It weighs little more than a coin, 1,25 grams. But when asked to lift something, that tiny piece of polymer does what no human muscle could: it supports 5 kilos, 4000 times its weight. It's not a circus stunt. It's the result of years of research into smart materials, the kind that change state as if they had a personality. Soft when flexibility is needed, rigid when it's time to support a load. A bit like a colleague who only wakes up when his back needs to be broken. Except this colleague is a magnetic actuator developed byUNIST University in South Korea, and it could change the way we think about robots, prosthetics and wearables.
The problem with soft actuators: either strong or flexible
Artificial muscles have always been a matter of compromise. You can have a soft and adaptable material, perfect for interacting with humans without hurting them, or a rigid and powerful actuator, capable of lifting heavy loads.But never both at the same time. It's the classic problem of soft robotics: flexible materials fail under load, rigid ones don't adapt to complex environments. Hoon Eui Jeong, professor of mechanical engineering at UNIST, decided to break this precarious balance. His team published the results in Advanced Functional Materials, proving that you can have your cake and eat it too. Or rather: a muscle that's rubber when needed and steel when it matters.
The secret lies in the double-lattice polymer. It's not an entirely new concept, but here it's been implemented with an elegance that makes the difference. The material uses chemical covalent bonds for the load-bearing structure, those that guarantee mechanical resistance, and thermosensitive physical interactions for flexibility. These form and break based on temperature, allowing the muscle to soften or tighten on command. Add surface-treated magnetic microparticles, and you have an actuator that responds to external magnetic fields with millimeter precision.
In the stiffened state, this 1,25 gram muscle supports 5 kilograms. In the softened state, it stretches to 12 times higher its original length. During contraction it lifts weights with a deformation of the86,4%, more than double the approximately 40% of human muscles. The work density reaches 1150 kJ/m³, 30 times higher than that of biological muscle tissue.
How stiffness change works
The trick lies in the two-layer polymer network. Imagine a structure made of steel beams (the covalent bonds) and elastic cords (the physical interactions). When it's cold, the cords stiffen and everything becomes solid. When it's hot, the cords soften and the structure can deform. Except here we're not talking about ambient heat and cold, but about controlled thermal stimuli which allow switching from one state to another in a matter of moments. Magnetic microparticles do the rest: they respond to external magnetic fields and allow the actuator's movement to be controlled without physical contact.
The great thing is that this system doesn't require heavy batteries or cumbersome cables. It's controlled remotely via magnets. Perfect for applications where weight and size make a difference: advanced prosthetics, robotic exoskeletons, wearable devicesAs long as there's a magnetic field nearby, the actuator responds. Need it to lift something? It stiffens. Need it to bend? It softens. Like a biological switch, but without the biology.
Applications: from humanoid robots to prosthetics
Where is such an artificial muscle used? Wherever delicate interaction and brute force are needed in the same system. collaborative robots Cobots are prime candidates: they must work alongside humans without harming them, but they must also lift industrial loads. Until now, this was an impossible equation. With actuators that vary in stiffness, the problem shifts from material to control: simply program when to stiffen and when to soften.
Le robotic prostheses are another obvious field. As already seen with the Italian GRACE musclesActuators that mimic biological muscles allow for more natural and precise movements. But the Korean ones go further: they don't just imitate, they surpass. A robotic hand equipped with these actuators could grasp a crystal glass without breaking it and then lift a 20-kilogram suitcase. All with the same set of "muscles."
Then there are the wearable devices for rehabilitation. Exoskeletons that assist people with motor disabilities, robotic suits for workers lifting heavy loads, gloves that amplify grip. All areas require flexibility to follow human movements and power to assist with effort. UNIST actuators could make these devices lighter, less bulky, and more energy efficient.
The numbers that count
Let's make a direct comparison. An average human muscle contracts by about 40%, develops a work density of about 40 kJ/m³, and can support loads up to about 30 times its own weight (in the best cases, with training). The Korean actuator contracts by 86,4%, develops 1150 kJ/m³, and supports 4000 times its own weight. It's not an incremental improvement, it's a paradigm shift.
Professor Jeong says it clearly in the study published in September 2025This research overcomes the fundamental limitation that traditional artificial muscles were either highly extensible but weak, or strong but rigid. Work density is particularly significant: it indicates how much energy per unit volume the muscle can deliver. Achieving high values while maintaining high extensibility has always been a challenge. Like trying to build a rubber band that is also a steel bar. These researchers have succeeded.
The research was funded by National Research Foundation of Korea and represents a significant step forward in the field of soft robotics. The developed actuators could find applications in the coming years in humanoid robots, industrial handling systems e advanced medical devices.
Actuators and Muscles 2.0: What's Still Missing?
Of course, it's not all gold. Laboratory prototypes work very well under controlled conditions, but the real world is more complicated. How long do these actuators last under repeated stress? How do they cope with extreme temperatures, humidity, and vibrations? And above all: how much do they cost to produce on an industrial scale? The study doesn't go into these details, but these are questions that will soon emerge.
Then there's the issue of control. Changing stiffness is useful, but it requires a sophisticated control system to decide when to do so. In a walking robot, for example, each leg must stiffen and soften dozens of times a second, synchronizing with the others. Sensors, algorithms, and real-time feedback are needed. The material technology is ready. Control technology still needs to evolve.
Finally, there's the issue of standardization. Each lab develops its own actuator with slightly different materials and processes. Common standards, reproducible tests, and shared metrics are needed. Otherwise, everything will remain confined to academic papers, which are great to read but useless for anyone wanting to build a real robot.
That said, UNIST's work is an important step forward. It doesn't solve all the problems of soft robotics, but it solves a major one: the tradeoff between strength and flexibility. And when you solve a fundamental problem, others become easier to tackle. Or at least, less impossible. Which is already an achievement for scientific research.
