A muscle-powered robot thin enough to resemble a small strip of gel has swum and steered through water using a single layer of living skeletal muscle cells. The device is not fast, autonomous or ready for the open ocean. Its importance is more basic: researchers have shown that living muscle does not need to be built into a bulky three-dimensional block to move a useful robotic body.
The work, published in Advanced Functional Materials, turns a centimetre-scale hydrogel sheet into a two-finned swimmer. Each side carries a patterned monolayer of muscle cells that has been engineered to contract when illuminated with blue light. By stimulating one fin, both fins or changing the frequency of the light, the team can alter direction and speed.
The muscle-powered robot is mostly a very carefully designed skeleton
The striking part of the demonstration is the living tissue, but much of the engineering problem sits underneath it. Skeletal muscle produces force by contracting along fibres. If those fibres are badly aligned, or if the surface underneath them is too soft, much of that force is lost before it can move the robot.
The researchers therefore treated the hydrogel as a microscopic skeleton. They tested different groove shapes, dimensions and material stiffnesses to encourage cells to line up in the same direction and pull together. Square-bottomed grooves about 25 micrometres across gave better alignment than flatter or wavier patterns in their experiments.
They also changed the hydrogel material. The final design used gelatin methacrylate, a tissue-engineering material whose stiffness can be tuned. The resulting film was about half a millimetre thick, while the living muscle layer itself was less than 15 micrometres thick. That geometry lets a thin sheet of cells bend the supporting material instead of wasting force inside a thick tissue block.
The MIT account of the work describes a robot roughly the length and width of a stick of gum. That comparison is useful because the device looks much more like a flexible flap than the metal-and-motor machine people normally imagine when they hear the word robot.
A thin sheet of muscle solved an old efficiency problem
Biohybrid robotics combines living tissue with engineered structures. Muscle is attractive because it is soft, responsive and capable of repairing some damage, but conventional laboratory-grown skeletal muscle is often built as a three-dimensional bundle. Those bundles can require large numbers of cells and can be difficult to align precisely.
The paper reports that its optimised two-dimensional actuators produced about 20 times more contractile force per unit of tissue volume than the three-dimensional muscle actuators used for comparison. The researchers also report a large increase in untethered lifetime compared with earlier thin-film designs, from less than ten minutes to more than 30 days in culture. Those figures describe this experimental platform, not a general guarantee for every biohybrid robot.
That efficiency is the reason the work could matter beyond an eye-catching swimming demonstration. A smaller amount of tissue that can generate useful motion gives engineers more freedom to make light, flat or highly compliant machines. It could also lower the biological burden of building them because fewer cells may be needed for a given mechanical effect.
LiveAIWire has covered robots learning skills that can transfer between different bodies. The muscle-powered swimmer attacks a different layer of the same broad problem: what the body itself should be made from, and how its material properties change the kind of control that is possible.
The swimmer moves by following flashes of light
The cells in the fins were genetically engineered to respond to blue light. When light hits one muscle sheet, the cells contract and the fin bends. Stimulating the opposite side changes the movement. In a dish, the team manually moved a light source to steer the robot through a simple watery maze.
At its fastest, the robot travelled at about four body lengths per minute in straight motion. That is slow by the standards of conventional aquatic robots, and the researchers explicitly describe the body as a basic first demonstration. The point was to show that the thin muscle system could create enough force for controlled locomotion, not to compete with a propeller.
The same limitation helps keep the result in perspective. The robot did not decide where to go, recognise obstacles or navigate an uncontrolled environment. It was externally stimulated in a laboratory setting. Any future field machine would need sensing, control, power or signalling systems and a way to keep its living tissue viable outside carefully maintained conditions.
That separates it from machines such as the robot dog that completed a marathon on one battery charge. The quadruped story is about energy storage and conventional robotic endurance. The aquabot is asking whether a living actuator can make an entirely different class of machine viable.
Living tissue changes what a robot can be
A soft robot powered by muscle does not have to beat an electric motor on raw speed to be useful. Its potential advantage is compliance. A living tissue actuator can bend around fragile surroundings rather than forcing a rigid mechanism through them. The researchers suggest long-term possibilities including environmental monitoring and, more broadly, tasks where soft interaction is valuable.
There is also an appealing design principle here: use the biology where biology is naturally good. Muscles already convert chemical energy into controlled mechanical motion. Instead of building a miniature imitation from gears and motors, biohybrid engineering asks whether the tissue itself can become the actuator.
That does not remove engineering complexity. Living cells need nutrients and suitable conditions. Genetic modification, sterility, long-term control and manufacturing consistency all become part of the robotics problem. A device that survives in a dish for weeks is still a long way from an inexpensive field robot that can be stored, transported and deployed on demand.
Safety also changes when machines combine engineered biology and robotics. A conventional robot can normally be switched off and left on a shelf. A living component has a lifecycle, biological waste and potentially different containment requirements. Those questions will matter more if biohybrid systems ever move beyond laboratory research.
The real advance is not the maze
It is easy to focus on the charming image of a tiny swimmer following a light around a dish. The stronger result is the actuator beneath that scene. The team showed that patterned, two-dimensional skeletal muscle can remain attached, generate useful force and control motion for much longer than earlier thin-film attempts.
That gives biohybrid robotics a new piece of engineering vocabulary. A robot does not necessarily need a motor bolted to a frame, or even a thick artificial muscle wrapped around a joint. It can be a sheet whose material and living cells are designed together so that the structure itself becomes the mechanism.
Other robotics research is already pushing in the opposite direction, giving increasingly conventional machines smarter software and stronger autonomy. LiveAIWire has examined whether robots can refuse unsafe instructions. The aquabot reminds us that intelligence is only one frontier. Reinventing the physical body may be just as important.
For now, the little swimmer remains a proof of principle. It is slow, externally controlled and dependent on living cells maintained in laboratory conditions. Yet that is exactly why the experiment is interesting: it demonstrates the minimum physical idea clearly. One carefully patterned layer of muscle can be enough to make a machine swim.
The manufacturing question may eventually be as important as the biology. A laboratory can hand-build a small number of patterned hydrogel devices and carefully culture the cells. A useful robot platform would need repeatable fabrication, predictable muscle strength and quality control across many units. Small variations in cell growth, alignment or gel stiffness could otherwise produce machines that behave differently even when they share the same design.
That makes the paper valuable as an engineering map as well as a demonstration. By measuring how groove geometry, substrate stiffness and training affect force, the researchers identify variables that future designers can tune rather than treating living tissue as an unpredictable black box. Biohybrid robotics becomes more credible when biological performance can be connected to controllable manufacturing choices.
For readers, the useful benchmark is therefore not whether this swimmer outruns a conventional robot. It is whether the thin actuator remains controllable, durable and manufacturable enough to justify using living tissue at all. That is the engineering contest the next versions will have to win.
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.