(auxetics). Mark et al. [39] provided an excellent example using auxetic and nonauxetic clutches to simplify the locomotion of a soft robot with only one actuator instead of three [Figure 4(d)]. More generally, auxetic metamaterials may endow a soft robot with the capability of shape matching upon actuation using cellular structures, which consist of auxetic and nonauxetic units [49]. Elastic beam elements may buckle when subjected to axial compressions. This simple phenomenon opens up a new avenue for reversible pattern transformations in metamaterials that consist of networks of elastic beams. This mechanism was exploited by Yang et al. [40] in soft grippers to produce several classical motions driven by a single negative pressure [Figure 4(e)]. To improve structural stiffness and enhance grasping force, the design was further improved in [50], where the output work was taken as the objective function. More generally, a metamaterial derives its properties by en-- coding its constituent microstructures [5]. Schumacher et al. Origami and kirigami represent a special category of metamaterials that lend themselves to programmable morphing of robots. Origami-based metamaterials are usually made by folding thin-walled sheets along predesigned creases to form ridges and valleys [46]. Driving rigid origami by vacuuming, artificial muscles were developed in [47] for use in soft grippers with excellent load capability. Jeong and Lee employed an origami twisted tower to fabricate the fingers of a robotic manipulator [Figure 4(b)], which potentially can be used to manipulate fragile objects [37]. Rafsanjani et al. [38] harnessed kirigami principles to remarkably improve the crawling speed of a soft actuator [Figure 4(c)]. The deformable kirigami surfaces buckle and induce remarkable directional frictional properties. Readers may refer to [48] for a comprehensive review of soft origami robots. Slender beams are widely used as basic units in flexible metamaterials. The designable arrangements of elastic beam elements may lead to desired mechanical behaviors that are otherwise difficult to achieve, such as negative Poisson's ratio Width Face Loop Back Loop Length Length Warp Weaving Loop Warp Weaving Length Weft Weaving Width Width 1) Woven Fabrics 2) Warp Knitted Fabrics Rib-Weft Knitting 3) Rib-Weft-Knitted Fabrics (a) (b) Supplied Volume Vmax Volume-Control Actuation Protocol Time Rough Surface (Foam) (c) Microstructure Material Space Optimization Sampling Preprocess: Metamaterial Family Construction (d) Database: Metamaterial Space (e) Input Microstructure Tiling Material Synthesis Optimization Parameters Run Time: Synthesis Output Printable Model (f) Figure 4. The metamaterial optimization. (a) Textile fabrics produced by weaving and knitting methods (left) are used to simultaneously program the motions and increase the load capabilities of a soft wearable assistive glove (right) [36]. (b) An origami twisted tower to fabricate the fingers of a robotic manipulator [37]. (c) An efficient crawling robot benefits from kirigami surfaces wrapped around an extending soft actuator [38]. (d) Auxetic and nonauxetic clutches simplify the locomotion of a soft robot with only one actuator [39]. (e) Elastic beams buckle under negative pressure and enable a soft gripper [40]. (f) An automatic design strategy of microstructures for physical prototypes with desired mechanical properties. Taking specified material parameters as the input, the method automatically optimizes the local microstructures that generate the target deformation behavior [41]. DECEMBER 2020 * IEEE ROBOTICS & AUTOMATION MAGAZINE * 33