Biohybrid Robots Use Living Muscle to Jump, Walk and Grip Objects

A robot made with living mouse muscle can jump about 5 millimetres, walk across a surface and grip a small object, all by turning a muscle contraction into movement through a soft artificial skeleton. Researchers at ETH Zurich demonstrated the system in a study published in Nature Communications on September 11, 2026, offering a new approach to building machines that move without conventional electric motors.

The development addresses a persistent problem in biohybrid robotics: growing muscle tissue is not enough to make a useful machine. Muscle cells can contract in a laboratory, but much of that movement is lost if the surrounding structure cannot transfer the force effectively. The ETH Zurich team designed the living tissue and its supporting structure together, using computer simulations to find configurations that produce larger movements from a limited amount of muscle.

Why living muscle has been difficult to turn into a useful robot

Most robots use electric motors, gears or other mechanical actuators to generate movement. Biohybrid robots take a different approach by combining living cells with manufactured materials, allowing biological tissue to supply some of the force. The appeal is that muscle naturally contracts and relaxes, while soft supporting materials can bend rather than collide with objects as rigid machinery might. The difficulty is making these biological components strong, predictable and durable enough to perform a defined task.

In earlier designs, researchers often developed the muscle tissue first and then attached it to a scaffold, the artificial structure that supports the tissue and converts contraction into movement. That approach can waste force if the scaffold bends in the wrong direction or the connection between the two materials slips. ETH Zurich's researchers instead treated muscle and scaffold as one mechanical system, asking the computer to determine how both should be shaped to maximise movement.

The experiments used skeletal muscle tissue grown from mouse-derived cells and a soft hydrogel scaffold. A hydrogel is a water-rich material that can deform like soft tissue while providing structural support. The researchers incorporated fine grooves into the scaffold to help the muscle fibres align and anchor themselves, improving the transfer of force from contracting cells to the surrounding structure.

Computer-designed muscle structures produced up to 11 times more movement

The team used computer simulations and an optimisation algorithm to search for combinations of muscle geometry and scaffold properties that would produce large deformations. Rather than assuming that more muscle automatically means stronger movement, the method examined how the placement and shape of the tissue influenced the mechanical response of the entire device. The researchers then manufactured the selected structures and tested them experimentally, feeding practical manufacturing information back into the design process.

According to the published study, the resulting bioactuators achieved up to an elevenfold increase in range of motion compared with earlier designs containing a comparable volume of muscle tissue. Range of motion describes how far a structure moves between its starting and contracted positions; it does not mean that the robot is eleven times faster or eleven times stronger. The distinction matters because a robot that bends substantially farther may still have limited speed, lifting capacity or endurance.

The individual actuators measured approximately 1.5 centimetres in length. Their supporting scaffolds were produced using xolography, a light-based three-dimensional printing technique that solidifies selected regions of a material to create a shaped structure. The researchers reported that printing a scaffold took about a minute, although that figure covers the printing process rather than the full production time, which also involves preparing and growing the living muscle tissue.

The same biological actuator can jump, walk, swim and grip

The researchers demonstrated several movements using the muscle-powered structures. In a jumping test, a single contraction lifted the device approximately 5 millimetres from the bottom of its container, around one-third of its body length. A walking version moved at 6.2 millimetres per second, demonstrating that the system could generate repeated forward motion rather than simply flexing in place. These are laboratory-scale movements, but they show how changing the geometry of the supporting structure can turn the same basic biological mechanism into different kinds of movement.

The team also demonstrated swimming and gripping. In the gripping experiment, a mounted actuator held an object and moved it when its muscle tissue contracted, then released it when stimulation stopped. This is a useful proof of mechanical control, but it should not be confused with an autonomous robot that can identify an object, decide to pick it up and carry it somewhere without human direction.

The experiments relied on electrical stimulation to trigger muscle contractions. That means the researchers could initiate movement by applying an electrical signal, rather than requiring the robot to make its own decisions about what to do. The work therefore demonstrates a new way to produce movement, not a complete robotic control system with independent perception, planning and navigation.

Why the scaffold matters as much as the muscle

The main engineering lesson is that a biological actuator's performance depends on how effectively its parts work together. A contracting muscle produces force by shortening, but the resulting motion depends on where that force acts and how the supporting material responds. If the scaffold deforms in an unhelpful direction or the tissue loses its grip on the surface, a substantial portion of the contraction may fail to move the robot as intended.

By designing both components together, the researchers could tailor the scaffold to the movement they wanted. Fine grooves helped orient muscle fibres and hold them in place, while the overall geometry converted contraction into a larger change in shape. This approach also provides a potential route to combining multiple actuators into more complicated structures, although building a larger machine introduces new problems involving tissue growth, nutrient delivery and consistent performance.

The findings are relevant to other soft robotic systems because they show that mechanical design can be as important as improving the material that generates force. Instead of treating living tissue as a component that must fit into a finished robot, engineers can optimise the entire structure around the tissue's behaviour. That could be useful for small devices that need to manipulate delicate objects or move through environments where a hard, rigid machine would be poorly suited.

Living muscle still needs a controlled environment

The most significant limitation is that the muscle tissue must remain alive. The demonstrated devices depend on a nutrient solution to sustain their cells, so they are not yet practical machines that can be placed in an environment and left to operate independently for weeks or months. As a device grows larger, delivering oxygen and nutrients throughout its tissue while removing waste becomes more difficult. The researchers identify the development of an artificial vascular system, which would perform some of the transport functions of blood vessels, as a major challenge for larger biohybrid robots.

Living tissue also varies in ways that manufactured components generally do not. Two muscle samples grown under similar conditions may develop differently, producing inconsistent contractions or movement. That variability matters for any application requiring repeatable positioning, predictable force or dependable operation. The team's optimisation method can improve the design, but it cannot eliminate biological differences or guarantee that every manufactured actuator will behave identically.

There is also a gap between demonstrating movement and proving that the technology is useful outside a laboratory. The study does not establish that these devices can perform sustained industrial work, survive uncontrolled outdoor conditions or operate without continuous biological maintenance. Their small size and soft construction make them interesting research platforms, but applications such as medical manipulation or environmental exploration would require substantial additional testing.

What researchers need to prove before these robots become practical

The next step is to determine whether the same design principles can produce more reliable actuators and more complex machines. Researchers will need to measure how performance changes as the tissue matures, how long the muscle remains functional and whether multiple units can work together without losing coordinated movement. Improving those properties would help establish whether the approach can move beyond carefully controlled demonstrations.

For now, the most important result is not that living muscle has replaced the electric motor. It is that engineers have shown how computer-designed biological tissue and a soft supporting structure can be built as a coordinated system, producing much larger movements than earlier designs with comparable muscle volumes. Whether that advantage translates into useful machines will depend on solving the less visible problems of keeping cells alive, controlling their variation and maintaining consistent performance over time.

Those are the questions to watch as biohybrid robotics develops: not simply how far a muscle-powered robot can jump, but how long it can keep moving, how reliably it can repeat a task and whether its biological advantages justify the additional care required to operate it.