NUS scientists have developed a self-training method that strengthens lab-grown muscle tissues around the clock, and used them to power a living-muscle robot that swims faster than any of its predecessors

Singapore, 19 March 2026 — Researchers at the National University of Singapore (NUS) have developed a platform that lets lab-grown muscle tissues train themselves to record-breaking strength, with no external stimulation required. By mechanically coupling two muscle tissues so they continuously pull against each other, their own natural contractions become a round-the-clock workout. The resulting muscles powered OstraBot, an ostraciiform (a type of fish locomotion) swimming robot that reached 467 millimetres per minute — the fastest speed reported for any skeletal muscle-driven biohybrid robot.

The advance removes a long-standing bottleneck in biohybrid robotics — machines driven by living cells rather than conventional motors. Because muscle-based actuators are soft, quiet and efficient at small scales, stronger versions could unlock minimally invasive biomedical tools, soft environmental sensors and fully biodegradable robots that safely degrade after completing their task.

“For years, researchers have been interested in building robots powered by living muscle because biological actuation is soft, adaptive and energy-efficient at small scales. However, the performance of these systems has been limited by the low force output of cultured skeletal muscle. If the actuator is weak, the robot cannot move fast, generate meaningful thrust, or perform useful tasks,” said Assistant Professor Tan Yu Jun from the Department of Mechanical Engineering in the College of Design and Engineering at NUS, who led the research.

“The purpose of this study was not just to build a faster robot, but to remove a fundamental bottleneck in the field and open the door to high-performance biohybrid systems designed with sustainability in mind,” Asst Prof Tan added.

The study was published in Nature Communications on 18 March 2026. In December 2025, the first author of the paper, Dr Chen Pengyu, won the Best Poster Award based on this study at the Materials Research Society (MRS) Fall Meeting 2025, one of the largest international conferences for materials science research.

Two muscles in an arm-wrestling match

The key insight came from a behaviour that biologists have long observed but rarely exploited: the spontaneous contractions that young skeletal muscle cells produce as they mature. Starting around day three of differentiation, engineered tissues begin twitching on their own, peaking by day five before fading as the cells reach full maturity. Although most researchers had treated this as a biological curiosity, the NUS team treated it as a training resource.

They designed a platform in which two muscle tissues are coupled through a sliding block, so that when one contracts, it stretches the other, which then contracts back. The result is continuous cycles of shortening and lengthening that run autonomously throughout the week of early maturation, with no external power source, control unit or manual intervention.

“As the cells mature, they naturally begin to contract spontaneously. Because the two tissues are connected, they continuously pull against each other, effectively exercising without any external control,” explained Asst Prof Tan.

The self-trained muscles generated a maximum force of 7.05 millinewtons and a stress of 8.51 millinewtons per square millimetre — the highest values recorded for this cell line in biohybrid robotics, and more than an order of magnitude above many previously reported figures. The method uses a commercially available muscle cell line found in labs worldwide, making it far more reproducible and cheaper than conventional approaches.

Optimising OstraBot to achieve personal bests

The team developed a physiology-based model tracing the full chain from electrical stimulation through calcium signalling and muscle activation to force output, then used it to guide OstraBot’s design. Inspired by the boxfish, which keeps its body rigid and propels itself entirely by oscillating its tail, OstraBot pairs this model-informed structure with a single trained muscle that drives two flexible tails. At optimal stiffness and 3 Hz stimulation, it swam more than three times faster than an identical robot powered by conventionally cultured muscle.

Beyond speed, the robot demonstrated something equally significant: precise controllability. Its speed could be tuned continuously by adjusting electrical field strength, and a sound-triggered system let it start and stop in response to clapping signals.

“The clap shows that the robot is not just alive — it is controllable. In the past, muscle-powered robots either moved constantly without clear control or were too weak to respond visibly. Our strengthened skeletal muscle allows the robot to react clearly to an external signal, similar to how nerves control muscles in the body,” said Asst Prof Tan. “This demonstrates that biohybrid robots can combine strength with precise regulation, which is essential for real-world applications.”

Robots with a vanishing act

The NUS team is now pursuing systems in which all structural materials are biodegradable — robots that perform their function and then safely break down. Possible applications include environmental monitoring devices deployed in sensitive ecosystems such as wetlands or coral reefs, as well as temporary implantable tools that perform a clinical task before dissolving inside the body, eliminating the need for surgical retrieval.

“Strength is one important milestone, but long-term stability, energy efficiency and lifecycle design are equally important,” said Asst Prof Tan. “Ultimately, we aim to develop biohybrid machines that are not only high-performance but also environmentally responsible by design.”

The team's next steps include integrating biodegradable structural materials, refining control strategies and improving the durability and efficiency of muscle-powered robotic systems.