Self-sensing artificial muscles made with graphene fibers play piano and move robotic eyes

Researchers at the University of Pennsylvania have developed artificial muscle fibers that sense their own contraction state and hold intermediate positions on demand, a capability that has largely eluded soft actuators. The fibers, called CoilLCE, combine a liquid crystal elastomer (LCE) with multilayer graphene powder and an ultrathin copper coil embedded along the fiber’s length, and translate electrical signals into muscle-like motion.

Ziyun Zhang, co-lead author of the study, with the robotic eyes and hand used to demonstrate the CoilLCE artificial muscles. Image credit: Penn Engineering

The fibers contain just 0.25 wt.% graphene. At that loading the composite reached a toughness of 632 MJ/m³, more than three times the 192 MJ/m³ of a pure LCE fiber, and its stiffness decreased rather than increased, bringing it closer to the Young’s modulus of skeletal muscle, 0.1 to 0.3 MPa. The researchers attribute this to graphene sheets acting as physical barriers between oligomer chains, lowering the effective crosslink density without disrupting the liquid crystal alignment that the actuation depends on. At 0.75 wt.% the alignment was substantially impaired, and both actuation strain and breaking stress declined. The reduced stiffness also lowers the force a passively stretched fiber exerts against its contracting partner. In antagonistic pairs the graphene-loaded fibers reached an actuation strain of 16.1%, compared with 11.2% for pure LCE, an improvement of about 44%. The graphene also improved the uniformity of heating along the fiber.

The copper coil, 50 µm in diameter, serves as both heater and sensor. It Joule-heats the fiber to trigger contraction, and because copper’s resistivity varies linearly with temperature, its resistance reports the fiber’s state without any separate sensor. “We used the resistance of the embedded copper coil as a feedback signal and adjusted the heating power in real time to keep the resistance near a target value,” said Xuelin Wu, a co-author on the study. With a PID controller closing the loop, the fibers hit strain targets to within about 2% and held a 30% contraction under a 13 g load for 30 minutes, and recovered from blasts of liquid nitrogen and hot air within five seconds. “This is the first time that feedback control has been demonstrated on an LCE actuator, and it is a necessary step for these sorts of artificial muscles to be able to be used in future robotic applications,” said Sung.

The fibers measure 0.55 to 0.73 mm across, comparable to small human muscle bundles, and deliver actuation strains above 40% at rates up to around 32% per second, with a work capacity of 343 J/kg and no measurable degradation over 116 heating and cooling cycles.

Two demonstrations put the fibers to work. Robotic fingers on a 3D-printed hand played piano keys with tunable force and duration, producing around 48 mN for a short note and 120 mN for a sustained one. A pair of artificial eyes driven by eight fibers in four antagonistic pairs rotated up to 30°, with measured angles closely matching the team’s geometric predictions. The researchers point toward humanoid faces, where dozens of small coordinated muscles must fit into a confined space, as a target application. Related work has included a graphene-enabled single-fiber actuator inspired by human muscles and a liquid crystal actuator toughened with functionalized graphene oxide that activates near body temperature. Funding came from the Army Research Office’s MURI program, the National Science Foundation and its Graduate Research Fellowship Program, and the National GEM Consortium.