A Soft Enveloping Gripper with Enhanced Grasping Ability via Morphological Adaptability

Hitherto, automated grasping with robotic grippers requires adjusting the posture and force of the fingers based on the size, geometry, stiffness, and pose of the objects. To provide a simpler but efficient grasping methodology, a soft enveloping gripper is presented and investigated how its morphological adaptability improves the grasping ability by comparing its performance with fingered grippers. Results show that this enveloping gripper can omnidirectionally envelop objects via active–passive interaction, which allows the gripper to 100% grasp the object located at different positions within range and keep their orientations. However, the grasping success rate and orientation error of the fingered grippers highly depend on the relative position and angle of the objects to the grippers, as well as the number of fingers. The dynamic vibration and decay time of the enveloping gripper when grasping a 500 g weight are, both, approximately one‐sixth of those of the two‐fingered gripper when grasping a 12.37 g cube. This enveloping gripper can automatically grasp objects (including deformable ones) lying in different poses without posture estimation and force control with a simple vision‐based automatic grasping method. The enveloping grasping method may open an avenue for simple, low cost yet powerful automatic grasping applications.

DOI: 10.1002/aisy.202200456 Hitherto, automated grasping with robotic grippers requires adjusting the posture and force of the fingers based on the size, geometry, stiffness, and pose of the objects. To provide a simpler but efficient grasping methodology, a soft enveloping gripper is presented and investigated how its morphological adaptability improves the grasping ability by comparing its performance with fingered grippers. Results show that this enveloping gripper can omnidirectionally envelop objects via active-passive interaction, which allows the gripper to 100% grasp the object located at different positions within range and keep their orientations. However, the grasping success rate and orientation error of the fingered grippers highly depend on the relative position and angle of the objects to the grippers, as well as the number of fingers. The dynamic vibration and decay time of the enveloping gripper when grasping a 500 g weight are, both, approximately one-sixth of those of the two-fingered gripper when grasping a 12.37 g cube. This enveloping gripper can automatically grasp objects (including deformable ones) lying in different poses without posture estimation and force control with a simple vision-based automatic grasping method. The enveloping grasping method may open an avenue for simple, low cost yet powerful automatic grasping applications.
scalability. They can be fabricated by casting [16] or direct 3D printing. [17] Their kinematic dexterity can be enhanced by increasing the chambers [18] or programming the bending sections. [19] Their load capacity can be improved via variable stiffness mechanisms [20][21][22] or rigid skeletons. [23,24] In addition, their universality can be raised by adjusting the grasp postures [25] or the active palm. [26] However, as with rigid grippers, the grasping implementation of these soft grippers is also highly dependent on the posture, stiffness, and geometry of the objects, as well as the relative pose of the gripper to the objects, which is the inherent limitation of the fingered gripper.
Except for the human-finger-inspired fingered grippers, some researchers are devoted to developing other grasping mechanisms to further simplify grasping tasks while increasing the universality. For example, the particle jamming-based universal gripper can adapt to objects with multiple shapes by vacuuming. [27] The origami "Magic-ball" gripper can grasp many kinds of objects by folding the waterbomb origami via a vacuum. [28] The programmed kirigami sheet can grasp deformable objects by simple stretching. [29] The bloodworm-inspired gripper can "swallow" various objects via passive form-fitting. [30] The envelope gripper can grasp various objects with the expansion and tightening of the membrane. [31] The geckoinspired elastomeric microfibrillar membrane can adhere to various 3D surfaces with simple normal pressure. [32] Our group also developed a multimodal, enveloping soft gripper that can morphologically adapt to a variety of objects through different modes actuated by a single air vent. [33] Compared to the fingered grippers, these grippers implement grasping by adaptively enveloping the objects. Therefore, they do not need to adjust their configurations or postures based on the pose and geometry of the objects. Nevertheless, most of these efforts focused on introducing the design and demonstrating the universality of the grippers. Profound investigations on why these grasping mechanisms can improve the grasping performance from the perspective of static, dynamic, and practical grasping application viewpoints have not been reported yet.
In this work, we investigate how the morphological adaptability of our enveloping gripper can improve the grasping ability by comparing the essential differences between our enveloping gripper and soft-fingered grippers. To this end, we prepare two soft grippers (one with two fingers while the other with three fingers) with the same material, wall thickness, and length as our enveloping gripper. The distance from the center of the gripper to the fingers is also the same as the radius of our enveloping gripper to ensure a similar enclosing space. Then, we compare the grasping mechanism of these two kinds of grippers via finite element (FE) simulation and grasping experiments. We also compare their permissible position errors and orientation errors via static experiments, and their dynamic stabilities via dynamic disturbance. Finally, we demonstrate that with only the position information provided by 3D vision, our enveloping gripper can grasp both soft and rigid objects lying in various poses without any posture estimation or force feedback control, as long as their size and weight are within the grasping limit.

Grasping Mechanism Comparison
Distinct from common soft-fingered grippers, which grasp objects by actively bending the fingers, our enveloping gripper can implement grasping by morphologically adapting itself to objects of various shapes and sizes, with an active-passive interaction mechanism endowed by the unique design of the gripper. The primary element of the gripper is a soft accordion structure, with three fasteners to seal and fix it ( Figure 1A). The operating principle of the gripper could be found in our earlier publication. [33] In short, when deflating or inflating the chambers of the accordion structure, the gripper actively deforms to touch the surface of the object. Upon touching, the gripper passively adapts to the morphology of the object until it fully envelopes the target. With a simple open-close operation, the gripper can grasp objects ranging from 10 to 50 mm, and weighing up to 2 kg.
We simulated and experimentally evaluated the grasping motion of our enveloping gripper and a common two-fingered soft gripper on three representative types of objects (a cube with a flat surface, a sphere with a convex surface, and a column with a concave surface) to study their different grasping mechanisms. The dimensions of the grippers are shown in Figure S1, Figure 1. Grasping mechanism comparison of the enveloping gripper and a common two-fingered soft gripper. A) The unique design of the enveloping gripper that produces the active-passive interactive enveloping of the gripper. B) The simulation and experimental results show that the enveloping gripper can wrap most areas of the cube with a flat surface, the sphere with a convex surface, and the column with a concave surface. C) The simulation and experimental results show that the two-fingered gripper can only contact these three types of objects with its tips, the contact area is much smaller than that of the enveloping gripper.
www.advancedsciencenews.com www.advintellsyst.com Supporting Information. Both the FE simulation and experimental results in Figure 1B show that the enveloping gripper can fully wrap the most area of the cube and the sphere. A small gap exists between the gripper and the upper area of the two objects because the base of the gripper is secured by the fasteners and, thus, cannot shrink as other sections under the vacuum pressure. For the column, the enveloping gripper can fully squeeze the whole area of the side concave surface, which is impossible to achieve for the fingered grippers, including the soft and rigid ones. These results indicate that the enveloping gripper can omnidirectionally envelop various objects with the active-passive interaction mechanism, contributing to a bigger contact area, a larger holding force, and higher grasping stability. In contrast, for the fingered gripper, only the tips of the gripper can touch the three objects during grasping ( Figure 1C), which results in a much smaller contact area, a smaller grasping force under the same pressure, and a lower stability under disturbing force pointed outside the contact area, comparing to that of the enveloping gripper. The FE simulation video (Supplementary video 1, Supporting Information) demonstrates that the contact area between the enveloping gripper and the three objects increases continuously with the increase of the applied negative pressure while the contact area between the fingered gripper and the objects shows a reverse trend. High-speed video observation of the fingered gripper (Supplementary video 2, Supporting Information) also proves that the angle between the finger and the object increases with the increase of the actuating pressure, leading to the separation of the contact surfaces of the finger and the object. Therefore, even if a higher air pressure could increase the picking force for the fingered gripper, it also decreases the contact area, which may cause great difficulty for the force control of the fingered gripper. When the fingered gripper grasps a specific object, a particular air pressure is needed to exert enough lifting force while keeping a suitable shape closure. Whereas the enveloping gripper only needs an open loop control, then its body can interact with the object via its morphological intelligence. In brief, compared to the bending-induced grasping mechanism, the active-passive enveloping mechanism of our enveloping gripper can ensure an omnidirectional force closure, a bigger contact area, and a larger picking force, which may enhance the static and dynamic grasping abilities as proved by the next experimental results.

Permissible Position and Orientation Errors
For fingered grippers, the successful grasping needs a relatively accurate position and posture guidance to ensure that their fingers can touch the relevant points on the object to form force closure or form closure, which challenges the feedback control based on the external sensors, especially for the visual systems which usually have a relative error around 1-2 cm. In contrast, our enveloping gripper can ignore the detailed geometric information of the objects and readily grasp them if only they are within the grasping range because the gripper can approach and envelop the objects from all sides. Static experiments were conducted to compare the permissible position errors of the enveloping gripper and fingered grippers (See Experimental Section for the details). Results clearly show that the enveloping gripper can 100% pick up the cylinder no matter where it is located relative to the center axis of the gripper (Figure 2A). For each of the 31 positions at different displacements and angles, the grasping success rate is 100%, as the gripper can fully envelop the cylinder provided the cylinder is within the grasping range. However, for the two-fingered gripper, the success rate is highly dependent on the position of the cylinder. As Figure 2B shows, the cylinder can be grasped by the two-fingered gripper only when it is in the space between the two fingers. Outside the space, the success rate is nearly zero. Inside the space, the success rate is 100% only when the cylinder is near the center or on the mid perpendicular of the two fingers, then decreases rapidly when the cylinder is farther from the mid perpendicular. Because when the cylinder is placed farther, the two fingers will asynchronously touch it and the first touched finger will probably push it out of the grasping range.
We also tested the permissible position error of a threefingered gripper. Results show that the success rate is also highly dependent on the position of the cylinder relative to the fingers (see Figure S2A, Supporting Information). The success rate is nearly 100% when the cylinder is in the central area of the three fingers or close to one of the fingers, then decreases dramatically when the cylinder is positioned in the gap between the fingers because it is out of the moving range of the fingers. Compared to the two-fingered gripper, the three-fingered gripper can successfully pick up the cylinder at more positions. However, as Figure S2B, Supporting Information shows, at many of these positions, the posture of the cylinder in the gripper after picking up is in an unstable state. Even in the same position (at 144°on the 10 mm circle), the posture of the cylinder also varies with the actuation pressure ( Figure S2C, Supporting Information). There is a great probability that the object will fall immediately if the gripper is subject to external disturbance or moves at a higher speed, or if the object is a little heavier or the actuation pressure is a little bigger. Therefore, a successful picking-up task of a fingered gripper needs to comprehensively consider the position of the target, the configuration of the fingers matching the object, and the actuation forces. While for our enveloping gripper, the only consideration is if the size and weight of the object are in the range as its permissible position error is improved a lot by the unique grasping mechanism design.
The placed object could keep the same posture as prepicking up is important for some precise grasping tasks. We performed static experiments to test the orientation errors of the two kinds of grippers on two different objects. The orientation error was defined as the angle between the axes of the object before picking up and after placing, see the Experimental Section for the detailed experiment process. For both the cube and the triangle, the orientation errors of the enveloping gripper are nearly zero no matter how many degrees the gripper rotates relative to the objects (see Figure 2C,D). Because for any graspable object, the gripper can morphologically deform to the same shape as it, plus the omnidirectional enveloping eliminates the internal unbalanced forces which could reorient the object to a new balance state. Therefore, the enveloping gripper could keep the original posture of the objects after placing them. Whereas the orientation error of the two-fingered gripper highly depends on the shape of the objects and the relative posture of the gripper to the objects (see Figure 2C,D). As shown in Figure 2C, the orientation error for the cube increases along with the minus direction when the gripper is gradually rotating from 0°to 30°. At 45°, the gripper cannot pick up the cube. From 60°to 90°, the orientation error gradually decreases to approximately 0°from the other direction. In Figure 2D, the orientation errors for the triangle at different angles are relatively small except at the angles of 30°, 45°, 90°, and 105°, where the gripper cannot pick up the object. The reason caused these results is that the fingered gripper has an unchangeable configuration, if the fingers are not at a suitable angle, they could inescapability push the object to an undesired direction while they touch it. This could be further explained by the diagrams at the bottom of the figures, which show the gripper angles relative to the objects. For example, when grasping the cube at 45°, the gripper can only touch limited points on the two edges of the cube, it is difficult to keep the cube stable in 3D space. Moreover, when grasping the triangle at 30°, 45°, 90°, and 105°, the resultant force of the fingers will push the object away from the gripper, and, therefore, yield a failed lift. For the same reason, the orientation errors of the three-fingered gripper also depend highly on the gripper angle when it grabs the triangle (see Figure S2D, Supporting Information), but the errors are within 2°for all the gripper angles when it grabs the cube so we did not show them in the figure. From the orientation error tests, we could conclude that the fingered grippers need an appropriate angle to keep the posture of the grasped object during grasping while the enveloping gripper does not need. Therefore, it requires no time and effort to develop complicated www.advancedsciencenews.com www.advintellsyst.com algorithms for the enveloping gripper to grasp different objects, as the gripper can accomplish different precise pick-place tasks with a fixed posture.

Dynamic Stabilities
The high stability of the gripper to resist sudden unpredictable disturbance during high-speed movement is indispensable to improve work efficiency and increase benefits. We evaluated and compared the dynamic stability of the enveloping gripper and fingered grippers (see Experimental Section for the detailed experimental process). During the tests, the grippers (moving at different speeds and with the object in hand) were suddenly stopped, then the vibration displacement of the object was recorded by a high-speed camera. Figure 3A shows that the vibration displacement of the cube in the two-fingered gripper is quite big under different velocities at the actuation pressure of 20 kPa. At 1.2 m s À1 , the initial vibration displacement is approximately 6 mm, then gradually diminished with time. With the increment of speed, the initial vibration displacement increases to 10 mm. Namely, the dynamic stability of the gripper descends obviously with the increase in speed. The collateral damage is that we need to limit the speed and work efficiency of the gripper in real applications. In addition, for all velocities, it took approximately 7 s for the amplitude to reach zero, which is relatively a long time for real-time grasping tasks. By increasing the actuation pressure to 30 kPa, the vibration amplitude and frequency of the object, as well as the time required to decay could be reduced (see Figure 3B). This indicates that the dynamic stability may be improved by increasing the actuation pressure. However, a higher pressure could also cause the problems highlighted in the static tests. It is worth emphasizing that the dynamic stability of the gripper is also affected by its moving direction. As shown in Figure 3C, the vibration amplitude of the cube is significantly diminished if the moving direction is changed by 90°. The decay time is also reduced to within 1 s. Supplementary video 3, Supporting Information, also compares the real-time vibration of the gripper in the two directions. Similar to the two-fingered gripper, increasing the actuation pressure of the three-fingered gripper could also decrease the vibration amplitude; but the initial vibration displacement (approximately 1.5 mm) and the decay time (approximately 2 s) of the three-fingered gripper are, respectively, 0.15 and 0.25 times of that of the two-fingered gripper when the velocity is 3.6 m s À1 and the pressure is 20 kPa (see Figure S3A, Supporting Information). The vibration tendencies in different moving directions are similar ( Figure S3B, Supporting Information). Hence, increasing the number of fingers could increase the dynamic stability of the gripper to a certain extent. When the enveloping gripper grasps the object under the actuation pressure of À20 kPa, the amplitudes at different velocities are almost zero (see Figure 3D), which is negligible compared to those of the fingered grippers. It evidences that the dynamic stability of the enveloping gripper significantly outperforms that of the fingered gripper. The vibration motion recorded by the highspeed video can also prove this point (Supplementary video 4, Supporting Information), which clearly shows that the twofingered gripper returns to the static state after a long time, while the enveloping gripper shows no vibration when it is stopped. The reason behind the results is that the intrinsic gripping mechanisms of the two grippers are different. The fingered grippers implement grasping by bending the fingers, thus 1) they can only touch a limited part of the object and other parts may still move under the action of inertial force; 2) the direction of the resultant force of all the fingers have a limited range, so these grippers cannot resist forces from all directions. In contrast, our enveloping gripper grasps objects by actively enveloping and passively www.advancedsciencenews.com www.advintellsyst.com adapting, so 1) it can fully constrain the motion of the object from all sides and the object has no freedom of motion; 2) by interacting with the object, it can generate force in any direction against inertial moving. To further demonstrate the dynamic stability of the gripper, we tested the vibration of the gripper while it grasps a 500 g weight under different driving pressures at the velocity of 3.6 m s À1 (Supplementary video 5, Supporting Information). The initial vibration displacement is approximately 1 mm ( Figure 3F), which is smaller than that (1.5 mm) of the three-fingered gripper, and much smaller than that (6 mm) of the two-fingered grippers, although the weight is increased by nearly 40 times. The decay time is approximately 0.5 s (Figure 3F), only one-fourth of that (2 s) of the three-fingered gripper and one-sixth of that (3 s) of the two-fingered gripper. The vibration trajectories at different pressures are similar ( Figure 3F) because of the buckling effect of the chambers, [33] which proves that the driving pressure is not a control variable for the gripper, and the gripper can be operated under a fixed pressure with open loop control. The excellent dynamic stability of the enveloping gripper demonstrates its potential for deployment in high-speed operating scenarios, without special requirements for trajectory planning or force control to prevent failure. It should be noted that the weight of the cube is only 12.37 g because it is difficult for the two-fingered gripper to pick up heavier objects and it is true that the heavier the object, the more violent the vibration.

Automatic Grasping Demonstration
Attributing to the unique active-passive interaction enveloping mechanism, our enveloping gripper can accomplish automatic grasping tasks in a simpler but more reliable way compared to the fingered grippers because 1) the gripper can grasp any objects in a fixed pose if only they are within the grasping range due to its big permissible position error and nearly zero orientation errors, the complicated algorithms to estimate the feasible gesture of gripper based on the pose and geometry of the object is unnecessary; 2) the gripper can work at a fixed actuation pressure without force feedback and control as its gripping force and dynamic stability do not depend on the driving pressure; 3) the gripper is dynamically stable in all directions and, therefore, it does not need special trajectory planning to avoid object falling during high-speed motion. We develop a vision-based automatic grasping method for the gripper to demonstrate its adaptability and simplicity. For the method, we first use the common SIFT algorithm to recognize only the position of the object based on the Kinect V2 3D camera, then the gripper automatically picks up the object in a fixed pose and driving pressure. We validate the method by using the gripper to grasp two ornaments in different poses. Whether the dogshaped ornament is in a standing, side-lying, or face-up state, the gripper can fully envelop it with a fixed gesture, then successfully grasps it (see Figure 4A and Supplementary video 6, Supporting Information). The gripper can also successfully grasp the ratshaped ornament placed in different poses (see Figure 4B and Supplementary video 7, Supporting Information). For both of the ornaments, the only information needed for the gripper is their positions in 3D space, their shape and posture information can be ignored. However, it is indispensable for the fingered gripper to replan the grasping posture whenever the shape or the pose of the object changes. [34] Therefore, the enveloping gripper can significantly simplify grasping tasks.
With the morphological adaptability, the enveloping gripper can also grasp deformable objects without force feedback. We prepared three deformable objects, the elephant-shaped rubber boy, the marshmallow, and the jelly, which can be easily flatted by pinching, to demonstrate this ability. For each object, the gripper can securely envelop and then lift it without causing any damage (see Figure 5 and Supplementary video 8, Supporting Information). During grasping, these objects start to deform as the gripper squeezes them, but the squeezing stops when the chambers of the gripper enter the buckling state. As a result, the objects are not damaged even without force feedback control. Furthermore, when the objects deform, the gripper can adaptively deform to keep enveloping them, which generates a geometrical interlocking between each other to guarantee stable grasping of the objects. By contrast, it is difficult for the fingered www.advancedsciencenews.com www.advintellsyst.com gripper, especially the rigid grippers to grasp deformable objects without force feedback because the grippers need to adjust the fingers to exert proper force on the objects to hold but not damage them whenever they deform. Although some soft-fingered grippers can grasp deformable objects, [16] they require proper driving pressures so that the objects do not slip when overpressed.
Compared to the fingered grippers, the enveloping gripper has some shortages. For example, its grasping range is smaller than the fingered gripper under similar conditions. The fingered gripper can achieve some manipulation tasks such as rotating a pen in hand if they have more degrees of freedom, whereas the enveloping gripper only excels at grasping tasks. Fingered grippers can grasp very long objects, like a pen, by pinching while the enveloping gripper cannot as they are out of the enveloping range. Nevertheless, the current enveloping gripper is intentionally designed for grasping tasks and its capabilities can be further improved by combining with other technologies, such as electroadhesion.

Conclusion
In summary, this paper investigates how the soft enveloping gripper can improve its grasping ability while simplifying grasping tasks by comparing its grasping mechanism, static permissible errors, and dynamic stabilities with the soft-fingered grippers. The FE simulations and experiments reveal that our enveloping gripper can morphologically adapt to objects with various curved surfaces via the active-passive interaction enveloping mechanism, which could omnidirectionally exert force on all sides of the objects. While the fingered gripper can only touch limited parts of the objects. The essential difference results that our enveloping gripper can successfully grasp objects within its grasping range and maintain their orientation after placement. While the success rate and orientation error of the fingered gripper depend on its relative position and angle to the object. Furthermore, the dynamic stability of our enveloping gripper also surpasses the fingered grippers. The initial amplitude and decay time of the vibration after the enveloping gripper, with the 500 g weight in hand, suddenly stopped is smaller than that of the three-fingered gripper with a 12 g cube in hand. With the simple vision-based automatic grasping method, we show that the enveloping gripper can grasp objects in different poses, even deformable objects, in a fixed posture and driving pressure. The great morphological adaptability of the enveloping gripper could improve the ability of robotic systems to meet the requirement of great economic and societal importance in many applications, such as food processing, logistical sorting, and pick-and-place sorting of heterogeneous objects on an assembly line.

Experimental Section
FE Simulation of the Grippers: The grasping simulation of the two kinds of grippers was analyzed by the commercial software Abaqus (Abaqus 2021, Dassault Systems Inc., France).
The material of the grippers was approximated as the incompressible hyperelastic Yeoh model considering its nonlinear property while the grasped objects were modeled as 3D discrete rigid bodies to reduce the computation cost. For the interaction property, the inner chambers of the gripper were set as self-contact, and the contacting surface between the grippers and the objects was set as surface-to-surface contact. The contact interaction property for both sets was molded as the "Hard" contact normal behavior and the penalty tangential behavior with the friction coeff. 0.6. The PINNED boundary constraint was applied on the base of the grippers and the reference point of the rigid body (the grasped objects) to restrict the freedom of movement in 3D space. The negative/positive pressure was set separately as the load on the surfaces of the chambers of the enveloping gripper/fingered gripper. The grippers were meshed by the hybrid element type C3D10H. Finally, the models were analyzed by the dynamic, implicit solver considering the complicated dynamic interaction during grasping.
Static Grasping Experiments: For the permissible position error test, we prepared a meshed circular grid, which has three concentric circles spaced by 5 mm and 10 radial lines spaced by 36°, so a total of 31 positions including the center of the circles. The central normal of the grid is aligned with the center of the grippers. During the test, we successively placed the 3D printed cylinder (weight 12.37 g, width 28 mm) at the 31 positions, then controlled the grippers to pick up it. The grippers were fixed to the end of the robot arm, whose speed is 0.2 m s À1 to eliminate the dynamic influence. We tried 50 times for each position and then calculated the success rate afterward. For the orientation error test, we prepared two 3D printed objects: the cube (weight 12.37 g, width 28 mm) and the triangle (weight 8.52 g, perpendicular bisector length 28 mm), then controlled the grippers to pick up the objects in different gripper angles. For the cube, the gripper angle changed from 0°to 90°with an interval of 15°. For the triangle, the gripper angle changed from 0°to 120°with an interval of 15°. For each gripper angle, the gripper first picked up the objects and then placed them in the same position. The deflection angle after placement was measured as the orientation error.
Dynamic Stability Tests: For the dynamic stability test, the grippers were attached to the end of the robot arm (Siasun GCR5, Siasun, China). They were moved to a specific place to pick up the objects, then moved up. After 1 s suspension, they were moved at different speeds (1.2, 2.4, 3.6 m s À1 ) and then suddenly stopped. The vibration of the object after the gripper stops moving was recorded by the high-speed video camera (Phantom VEO 410, Vision Research Inc., America) at the frequency of 1,000 Hz. Then, the displacement data were obtained from the videos via Tracker (a free and open-source video analysis software). www.advancedsciencenews.com www.advintellsyst.com