宾夕法尼亚大学《AFM》：液晶弹性体-石墨烯复合纤维，构筑多级自感知人工肌肉

人工肌肉凭借可模拟生物肌肉大变形与柔顺驱动的特性，在软体机器人领域展现出巨大潜力。然而，多数软体驱动器缺乏对中间驱动状态的实时、按需精确控制，制约了其灵巧操作能力。因此，开发兼具自感知与可编程多级驱动的人工肌肉具有重要意义。基于此，本文，宾夕法尼亚大学Cynthia Sung、 Shu Yang等研究人员在《Advanced Functional Materials》发表名为”Multistage, Self‐Sensing Artificial Muscles With Coordination From Liquid Crystal Elastomer‐Graphene Composite Fibers”的论文，研究提出了一种由液晶弹性体/石墨烯（LCE/石墨烯）复合纤维与嵌入式柔性自感知焦耳热铜线圈构成的电子可控人工肌肉（CoilLCE）。该复合纤维通过在 LCE 中仅添加 0.25 wt.% 石墨烯，制备出兼具柔软与韧性的类肌肉力学响应材料。

此外，利用温度与驱动应变之间明确的单调关系，CoilLCE 实现了闭环、多级可编程中间态驱动；石墨烯的引入降低了模量，促进了拮抗协调，使拮抗致动应变达 16.1 ± 0.7%，较纯 LCE 的 11.2 ± 0.8% 提升约 44%。经测试，CoilLCE 在保持高驱动性能的同时实现了大于 40% 的大驱动应变、最高 30% s⁻¹ 的应变速率以及 343 J kg⁻¹ 的做功能力。同时，该材料还集成了自感知焦耳热与闭环控制能力，使其适用于灵巧、自适应的软体机器人。本研究提出了一种 LCE/石墨烯复合纤维与嵌入式自感知铜线圈协同的人工肌肉构筑策略，该策略通过温度–应变单调关系实现闭环多级驱动，并为可系统集成的自感知软体人工肌肉提供了新的设计思路。

图文导读

图1、Conceptual framework and the multistage actuation mechanism of the CoilLCE platform. (A) Conceptual schematic comparing the human neuromuscular system with the CoilLCE-based artificial muscle platform. Insets (i) shows coordinated actuation of 4 pairs of CoilLCEs, mimicking antagonistic behavior of extraocular eye muscles, and (ii) highlights independent force and speed control of individual “fingers”. (B) Photos of the real-time continuous control of the CoilLCE actuator in a finger model at different stages (top) and the corresponding actuation mechanism (bottom) at different temperatures. (C) Illustration of the closed-loop control in CoilLCE enabled by the self-sensing feedback mechanism. (D) Schematic of the actuation strain as a function of time, corresponding to the three representative states shown in (B), from an initial rest state (T₀, R₀) to intermediate states (T_i, R_i), where T and R denote temperature and resistance, respectively. All scale bars: 1 cm.

图2、Fabrication and material characterizations of the CoilLCE. (A) Schematic illustration of the fabrication process of CoilLCE. (B) Optical image of the as-fabricated CoilLCE fiber actuator, accompanied by a schematic illustrating graphene dispersion within the LCE precursor (purple box indicates the precursor). (Bi, Bii) The upper half of the box represents the interfacial region between graphene and the LCE precursor. The black rectangle in the middle marks the graphene edge, while the lower half of the box represents the bulk LCE region without graphene. Red spots denote unreacted acrylate groups: (Bi) Stage I: the first-step thiol–acrylate Michael addition reaction, and (Bii) Stage II: the second-step photopolymerization of acrylate groups. (C) Tensile stress-strain curves of the LCEs with different graphene loadings (without coil). (D) Storage modulus (E′) and loss modulus (E″) of the CoilLCE fibers with different graphene loadings as a function of temperature. (E) Corresponding length change of the fibers as a function of temperature.

图3、Self-sensing mechanism of the CoilLCE. (A) Schematic illustration of the circuit setup and the muscle-like uniaxial actuation mechanism with a hanging weight (clip). (B) Temperature-dependent resistivity of the CoilLCE. The experimental R was obtained at various ambient temperatures under equilibrium conditions. (C, D) Relationship between the actuation strain and (C) the resistance and (D) the surface temperature at different input current levels (n = 3).

图4、Actuation properties of CoilLCE. (A) Actuation strain vs. time under 23 g load with different graphene loadings. (B) Actuation stress under different fixed strains. Inset: the experimental setup, where the CoilLCE was fixed at its original length (L) and stretched to a fixed desired length (0.8L or 1.2L), and then actuated by a constant current of 1 A. (C) Comparison of antagonistic actuation strain between pure LCE and 0.25 wt.% graphene-loaded CoilLCE. Left: schematic of the experimental setup, in which two identical actuators of the same length are assembled in opposition, with one activated to drive bidirectional motion while the other elongates passively. Right: measured antagonistic actuation strain (ε), defined as shown in the inset formula. (D) Actuation Strain with various input currents for CoilLCE with 0.25 wt.% graphene. (E) Cyclic actuation performance of the 0.25 wt.% CoilLCE, showing the first 5 and last 5 cycles of 116 cycles of continuous heating and cooling studies under 0.49 MPa applied stress. The intermediate cycles are omitted for clarity and the full 116-cycle can be found in Figure S10. F, Comparison of the maximum work capacity with literature values from electrothermally actuated uniaxial LCE-based actuators.

图5. Closed-loop control of CoilLCE actuator for small muscle-like behavior. (A) Precision control of actuation strain at target levels of 10%, 20%, 30%, and 40%, each held for approximately 30 s. CoilLCE initial length:11.1 cm. (B) Long-duration actuation of the CoilLCE actuator under closed-loop control, maintaining ∼30% strain for 30 min while lifting a 13 g load (0.49 MPa). (C) Infrared images (top) and the corresponding R/R₀ of the closed-loop-controlled CoilLCE under fluctuating environmental temperatures. The shaded parts present the duration of temperature fluctuations. Scale bar: 1 cm. (D) Calibration curve correlating the actuation strain and resistance across three environmental conditions. Each dataset is collected by 2 fibers over 100 data points. (E) Relationship between passive elongation L_tensile, and active contraction L_actuation of the CoilLCE fiber under varying applied stress. All are operated under the set resistance (R/R₀ = 1.3).

图6、Humanoid robotics applications: individual control of artificial fingers. (A) The schematic illustration for converting linear actuation of the CoilLCE into bending motion using a soft rubber finger with elastic hinges. (B) Demonstration of dynamic gesture changes achieved through continuous programming of the actuator. The small hand icons indicate the actuated finger, as indicated by the yellow lines. (C–F) Artificial fingers playing the piano, with (C) and (E) showing force output measured by sensors placed underneath the piano keys. (C) and (D) illustrate the control of short and long notes (1/8 note and whole note, respectively) by adjusting the set resistance of the actuator. (E) and (F) demonstrate chord playing through independent control of individual fingers. All scale bars: 1 cm.

图7、Humanoid eye coordination system using CoilLCEs. (A) The working mechanism of the artificial eyes. During leftward rotation (left panel), the left-side CoilLCE actuator contracts (active), pulling the eyeball to 30°, and vice versa, enabling fast directional switching through alternating actuation. (B) The linear prediction of the actuators’ length vs passive (R/R₀ = 1) and active force with different actuation levels. (C) The relationship between the actuators’ length and the rotational angle of the eyeballs is compared between theoretical calculations and experimental measurements. (D) Rotation angle of the artificial eyeball as a function of resistance, demonstrating precise control of angular displacement. Scale bar: 1 cm. (E) Full-range eyeball rotation corresponding to the maximum actuation strain of the CoilLCE actuator.

小结

LCE 具有许多令人称道的特性，包括可编程性、可逆的形状变化以及高驱动应变。然而，与形状记忆合金（SMAs）或气动执行器相比，LCEs在机器人领域的应用仍明显滞后。后者在机器人应用中更为成熟，但在单向驱动行为、系统复杂性或便携性方面往往存在取舍。迄今为止，LCE的大部分演示都侧重于编程介晶取向以预先确定变形，例如在可重构或可展开表面等形态变化结构中，或者在小型独立机器人中，其中执行器本身提供简单的爬行或行走运动。LCE研究主要停留在材料层面，而能够实现与机器人技术集成的可操作控制策略和接口仍未得到充分探索。

在本研究中，我们通过提供将LCE致动器应用于机器人系统的实用框架，弥合了材料与机器人技术之间的鸿沟。我们没有将LCE仅视为活性材料，而是开发了具有明确性能校准曲线和实用控制方案的复合纤维致动器。通过嵌入铜线圈并利用其随温度变化的电阻特性，我们建立了一个能够进行精确、稳定调节的闭环系统，从而实现按需驱动。石墨烯的加入进一步提高了加热均匀性，并实现了驱动应变与应力的平衡，从而支持对抗性配置。基于这些要素，我们展示了两个类人机器人演示：一个手指模型，展示了对驱动应变和速度的连续且富有表现力的控制；以及一个人工眼肌系统，其中四组纤维协同工作，以实现可控且可重复的拮抗驱动。这些成果共同将基于LCE的致动器从可编程材料提升为系统化、可集成的组件。

展望未来，我们的方法为开发热可寻址驱动平台提供了见解。其灵活性体现在能够调节模具尺寸、线圈电阻和致动器几何形状，使其能够适应广泛的结构设计和材料选择。相同的基于电阻的温度传感和反馈控制技术可轻松扩展至其他热激活材料。未来的研究还可侧重于优化控制逻辑和电路架构，以实现更智能、更紧凑的运行，同时改进基于LCE的致动器的热管理和能效，从而拓宽其实际应用范围。

文献：https://doi.org/10.1002/adfm.78054