A muscle can contract strongly and still move a robot very little if its supporting structure wastes the force. ETH Zurich researchers redesigned both parts together: living muscle tissue and the soft scaffold that converts its contraction into motion. The resulting small biohybrid robots can jump, swim, walk and grasp.[1]
ETH highlighted the work on 22 October, following a Nature Communications paper published on 11 September. The team, including co-lead authors Aiste Balciunaite and Mike Y. Michelis in Robert Katzschmann's Soft Robotics Lab, used computer simulation to determine where muscle should sit, how much to use and how the surrounding hydrogel should deform.[1]
The shape turns contraction into movement
The actuators combine skeletal muscle grown from mouse cells with a water-rich hydrogel scaffold. An optimisation algorithm searches for configurations in which muscle contraction creates a large movement of the whole component. That treats muscle geometry as a design variable rather than a fixed lump of tissue to accommodate.[1]
The researchers then build and test the designs, feeding observations about material behaviour and manufacturing back into the simulation. Fine grooves help the tissue anchor to the scaffold and align its fibres, improving the transfer of force. A useful design depends on that interface as well as on the strength of the muscle itself.[1]
Reported prototype measurements
| Measure | Value | Context |
|---|---|---|
| Actuator length | About 1.5 cm | Living muscle and hydrogel designed as one component |
| Range of motion | Up to 11× greater | Compared with earlier systems using a comparable amount of muscle |
| Jump height | About 5 mm | Roughly one-third of the actuator's body length |
| Walking speed | 6.2 mm/s | Reported for a walking version |
| Scaffold printing | About 1 minute | Xolography printing time; muscle growth is a separate process |
Why a small jump is an informative test
A jump requires enough coordinated deformation to lift the device away from its support. It therefore demonstrates a different mechanical capability from a barely visible contraction. The swimming, walking and grasping examples show how related muscle–scaffold designs can translate force into different movements.[1]
The one-minute fabrication figure applies to the scaffold, which is printed with light through a process called xolography. Growing and maturing the living tissue takes additional work. Keeping these steps separate is essential to understanding what the fabrication method has accelerated.[1]
Living material changes the engineering requirements
The prototypes operate with nutrient solution to keep their muscle tissue alive. Larger or longer-lasting systems would need to deliver oxygen and nutrients and remove waste throughout the tissue. ETH identifies an artificial vascular supply as a major challenge for scaling the approach.[1]
Living muscle also varies between samples, unlike a component manufactured to a narrow industrial specification. The team wants to understand and control that development more consistently. These biological requirements shape the design space just as directly as the stiffness of the scaffold or the amount of force the muscle produces.[1]
The computer's role here is optimisation and simulation. The reported result does not require a claim that a general-purpose AI model controls the robot or that the device makes autonomous decisions. Its relevance to robotics is physical: a better arrangement lets the same kind of active tissue produce a more useful movement.[1]
That gives soft-robotics designers another variable to work with. The body can be engineered around how muscle develops, attaches and changes shape, allowing the mechanics of the structure to do more of the work that a conventional robot delegates to motors and rigid joints.[1]
Sources & further reading
- One small leap for a muscle, one giant step for robotics ETH Zurich · 22 Oct 2026 · Accessed 27 Oct 2026