A robot bird built by engineers at MIT and EPFL has done something most conventional drones cannot: swim underwater and then flap its way out into the air. The small experimental machine uses flexible wings and a movable tail to travel between two environments that demand very different kinds of movement. Its designers took inspiration from birds such as puffins and petrels, which are equally at home flying and diving beneath the surface.
The research, published in July 2026, describes a prototype weighing less than 300 grams. In water-tank and lake tests, the engineers adjusted its wings, tail and flapping patterns until it could complete the transition from water to flight. This is a laboratory demonstration of an unusual vehicle, not an announcement that autonomous seabird drones are now commercially available to explore the oceans.
What makes it memorable is the seemingly contradictory task. Wings need to push effectively against air, yet water puts far greater pressure on them. Instead of using separate propellers and wings for different journeys, the researchers asked whether one adaptable flapping mechanism could do both jobs. The answer from their controlled tests was encouraging.
Why a robot bird faces different forces in air and water
Air is thin compared with water. A flying machine has to move enough air to remain aloft; an underwater vehicle must contend with much greater resistance when it pushes its surfaces through liquid. A wing that is stiff enough for flight might meet excessive drag underwater. A wing that yields easily in water might fail to produce useful lift in the air. Designing around these contradictory demands is the central challenge.
Animals have evolved combinations of body shape, flexible tissue and movement that deal with the trade-off. Diving birds do not swap their wings when they enter the sea. They adjust the way they flap and use their bodies to move between media. The researchers studied the way several species move and treated those observations as engineering clues rather than copying the appearance of one particular bird.
The resulting machine has a central body containing a battery and motor, a pair of flexible wings and a steerable tail. Its motor drives a mechanism that moves the wings up and down, while adjustments to the tail help control whether it rises or dives. The team also made the wings interchangeable, which allowed it to test how changes in size affected performance.
That flexibility matters. A conventional drone achieves stability by controlling separate propellers, and an underwater robot may rely on propellers and fins. This prototype seeks a shared motion for both environments. It is mechanically simpler in some ways, but the wing design and control challenge become harder because the same surfaces must work in air and water.
The wing size that made both journeys possible
MIT’s account describes three sets of wings tested by the researchers: narrow, medium and wide versions. The medium span, measuring about 80 centimetres, offered the most dependable compromise between swimming and flying. The largest wings could produce useful aerial forces but were harder to work through the water. The smaller ones did not deliver the same balance. The result shows why an elegant idea still has to be tuned against real physical forces.
For several trials, the team placed the robot around half a metre below the water’s surface, programmed its flapping and tail angles, and watched whether it could rise through the water and enter sustained flight. Experiments began in a tank and moved to Lake Geneva in Switzerland. The robot was reported to swim at close to one metre per second and fly at around six metres per second under the test conditions.
The underlying study, published in Science, sets out the movement experiments. It is important to read those figures as results from a particular small prototype, not promised operating speeds of future field robots. The researchers were finding combinations of physical design and movement that permitted the transition. They were not demonstrating long-term autonomous missions in rough seas.
The machine’s wings use thin membranes with a water-repellent treatment. That helps them move from wet to dry conditions, but the more important behaviour is their ability to bend in a useful way. They need enough compliance to limit resistance underwater while retaining the characteristics needed for flight. Nature provided the inspiration; testing revealed which engineered compromise worked.
A future tool for seeing what happens beneath the waves
One possible application is environmental observation. Imagine a lightweight vehicle travelling over water, dipping below the surface to take a measurement and then returning through the air to another location. Researchers could study a small area without sending a large vessel to every sampling point. The MIT team has discussed future missions near ice, ports and marine animals, but those scenarios remain prospective uses of the technology.
Today, underwater measurement generally involves a choice of platforms. Boats can carry large instruments and operate for longer periods, but they may be costly or difficult to position in shallow or hazardous areas. Flying drones can reach places quickly but normally cannot dive beneath the water and continue flying. Submersible robots do the opposite job well, yet are designed around moving through water. Combining two capabilities may become useful where repeated movement across the surface is part of a mission.
LiveAIWire has covered technology intended to help identify whales at sea. A diving and flying research vehicle would answer different questions from an image-recognition system, but both highlight the need to observe difficult environments without placing people close to hazards. A drone could eventually carry measurements that complement remote cameras or acoustic instruments. The present prototype, however, does not establish that it can collect those measurements on a normal field deployment.
Instrumentation creates another engineering problem. Adding cameras, water sensors, sample containers or communications hardware changes the vehicle’s weight and balance. Extra equipment may affect how it leaves the water. Its designers will need to assess endurance, navigation, recovery and what happens if a component stops responding while the machine is submerged.
What can go wrong when a drone has two homes
The transition through the water surface is an especially demanding moment. The forces acting on the wings change rapidly. The wings also carry water as they emerge, which adds weight and alters airflow. A small disturbance in the tail angle or flapping cycle could mean the vehicle fails to take off. Researchers must therefore understand the whole transition, not just underwater movement and flight considered separately.
Natural water adds uncertainties absent from a test tank. Wind can change the conditions for flight while waves shift the surface the robot must cross. Currents change the starting point underwater and may push the vehicle towards obstacles. Communications can also become difficult when a transmitter is submerged. The published work describes a promising mechanism, but it does not demonstrate reliable operation through all those conditions.
A safety-conscious field system would need to recover if it fails to leave the water. That could mean flotation, a return-to-surface procedure or an operator who can retrieve it. The correct response would depend on the mission and location. No researcher should have to call a laboratory success a deployable product before these questions are answered.
Other research into very small flying robots faces similarly sharp compromises between weight, power and control. A prototype can be impressive because it proves that a movement is possible, while still lacking the endurance and reliability expected of practical equipment. Both facts can be true at once.
Why copying animals is useful without copying everything
Biology offers an extraordinary range of approaches to movement. Birds dive, fish leap and insects hover, but an engineered machine is built from different materials and has different power constraints. Scientists therefore often learn the most by isolating a physical principle rather than making a machine that looks completely natural. In this case, the principle concerns flexible wings that can contribute to propulsion on either side of the water’s surface.
A successful design could eventually inspire several kinds of machines. Some might be small research instruments; others might test methods of movement that work in more complicated environments. None of those applications follows automatically from this one prototype. Developing them requires engineers to choose new payloads, understand the relevant water conditions and determine whether the flapping mechanism is better than existing alternatives.
The distinction is worth retaining when watching demonstration videos. A machine that hops from water into air makes a striking visual story. The engineering achievement is subtler: a workable balance between forces that usually encourage entirely different designs. If future prototypes reproduce the behaviour reliably outside controlled tests, the technology could become a practical way to travel across one of nature’s most awkward boundaries.
The next challenge is an ordinary working day
The researchers have demonstrated movement that could eventually support new methods of ocean observation. A full operating system would have to do more than make one successful transition. It would need to launch, reach a target, gather useful information, return and survive repeated journeys without constant adjustments from its designers. Those are longer and more demanding tests than the experiments reported so far.
The lessons also apply to robots trained for different physical environments. Learning one action is valuable, but the world is untidy and a practical machine must handle changes that were not in its demonstration conditions. Physical adaptability comes from a combination of good hardware, careful control and a realistic understanding of failure.
For now, the diving robot’s achievement is simple to grasp and difficult to accomplish. A lightweight, bird-inspired machine has swum through water and flown out again using flapping wings. It points to an intriguing possibility for future research vehicles, while reminding us that a remarkable new motion is the beginning of an engineering project rather than its finish.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.