Analysis of Contact Force and Shape Change on Grasping a Square Object Using an Actual Fin Ray Soft Gripper

In recent years, robots that perform tasks in place of humans, such as cooperative robots, have been attracting increasing attention in the field of high-diversity, low-volume production. Research and development have been progressing accordingly. One of the important functions of such robots is grasping, and gripper research and development has been actively conducted. Major gripper configurations are rigid-link, multi-degree-of-freedom motor structures that realize stable grasping of a wide variety of objects through precise angle control based on sensor feedback. However, such grippers are expensive to control and require time-consuming mounting work each time the gripping object is changed. In contrast, with the development of 3D model design and 3D printer fabrication, soft grippers that realize stable grasping with simple control, assuming flexible materials and motor structures with low DOF, are attracting attention [1]. There are many soft grippers with various flexible structures, such as Soft Robotics’ mGrip [2] and Empire Robotics’ VERSABALL^® Gripper [3]. Soft grippers can be classified into two main types. One is the pneumatically driven soft gripper, which allows compliant wrapping and a grasping motion due to the compressibility and fluidity of air; Dilibal et al. [4] and Sun et al. [5] have proposed this type of soft gripper. The second is a soft gripper that mimics a biological system. Soft grippers, which mimic biological systems, are known for their unique flexible structure and functionality to realize stable grasping. These functions include the ability to passively deform with the force generated by contact with an object and to adapt to the shape of the object [6]. Such functional structures not only reduce the control burden, but also enable the stable grasping of objects of different sizes within a certain range with the same soft gripper. In addition, it has been found that changing the angle and thickness of the structure of a specific part further contributes to the stable gripping of objects in a wide range of different shapes. These advantages are thought to reduce the time cost of frequent gripper mounting operations in a high-mix, low-volume production, and the gripper is attracting attention as an effective gripper for recent production sites. A representative example is the Fin-Ray-type soft gripper (FRSG), which uses the Fin Ray^® effect [7] proposed by Leif Kniese et al. in 1997. The Fin Ray^® effect is a bio-mimetic structural mechanism (Fin Ray structure). It is based on the structure of fish fins, and consists of a front beam and back beam, which are the oblique sides of a triangle, and a crossbeam connecting them. When the front beam receives a contact force from the grasping object, the internal force is transmitted to the crossbeam, and the entire structure is passively deformed by wrapping in the direction of the grasping object. In other words, the passive grasping motion is realized by using two Fin Ray structures to clamp the grasping object. The Fin Ray structure has been released as an open-sourced design [8,9], and many researchers and developers have customized the Fin Ray structure according to the target grasping object. For example, Hemming et al. [10] developed a robot for harvesting agricultural crops such as green peppers, and realized the grasping of various types of peppers with different sizes and shapes. However, to introduce the most efficient FRSG for a production site, it is necessary to analyze whether the structure is effective for the grasping object. For this reason, finite element method (FEM) analysis has been the mainstream method used in conventional research, especially in attempts to verify the grasping performance when the crossbeam structure is changed. For example, Elgeneidy et al. [11] reported the ability to change the stiffness of the Fin Ray structure due to contact between crossbeams by changing the angle and arrangement of the crossbeams. Manoonpong et al. [12] used a Fin Ray structure for the feet of a multi-legged mobile robot and verified the angle at which it becomes stable on each uneven terrain by changing the crossbeam angle. This FEM has shown some degree of success in terms of demonstrating the grasping performance of the FRSG model in the simulation. However, the material properties and geometry of the actual 3D printed objects may not always be in perfect agreement with the FEM analysis model, and it is difficult to achieve perfect agreement in the grasping behavior. In addition, it is difficult to analyze buckling and slippage by FEM simulation. Therefore, we recognize that it is important to produce an actual FRSG using a 3D printer, in order to carry out analysis and evaluation, instead of using a simulation. In the grasping state analysis evaluation of an FRSG, it is important to analyze the relationship between the force and shape of the Fin Ray structure. There have been four types of evaluation methods used in previous studies using actual equipment. The first method involves building sensors into the Fin Ray structure: Zhou et al. [13] used piezoelectric elements, Zapciu et al. [14] used capacitance sensors inside the Fin Ray structure, and Hashizume et al. [15] used Gecko sheets and taxel sensor sheets inside the FRSG to analyze the force at each structural position. The second method is to place a flexible structure with feature points inside the Fin Ray structure, record its deformation with a camera, and use image processing to estimate the contact force state from the amount of deformation. Sandra et al. [16] placed a gel sheet with feature points and a small camera inside the Fin Ray structure and used image processing to estimate the contact shape and force by linearly deforming the gel sheet. Furthermore, GelSight Baby Fin Ray [17,18], which minimizes the sensor-unrecognizable area by devising the camera arrangement, has also been developed. However, these two methods are considered to affect the grasping characteristics of the FRSG, and the part with a built-in sensor is easily damaged. Furthermore, when multiple sensors are arranged, there is a problem in that the part that cannot be measured is generated. The third type is a method in which the force sensor is built into the grasping object. This method does not interfere with the grasping motion of the FRSG, and the sensor position and contact area can be changed, enabling flexible operation. Furthermore, it enables more accurate measurement and analysis than capacitive or piezoelectric methods. For example, Shan et al. [19] and Suder et al. [20] have analyzed 6-axis force applied to the center of a cylindrical object by dividing the object into two parts and installing a 6-axis force sensor in the center joint of the part. However, because this method analyzes the composite force applied to one central point, it has not yet demonstrated the wrapping and grasping characteristics of the FRSG with quantitative data. To quantitatively evaluate the wrapping grasp, analysis using a measurement system with high spatial resolution is necessary. The fourth method type is one in which the elastic properties of each part of the Fin Ray structure are known in advance, the amount of deformation is measured from camera image analysis, and force estimation is performed based on the relationship between the elastic properties of the specific part and the amount of deformation. Stuhne et al. [21] succeeded in estimating the internal force from camera image analysis of the side profile of a Fin Ray structure. However, this method is technically difficult because it requires a high-precision understanding of the elastic properties of the Fin Ray structure to achieve a highly accurate estimation. In contrast, Barrie et al. [22] and Deng et al. [23] constructed an elastic property model of the Fin Ray structure based on the actual device in an FEM simulation. This method numerically analyzes grasping with an object model under various conditions and machine learning is employed for the relationship between force and shape. Then, the shape of the object grasped by the actual FRSG is input to the machine learning system, and an operationally efficient system for estimating the force is realized. However, this FEM–machine learning method is currently applicable only under the precondition that stable grasping is possible with little buckling or slippage. In other words, there is no problem with cylinders, which are the basic object of the FRSG, but the grasping of objects with angles, such as a square object, results in a different behavior. This is because the shape and material properties of the FRSG and the object have a large influence on the behavior of the object, and it is important to evaluate the actual object instead of FEM analysis. Based on the above four previous studies, we believe that it is essential to establish an evaluation process for FRSGs to achieve efficient operation at high-mix, low-volume production sites. To realize such an evaluation process, it is necessary to use a system that combines a high-spatial-resolution force measurement device with multiple force sensors built into the gripping object and a camera image analyzer for measuring FRSG side deformation. In the previous study, we developed a 16-way object contact force measurement system and a structural deformation measurement system based on camera image analysis and analyzed quantitative data from a case of grasping a cylinder, which is the basic grasping object, using the standard FRSG open-sourced by Festo. The results show that the Fin Ray structure deforms passively and successfully exhibits a distributed contact force state that indicates an enveloping grasp on the object [24].

As the next step in this study, we will use these systems to analyze five types of enveloping grasping characteristics with different crossbeam angles, which are used in many applications of FRSGs. Furthermore, we will attempt to analyze not only cylinders, which are basic grasping objects, but also square objects with angles (two different arrangements), which are difficult to analyze properly with FEM simulations.

This chapter describes the configuration of built-in and contact force measurement systems for two types of grasping object for the proposed Fin-Ray-type soft gripper (FRSG) and a camera-image-based structural displacement measurement system.

In Experiment 1, five different FRSGs were used to analyze the grasping state of a cylindrical object. The results of cylinder object grasping are shown in four figures; Figure 5 shows the results of the situation in which a particularly stable grasp is achieved among the combinations of each grasping position and FRSG type, Figure 6, Figure 7 and Figure 8 show the contact force and structural deformation for each Fin Ray structure and for each of three different gripping positions (center of contact surface, 10 mm tip from center of contact surface, 20 mm tip from center of contact surface). First, from Figure 6, Figure 7 and Figure 8, the contact force distribution under each condition shows that the center of the contact surface at the grasping position has the widest contact force distribution. It is considered that the pressure is evenly applied to the surface of the grasped object to realize an enveloping grasp. The peak values of the contact force distribution when grasping at the center of the contact surface for the FRSG were 3.3 N (−30°), 3.6 N (−15°), 2.4 N (0°), 4.0 N (15°) and 2.3 N (30°) immediately after the beginning of contact. Furthermore, looking at the movement locus of the marker, it can be confirmed that it deforms most along the shape of the circle when grasped at the center of the contact surface. On the other hand, the position of the cylindrical object during grasping and the position where the object is released after grasping have different characteristics (Table 1, Table 2 and Table 3). In the FRSG (−30°) and FRSG (+30°), when the object is grasped at the center of the contact surface, it is pushed farther out than the starting point of the grasp. In other grasping positions, it tends to be pushed farther out than in the FRSG (−15°), FRSG (0°), and FRSG (+15°). In terms of contact force distribution, the FRSG (−15°), FRSG (0°), and FRSG (+15°) are distributed over a wide range, while FRSG (−30°) and FRSG (+30°) are distributed over a narrow range, which cannot be considered as a wrapping grasp. In the FRSG (−30°) and FRSG (+30°), when the object is grasped at the center of the contact surface, it is pushed farther out than the starting point of the grasp. In other grasping positions, it tends to be pushed farther out than in the FRSG (−15°), FRSG (0°), and FRSG (+15°). In terms of contact force distribution, the FRSG (−15°), FRSG (0°), and FRSG (+15°) are distributed over a wide range. However, the FRSG (−30°) and FRSG (+30°) are distributed over a narrow range, which cannot be considered as a wrapping grasp. This may be due to the influence of the crossbeam angle. Therefore, we discussed the physical properties of each FRSG during grasping: from Figure 5a,c, the FRSGs (−30° and −15°) have a passive deformation function where the contact surface tends to wrap the grasped object within the contact. However, as the crossbeam angle increases in the negative direction, contact between the crossbeams is more likely to occur. In other words, the rate of entrapment increases, but the allowable range of contact deformation decreases. However, the contact between crossbeams causes buckling at the back of the FRSG, generating an S-shape. This prevents the Fin Ray structure from spreading to the outer left and right sides of the FRSG. As a result, the FRSG (−30°) results in a grasping state with a narrow contact area, as shown in Figure 5a. Furthermore, the outward deformation of the FRSG due to the S-shape deformation slides on the contact surface of the grasped object and is pushed forward. In contrast, the FRSG (+30°, +15°) is designed to suppress the crossbeam-to-crossbeam contact that occurs when it contacts the cylinder, as shown in Figure 5b. This allows the raised area near the base of the contact surface of the FRSG to be seen. This raised shape also pushes objects forward with the FRSG. On the other hand, the ideal result can be shown as the FRSGs (0°, 15°) shown in Figure 5c,d. Within these Fin Ray structures, no contact occurs between the crossbeams, the contact force distribution range is widest. It shows the enveloping grasp. And the FRSG deforms along the grasping contact surface, even near the tip of the FRSG; the FRSG is grasped without sliding due to friction from the start of the FRSG grasp until it is closed. How-ever, when the FRSG is opened, it opens from the tip of the FRSG. And it caused slippage via a decrease in the contact area. This prevents the FRSG from returning to its initial position, which is a problem. These misalignments of the object position after grasping need to be considered in the operation of the gripper.

In this experiment, the effect of the crossbeam angle of the FRSG on cylindrical object grasping was verified. The more negative the crossbeam angle is, the more the left–right deformation caused by contact between the crossbeams is suppressed. In addition, the more positive the crossbeam angle is, the more the raised buckling phenomenon at the base of the contact surface is confirmed. Furthermore, it was found that wrapping grasping was best achieved when the crossbeam angle was near 0°. This means that the crossbeam angle has a significant influence on the grasping condition. Countermeasures against object misalignment on the field mounting surface of the FRSG include (1) introducing a position sensor to detect the object position, (2) increasing the friction of the contact surface, and (3) assuming and controlling an error when releasing the object from the FRSG.

In Experiment 2, we analyzed the crossbeam angles that can realize stable grasping based on the grasping conditions of the corners of the FRSG in a square grasp, which is a common grasping object shape.

In the study of Fin-Ray-type soft grippers (FRSGs), researchers have developed a wide variety of FRSGs and evaluated their operation. In many studies, buckling and slippage were difficult to describe by FEM analysis, and contact analysis of angled objects was not consistent with the actual motion of the actual device. On the other hand, it has been reported that the grasping characteristics of FRSGs vary depending on the crossbeam angle, and it is necessary to verify this in the actual machine. Therefore, in this study, we analyzed the influence of the crossbeam angle on the grasping characteristics and the grasping state of an angled object to establish a guideline for determining the FRSG structure to suit the task. The analysis was carried out using a discarded body developed in a previous study by our research team, which is a combination of a contact force measuring device (type 16/16) with a force sensor built into the grasping object and a structural deformation measuring device using camera image processing. Five different FRSGs (−30°, −15°, 0°, +15°, and +30°) were prepared, differing only in crossbeam angle. Using these fingers, grasping experiments were conducted with three types of grasping objects: a cylinder, which is the basic shape of the object to be grasped, a square 0° configuration and a square 45° configuration, which is assumed to be used in daily life. As a result, it was confirmed that the FRSG (0°, −15°) performs a particularly extensive enveloping grasp for cylinders. However, the initial position of the grasped object tended to differ from the position after the end of grasping, indicating that countermeasures were necessary. Next, for the square (0°), the best results were obtained for the FRSG (+15°). These results indicate that for stable grasping of a square object, one Fin Ray structure must make two points of contact with the corners of the square object. Therefore, we also found that there are two important grasping behaviors: the first is that the FRSG raised buckling structure can contact the lower corners of the square body and maintain the crossbeam spacing. The second is that the tip of the FRSG is at the top of the square body and can create a wraparound shape change. Finally, the best results were obtained when the FRSG (0°) was square (45°): above the contact position between the contact surface of the FRSG and the corner of the square body, a wrapping grasp shape change is required and a crossbeam angle of −30° was used. Furthermore, below the contact position, it is appropriate to configure the crossbeam angle at +30° to prevent crossbeam-to-crossbeam contact. It was estimated that this would enable stable square wraparound grasping.