumes of imagery. Instead of requiring archaeologists to inspect every sonar image manually, Clark’s team trained neural networks to identify locations likely to contain shipwrecks or other archaeological remains. Once a promising site is detected, robots can return for closer inspection, collecting imagery detailed enough to reconstruct entire wrecks in three dimensions.
Those digital reconstructions allow researchers to study fragile sites while reducing the need for repeated dives, helping preserve underwater cultural heritage.
Clark's team assists with tagging sharks.
Clark's team assists with tagging sharks.
Clark's team assists with tagging sharks.
Clark's team assists with tagging sharks.
That strategy could dramatically reduce the time and cost required to develop capable underwater robots while opening new possibilities for exploration in environments where direct human access is dangerous or impossible.
Clark’s newest projects already point toward that future. One autonomous surface vehicle under development will carry hydrophones to detect tagged sharks while simultaneously collecting water samples, allowing researchers to correlate animal behavior with changing ocean conditions. Eventually, he envisions fleets of cooperating surface and underwater vehicles continuously monitoring coastal ecosystems.
The same technologies could support archaeological surveys, environmental monitoring, offshore infrastructure inspection and scientific expeditions in some of Earth’s least explored places.
Despite remarkable advances in autonomous systems over the past decade, more than 80 percent of the ocean remains poorly mapped and largely unexplored. For Clark, that statistic is less a limitation than an invitation.
Each improvement in robotic intelligence allows scientists to spend less time wrestling with the mechanics of exploration and more time asking deeper questions about the hidden world beneath the waves. In that sense, underwater robots are becoming more than machines. They are evolving into scientific partners, extending human curiosity into an environment that has remained beyond our reach for centuries.The robots continuously monitor individual animals, recording detailed information about where they travel, how quickly they swim and how they move through different parts of the water column. Those fine-scale behavioral measurements can then be compared with environmental variables such as water temperature, salinity, habitat and seasonal changes.
The resulting data help scientists answer questions that extend well beyond shark biology. Understanding where marine predators spend their time can inform fisheries management, conservation planning and decisions about marine protected areas.
Clark’s robots have also become unlikely tools for archaeology. In collaboration with researchers at the University of Malta, his team has spent years exploring submerged cultural heritage sites throughout the Mediterranean. Early expeditions focused on flooded tunnels, cisterns, and underground chambers beneath castles and ancient fortifications. Small robotic vehicles collected sonar scans and thousands of photographs that were later combined into detailed three-dimensional reconstructions.
More recently, those efforts have expanded offshore. Autonomous vehicles equipped with side-scan sonar survey the seafloor in systematic search patterns, generating enormous volumes of imagery. Instead of requiring archaeologists to inspect every sonar image manually, Clark’s team trained neural networks to identify locations likely to contain shipwrecks or other archaeological remains. Once a promising site is detected, robots can return for closer inspection, collecting imagery detailed enough to reconstruct entire wrecks in three dimensions.
Those digital reconstructions allow researchers to study fragile sites while reducing the need for repeated dives, helping preserve underwater cultural heritage.
Robots That Swim Like Fish and Think Like Engineers
When a trout holds itself effortlessly still in a fast-moving stream, it is not fighting the current. It is using it. The fish’s body, shaped by millions of years of evolution, passively harvests energy from the turbulence around it, converting what seems like an obstacle into forward motion. To most observers, it is simply a beautiful thing. To Nana Obayashi, it is an engineering problem — and a solution.
Obayashi is an Assistant Professor of Mechanical Engineering at NYU Tandon, and the director of the Prema Lab dedicated to a deceptively simple-sounding goal: building robots that interact efficiently through complex environments, like fluids. She often draws her inspiration from the ocean, from the aerodynamics of her earlier training in aerospace engineering, and from a conviction that the most elegant machines are the ones that work with their environment rather than against it.
“Biology has evolved to have very efficient and intelligent body behaviors,” she explains. The trout example she cites is not hypothetical. It refers to a well-documented phenomenon in which a fish placed in a structured vortex field can propel itself upstream without any active muscular effort — purely because of how its body is shaped. In robotics, this principle is called embodied intelligence: the idea that the body itself can encode behavior, so that the brain — or the controller — doesn’t have to do all the work.
Her lab pursues this idea through soft robotics — robots made from flexible, compliant materials like silicone, rather than rigid metal or plastic. One ongoing project involves an underwater robot designed to vary its own stiffness in real time, much as a fish stiffens parts of its tail when it needs to sprint or loosens its body when cruising at low energy cost. “We’re trying to create robots that can manipulate their stiffness based on the environment,” she says. The mechanical challenge is formidable: fluid-solid interactions at the boundary between a robot’s body and the water surrounding it are notoriously difficult to model in simulation, which is why Obayashi’s lab invests heavily in physical prototyping and real-world experimentation.
“When I see marine creatures underwater, I see how beautiful they are. That has been a constant inspiration.”
One particularly striking demonstration involves a silicone tentacle — a soft, flexible appendage that can generate lift simply by moving through water. When researchers vary the tentacle’s speed, length, or material stiffness, its behavior changes dramatically, even visually. More remarkably, an optimized version of the tentacle can swim faster than an unoptimized one executing the identical movement pattern — a result that underscores how much performance is encoded in physical structure rather than in motion commands.
The environmental stakes of this work are not merely academic. Conventional underwater robots are often loud and mechanically alien to the creatures they are deployed to observe. A robot that disturbs the fish it is meant to study is not a useful scientific instrument. Obayashi envisions soft, biologically inspired machines that can monitor marine ecosystems without disrupting them — tools that look and move enough like the creatures around them to become, effectively, invisible.
But Obayashi’s ambitions extend beyond the ocean. A parallel strand of her research addresses the challenge of human- robot interaction, particularly in domestic and social settings. As robots become more common in everyday environments — already, robotic vacuum cleaners navigate living rooms autonomously — the question of how humans and machines can collaborate safely and intuitively grows more urgent. Rigid robots that are heavy, noisy, and inflexible may perform tasks competently but engender little trust. Soft robots, by contrast, can be made physically safe and behaviorally predictable in ways that hard machines cannot.
One experiment from her lab illustrates this collaborative potential vividly. A robot tasked with designing a paper airplane that could fly five meters became stuck in a local optimum — cycling through design iterations that never quite reached the target. When a human collaborator stepped in, however, the partnership broke the impasse. The robot supplied systematic data and an efficient search through design space; the human contributed intuition. Together, they reached a successful design in nine iterations. Neither would likely have done it as quickly alone.
Obayashi is also exploring a subtler kind of human-robot communication. By reading electromyography signals — electrical activity generated by muscles 70 milliseconds before a movement is even executed — her lab aims to build robots that can anticipate human intent rather than simply react to commands. The ambition is a robot that feels less like a tool and more like a partner.
It is a vision shaped equally by engineering rigor and genuine wonder. “When I see marine creatures underwater,” she says, “I see how beautiful they are. That has been a constant inspiration.”
Can Solar Power the Next Generation of Underwater Robots?
Most people associate solar energy with rooftops, sprawling desert installations or satellites orbiting Earth. Water, by contrast, seems like one of the least likely places to harvest sunlight. It absorbs and scatters light, growing darker with every meter of depth. Marine organisms quickly coat submerged surfaces, blocking what little sunlight remains.
Yet those challenges have not stopped researchers from asking an intriguing question: Could solar energy power autonomous underwater vehicles capable of exploring the ocean for months — or even years — without returning to the surface?
At NYU Tandon, Jason Röhr, Industry Associate Professor of General Engineering, is exploring exactly that possibility. His research examines how advances in photovoltaic materials could help underwater robots operate far longer than today’s battery-powered systems, opening new possibilities for environmental monitoring, ocean science and marine infrastructure inspection.
Jason Röhr
Jason Röhr
“The idea isn’t that solar is a perfect solution,” Röhr says. “It’s that it’s an interesting one.”
The work sits at the intersection of renewable energy, materials science and robotics, challenging long-held assumptions about where solar power can — and cannot — be used.
At first glance, the concept seems counterintuitive. Water rapidly absorbs sunlight, particularly infrared wavelengths that conventional silicon solar cells depend on. Dive deep enough, and darkness quickly follows. But shallow coastal waters, especially in tropical regions, receive far more sunlight than many people realize.
“If you’ve ever gone swimming in Florida,” Röhr says, “you know those waters can be incredibly clear.” That observation has prompted researchers to rethink underwater photovoltaics. Instead of relying on traditional silicon cells, Röhr studies materials that better capture the wavelengths of light that penetrate water most effectively.
One promising candidate is cadmium telluride, a commercially available solar technology that his group has already tested. Even more exciting are newer materials, including organic solar cells and perovskites, which have become major areas of photovoltaic research over the past several years.
Organic solar cells offer several advantages for underwater environments. They perform well under dim light, absorb light arriving from many different angles and can be manufactured on flexible materials. Rather than rigid panels, future solar cells might wrap around the curved hull of an underwater vehicle or even be sprayed onto surfaces during manufacturing.
“They absorb diffuse light quite nicely,” Röhr explains. “As you go deeper, the light intensity drops very quickly, so that becomes an important advantage.”
Perovskites, meanwhile, achieve even higher efficiencies, although they remain difficult to deploy in marine environments because many contain lead and degrade rapidly when exposed to moisture.Developing an underwater solar cell involves much more than choosing the right material.
One of the biggest practical challenges is biofouling — the gradual accumulation of algae, microorganisms and other marine life that coats submerged surfaces. Even highly efficient solar cells become ineffective if sunlight cannot reach them.
Researchers in other fields have developed anti-biofouling coatings for ship hulls and underwater optical equipment, and Röhr believes those technologies may eventually translate to solar applications. But they have yet to be thoroughly tested on photovoltaic devices.
His group is also investigating another, less obvious problem: the tiny electrical imperfections inside solar cells themselves.
Early theoretical work assumed nearly ideal solar cells. Real devices, however, contain microscopic defects that create unwanted pathways for electrical current. Under bright sunlight those flaws have relatively little impact. Under the weak illumination found underwater, they can dramatically reduce performance.
Röhr’s current research examines how these internal electrical losses affect underwater solar systems and how future designs might minimize them.
The work borrows concepts from indoor photovoltaics, another field where solar cells must generate electricity under extremely low light conditions.
“We’re asking how much we have to improve these devices to get back to that idealized performance,” Röhr says.
“They absorb diffuse light quite nicely,” Röhr explains. “As you go deeper, the light intensity drops very quickly, so that becomes an important advantage.”
Perovskites, meanwhile, achieve even higher efficiencies, although they remain difficult to deploy in marine environments because many contain lead and degrade rapidly when exposed to moisture.Developing an underwater solar cell involves much more than choosing the right material.
One of the biggest practical challenges is biofouling — the gradual accumulation of algae, microorganisms and other marine life that coats submerged surfaces. Even highly efficient solar cells become ineffective if sunlight cannot reach them.
Researchers in other fields have developed anti-biofouling coatings for ship hulls and underwater optical equipment, and Röhr believes those technologies may eventually translate to solar applications. But they have yet to be thoroughly tested on photovoltaic devices.
His group is also investigating another, less obvious problem: the tiny electrical imperfections inside solar cells themselves.
Early theoretical work assumed nearly ideal solar cells. Real devices, however, contain microscopic defects that create unwanted pathways for electrical current. Under bright sunlight those flaws have relatively little impact. Under the weak illumination found underwater, they can dramatically reduce performance.
Röhr’s current research examines how these internal electrical losses affect underwater solar systems and how future designs might minimize them.
The work borrows concepts from indoor photovoltaics, another field where solar cells must generate electricity under extremely low light conditions.
“We’re asking how much we have to improve these devices to get back to that idealized performance,” Röhr says.
Today’s autonomous underwater vehicles generally rely on batteries. Those batteries have become increasingly capable, but they still impose limits. Many vehicles must periodically return to the surface for recharging or recovery, interrupting long-term monitoring missions.
Solar power offers a different vision. Rather than treating sunlight as something collected only at the surface, future vehicles might continuously harvest available light while operating, extending mission durations without human intervention.
The approach would not work everywhere. Murky coastal waters or deep-ocean environments receive too little sunlight. But tropical regions with exceptionally clear water could provide enough energy to support certain classes of sensors or low-power autonomous systems.
Röhr emphasizes that solar is unlikely to replace batteries entirely. Instead, it could become another tool for designing truly autonomous marine systems.
“If you want something that’s continuously functioning while it’s doing the thing it’s supposed to be doing,” he says, “you need a power source that’s readily available.”
Building Tomorrow’s Ocean Robots
NYU Robosub team captains Filip Ovcarov, Eddy Hu, and Sohan Sondhi.
NYU Robosub team captains Filip Ovcarov, Eddy Hu, and Sohan Sondhi.
Röhr’s interest in underwater technology extends beyond his research laboratory.
He also serves as academic supervisor for NYU Tandon’s RoboSub Vertically Integrated Project (VIP), a multidisciplinary student team that designs fully autonomous underwater vehicles. Although Röhr’s own expertise lies in solar energy rather than robotics, he describes the project as overwhelmingly student driven.
Typically involving between 15 and 20 students, the team is divided into three major groups. One builds and maintains the vehicle itself, integrating its mechanical and
electrical systems. Another develops the software that controls navigation and motion. A third focuses on computer vision, enabling the robot to perceive and interpret its underwater surroundings.
Together, the students are building a completely autonomous submarine capable of navigating underwater without human control.
The project serves primarily as an educational experience, giving students hands-on exposure to robotics, embedded systems, computer vision and systems engineering. But it also has ambitious competitive goals.
Last year, the team entered its first international RoboSub competition after several years of development — a major milestone for the program. Students are now refining the vehicle for another competition in 2027, testing it in local swimming pools before deployment.
NYU Robosub team captains Filip Ovcarov, Eddy Hu, and Sohan Sondhi.
NYU Robosub team captains Filip Ovcarov, Eddy Hu, and Sohan Sondhi.
Röhr’s interest in underwater technology extends beyond his research laboratory.
He also serves as academic supervisor for NYU Tandon’s RoboSub Vertically Integrated Project (VIP), a multidisciplinary student team that designs fully autonomous underwater vehicles.Although Röhr’s own expertise lies in solar energy rather than robotics, he describes the project as overwhelmingly student driven.
Typically involving between 15 and 20 students, the team is divided into three major groups. One builds and maintains the vehicle itself, integrating its mechanical and
electrical systems. Another develops the software that controls navigation and motion. A third focuses on computer vision, enabling the robot to perceive and interpret its underwater surroundings.
Together, the students are building a completely autonomous submarine capable of navigating underwater without human control.
The project serves primarily as an educational experience, giving students hands-on exposure to robotics, embedded systems, computer vision and systems engineering. But it also has ambitious competitive goals.
Last year, the team entered its first international RoboSub competition after several years of development — a major milestone for the program. Students are now refining the vehicle for another competition in 2027, testing it in local swimming pools before deployment.
For Röhr, the project demonstrates how complex engineering problems increasingly demand expertise from multiple disciplines.
No single student builds the robot alone. Mechanical engineers, programmers, electrical engineers and vision specialists must all work together to create a functioning autonomous system
That collaborative approach mirrors the future of ocean engineering itself.
As renewable energy technologies continue to improve, underwater robots may eventually remain at sea longer, travel farther and collect more data than ever before. Whether powered by sunlight, batteries or a combination of both, the next generation of autonomous vehicles will depend on engineers willing to rethink familiar technologies in unfamiliar environments.
Sometimes, the next frontier for solar energy isn’t on a rooftop. It’s beneath the waves.
Other Stories
Building a
Quantum Ecosystem in Our Backyard
A new initiative is set to turn New York City into the new quantum hub.
Three Unanswered Questions About Spaceflight
Long-term space flights require long-term solutions for astronauts’ health and safety.
Mapping the
Invisible City
Data science is uncovering some of the most pressing issues of urban living.
LLMs, Ransomware and the AI Boom’s Cybersecurity Challenges
Cybersecurity professionals face down new attack vectors.