At the threshold where engineering meets biology, a team of researchers has built a 140-milligram robot that moves through air, land, water surface, and underwater depths using a single morphing body — no swapped parts, no separate systems. The machine embodies a principle called functional recursion: one structure, reassigned again and again, performing what nature has long achieved through millions of years of evolutionary refinement. In an era of increasing specialization, this tiny device asks a quiet but profound question about whether simplicity, not complexity, is the deeper form of int
140-mg morphing microrobot masters flight, crawling, and water locomotion through shape-memory muscles
One structure performs multiple locomotion modes.
Mark
So the robot uses the same muscles to fly, crawl, and swim. How does it know which mode to be in?
Mimi
It doesn't, really. The robot receives different electrical signals—different frequencies and amplitudes—and those signals cause the frame to deform in different ways. The deformation pattern determines the mode. In flight, the muscles modulate wing geometry. On land, they oscillate the frame to drive the legs. On water, they stretch the footpads.
Luke
But that's not autonomous decision-making. The robot isn't sensing which environment it's in and switching modes. A human operator is sending the right signal for each mode.
Mimi
Correct. The robot is tethered and controlled externally. The open-loop part means once you send the signal, the robot responds without feedback. It doesn't correct itself mid-flight or mid-crawl.
Mark
What's the practical limit here? How long can the artificial muscles actually work?
Mimi
The team tested them over 10,000 cycles. They stayed stable for the first 6,000, then started degrading. Displacement dropped below 1 millimeter after that.
Luke
So if the robot is crawling at 8 hertz, that's 6,000 cycles in about 12.5 minutes. That's not very long for a real mission.
Mimi
No, it's not. The team acknowledges this is a durability problem that needs solving before the robot can do anything truly autonomous or long-duration.
Mark
The water immersion part—225 volts to sink. That seems like a lot of power for something so small.
Mimi
It is. The voltage is applied across the footpad to induce electrowetting. The higher the voltage, the faster the wetting transition. At 225 volts with 1 millimeter of footpad deformation, the robot sinks. If the footpad is stretched more, you need 300 volts.
Luke
And that voltage is coming through the tether, not from an onboard battery?
Mimi
Yes. Everything is tethered. Power, control signals, communication. The 140-milligram mass doesn't include any of that infrastructure.
Mark
So the real weight of the system is much higher.
Mimi
Much higher. The robot itself is 140 milligrams, but you need external power supplies, control electronics, a tether. The practical system is orders of magnitude heavier.
Luke
That's an important caveat. When people read "140-milligram robot," they might imagine something truly autonomous and portable. This is a lab demonstration.
Mimi
Fair point. But the principle—that you can encode multiple functions into one morphing structure—that's real and novel. The execution is tethered, but the concept scales.
Mark
What happens if the robot lands on a surface it wasn't designed for?
Mimi
The team tested crawling on four different surfaces: glass, smooth paper, balsa wood, and expanded polypropylene. On glass, the robot barely moved because there's no friction to grip. On rough foam, the legs got tangled. On paper and wood, it worked, but at different speeds depending on roughness.
Luke
So the anisotropic friction mechanism only works within a certain range of surface properties. Too smooth or too rough, and the mode breaks down.
Mimi
Yes. The design assumes a specific range of friction coefficients. Outside that range, the locomotion principle fails.
Mark
And in water—the robot glides on the surface at 6.3 millimeters per second. How does that compare to real insects?
Mimi
Water striders move much faster, but they're larger and have different leg structures. At the robot's scale, 6.3 millimeters per second is reasonable for a proof of concept. The key is that it's using the same wing-beating pattern as aerial flight, just with the body tilted.
Luke
But the robot is open-loop in water too. It's not sensing the water surface or adjusting its posture in real time. If there's a wave or current, what happens?
Mimi
The robot would likely capsize or drift. The open-loop assumption is that conditions are calm and predictable. Real-world water environments are neither.
The Pulse
- A robot the weight of a few grains of rice has demonstrated flight, crawling, water-surface gliding, and controlled underwater immersion — four distinct locomotion modes — without changing its physical body.
- The central tension is one of mass and function: packing multiple capabilities into 140 milligrams demands that every component do more than one job, leaving no room for redundancy or specialization.
- Silk-reinforced shape-memory alloy muscles — contracting when heated, recovering when cooled — solve this by bending the H-shaped frame 45 degrees while weighing 408% less than conventional designs, surviving 6,000 actuation cycles.
- Electrowetting at 225 volts breaks the surface tension that keeps the robot afloat, flipping water from a supportive medium to a pulling force and allowing controlled descent — the most delicate transition of all four modes.
- The robot remains tethered and untested as a continuous multi-environment mission, meaning the four demonstrations are still separate chapters rather than a single unbroken journey.
At the threshold where engineering meets biology, a team of researchers has built a 140-milligram robot that moves through air, land, water surface, and underwater depths using a single morphing body — no swapped parts, no separate systems. The machine embodies a principle called functional recursion: one structure, reassigned again and again, performing what nature has long achieved through millions of years of evolutionary refinement. In an era of increasing specialization, this tiny device asks a quiet but profound question about whether simplicity, not complexity, is the deeper form of intelligence.
A robot the size of a quarter has been built to do what insects do naturally: fly, crawl, glide on water, and sink beneath it — all without swapping out its body. Developed by researchers at the University of Washington and collaborating institutions, the 140-milligram, 35-millimeter machine encodes multiple locomotion modes into a single H-shaped morphing frame, a design philosophy the team calls functional recursion.
The key innovation is not adding more parts but making one structure do more work. Artificial muscles made of shape-memory alloy wire — threaded through silk reinforcement — contract when heated and recover when cooled, bending the frame 45 degrees. The silk amplifies bending force by 325 percent while cutting mass by 408 percent compared to conventional designs, and the muscles remain stable across 6,000 actuation cycles.
In flight, asymmetric muscle contraction deforms the wing linkages unevenly, tilting lift vectors to produce pitch, roll, and yaw without real-time sensor feedback. The robot achieves a thrust-to-weight ratio of 1.2 and can rotate 160 degrees in yaw within a single second. On land, the same oscillating frame drives four legs with anisotropic feet — gripping harder in one direction than the other — converting symmetric wobble into forward motion at 0.46 body-lengths per second.
On water, hydrophobic kirigami footpads expand under muscle actuation, widening the contact line with the surface and amplifying surface-tension support enough to carry 1.7 times the robot's own weight. Wing beats at 90 hertz propel it across calm water at 6.3 millimeters per second. To descend, a 225-volt signal induces electrowetting across the footpads, reducing their water repellency until surface tension reverses — pulling the robot down rather than holding it up — while wing flapping stabilizes posture during the transition.
The robot currently operates tethered, and its four locomotion modes have been demonstrated separately rather than as a continuous sequence. Muscle fatigue beyond 6,000 cycles and the challenge of autonomous multi-environment navigation remain open problems. But the principle has been established: a single morphing structure, driven by minimal actuation, can recursively become something new in each environment it enters.
A robot the size of a quarter has learned to do what insects do naturally: move through air, crawl across ground, glide on water, and dive beneath the surface—all without changing its core body. Researchers at the University of Washington and collaborating institutions have built this 140-milligram machine by encoding multiple functions into a single morphing frame, a design principle they call functional recursion. The robot is 35 millimeters long. It weighs as much as a few grains of rice. And it can fly, walk, and swim.
The breakthrough lies not in adding more parts but in making one structure do more work. At the heart of the robot sits an H-shaped frame driven by artificial muscles made of shape-memory alloy—a metal that contracts when heated and returns to its original shape when cooled. These muscles are threaded through silk reinforcement, a biohybrid approach that amplifies their bending power by 325 percent while cutting mass by 408 percent compared to conventional polymer encapsulation. The silk works by converting the metal's contraction into outward deflection, allowing the frame to bend 45 degrees with minimal added weight. This is the engineering problem the team solved: how to pack multiple locomotion modes into something so light that a single coin could balance it.
In the air, the morphing frame acts as a transmission system. Electromagnetic actuators drive four wings through planar linkages, but the artificial muscles attached to these linkages change the wing geometry itself. When the muscles on one side contract more than the other, the wings beat asymmetrically, tilting the lift vectors and allowing the robot to pitch, roll, and yaw without feedback control. In open-loop flight tests—meaning the robot receives no real-time sensor data to correct its motion—it achieved a thrust-to-weight ratio of 1.2, enough to lift itself and modulate attitude in three axes. The pitch response reached 40 degrees in 0.7 seconds. Yaw rotation hit 160 degrees in one second. These are crude maneuvers by insect standards, but they prove that deforming the body itself can steer the machine.
On land, the same frame becomes a crawler. The robot has four legs with claw-like feet attached to the ends of the H-frame. When the artificial muscles oscillate, they bend the frame back and forth in a symmetric pattern. But the feet are designed with anisotropic friction—they grip harder in one direction than the other, like a ratchet. This asymmetry converts the frame's symmetric wobble into net forward motion. At 8 hertz, the optimal driving frequency, the robot crawls at 0.46 body-lengths per second on smooth paper, or about 16 millimeters per second. It can steer by driving one side faster than the other, achieving a minimum turning radius of 42.6 millimeters. The same muscles that modulate wing kinematics now propel legs across the ground.
Water presents a different problem. At the scale of milligrams, surface tension dominates. The robot's feet are covered with hydrophobic kirigami—a material patterned with cuts and folds that can expand and contract. When the artificial muscles stretch these footpads, the gaps between the kirigami ligaments widen, increasing the contact line where water meets air. This amplifies surface-tension support, allowing the robot to carry a payload 1.7 times its own body weight while floating. With wings beating at 90 hertz in a pattern similar to aerial flight, the robot glides across calm water at speeds up to 6.3 millimeters per second. The inclined stroke plane converts vertical lift into horizontal thrust, and subtle yaw adjustments—the same attitude modulation used in flight—change the direction of gliding without altering the net force.
To sink, the robot must break the surface. The team applied a voltage across the kirigami footpads, inducing electrowetting—a phenomenon where an electric field reduces the contact angle between water and a hydrophobic surface, making it wettable. At 225 volts, with the footpads deformed to 1 millimeter, water wicks into the gaps. The menisci merge. Surface tension flips from supporting the robot to pulling it down. The robot sinks feet-first, and wing flapping provides open-loop posture correction to prevent capsizing during descent. This is the final mode: controlled immersion, where the same frame that flew and crawled and glided now transitions into a new medium.
The artificial muscles themselves are a feat of miniaturization. Each consists of a 48-micrometer-diameter shape-memory alloy wire anchored above a carbon-fiber substrate and wrapped in silk. When current flows through the wire, Joule heating triggers a phase transformation—the metal shifts from its martensitic (soft) state to its austenitic (stiff) state, contracting sharply. Cooling allows recovery. The team characterized this cycle over 10,000 repetitions and found the muscles remained stable for the first 6,000 cycles, with displacement degrading afterward due to structural fatigue at the clamping interface. For the demonstrations in this work, the durability was sufficient. But scaling to longer missions will require addressing this wear.
The robot operates in a tethered configuration, with power and control signals delivered through a cable. It is not yet autonomous. The demonstrations are separate functional tests—takeoff, crawling, water gliding, immersion—rather than a single continuous mission chain. Connecting these modes into a complete sequence remains future work. Yet what the team has shown is a principle: that a single morphing structure, driven by minimal actuation, can recursively reassign its form and function across radically different environments. This is how insects move through the world. Now, at 140 milligrams, a machine can too.
Notable Quotes
The robot demonstrates takeoff, terrestrial crawling, water-surface gliding, and water-surface-to-underwater transition as separate functional tests rather than a single continuous autonomous mission.— Research team, Nature publication
By integrating soft actuation, structural mechanics, and interfacial control into one reconfigurable structure, this work establishes a paradigm of functional recursion, where a minimal physical platform repeatedly reassigns form and function.— Research team, Nature publication
Want the full story?
Read the original at Nature ↗
Economy & Finance