Rigid robots can achieve precise motions but expose shortcomings in system complexity, fabrication cost, and humanrobot interaction, which motivates researchers to develop various soft robots to fill these gaps. Electro-hydraulic actuators (EHAs) have received widespread attention and been used in many soft robots due to impressive high-strain, fast-speed and rapidresponse characteristics. However, existing EHAs face challenges in achieving large-deformation, high-robustness, and low-weight simultaneously. This limits the application of EHAs in robotic systems that are weight-sensitive or require fail-safe and faulttolerant behavior. Here, we present a lightweight (0.98 g) electropneumatic actuator (EPA) filled with air and only 0.1-mL liquid dielectric, which achieves high-speed bending from 11° to 93.5° in 60 ms, large-angle bending from 11° to 104° in 2 s (the largest in current EHAs), and high-frequency swing at 20 Hz. The EPA is ultrarobust and can operate properly after being punctured by four needles or crushed twice by a 1500-kg vehicle. Furthermore, to validate the above features of EPAs, three applications are demonstrated at a voltage of 6 kV, including four-finger grippers, fast-crawling robots, and water-walking robots. This work pushes the boundaries of robustness and lightweight for EHAs, providing a foundation for the application of electro-pneumatic actuation in soft robotics.

This paper presents a novel synthesis method for the design of micro-scale robotic flexure mechanisms. A structural optimization method, termed the mechanism-based approach, is used to identify the optimal topology and shape of the flexure mechanisms based on their lump stiffness characteristics. Using several different fitness functions, several optimal flexure designs have been synthesized for use in millimeter-scale mobile grippers (μ-grippers). The stiffness characteristics of the optimal μ-grippers are shown to be better than the thin-beam designs developed using human intuition. Two large-scale prototypes are constructed and experiments are conducted to validate the stiffness analysis. The experimental results are within 20% of the analytical expectations. As a proof of concept, at-scale μ-grippers are constructed based on photolithography and replica molding methods, and demonstrated in simple actuation. The optimal μ-grippers can be applicable for cell manipulations in future works.