Finger Flexion and Extension Driven by a Single Motor in Robotic Glove Design

Pneumatic and tendon‐driven actuators are widely used in soft robotic glove design. Tendon‐driven robotic gloves are generally better in controllability, dexterity, and force output, but they are less comfortable than pneumatic ones. Most soft gloves focus on only one actuation mode where either motor‐driven tendon or pump‐driven pneumatic transmission is used. Herein, a double‐acting soft actuator (DASA) that provides both tendon‐driven flexion and pneumatic extension of fingers by a single motor is presented. This is achieved by a smart pulley and bellow system. The kinematic model of the tendon‐driven flexion and the torque model of the fabric‐based pneumatic extension actuator (FPEA) are developed to analyze the DASA performance. The bending angle of the index finger actuated by the tendon and the FPEA extension torque of a joint are characterized by experimental studies. A cycle test of the DASA is conducted 3000 times, demonstrating its high repeatability. A prototype soft glove (68 g) based on the proposed DASA with a control box (835 g) is fabricated to demonstrate finger flexion and extension assistance. Based on electromyography signals, the performance of the robotic glove is evaluated by a squeezing sponge test.

compared with robotic exoskeleton devices. The actuation of the tendon generally uses electric motors. Meanwhile, the tendon wire is guided through a tendon routing path to achieve proper finger flexion. The tendon routing paths are usually made of Teflon tubes (Polytetrafluoroethylene tubes), Velcro straps, or 3D-printed parts to generate specific routes. [21][22][23] The tendon-driven rehabilitation gloves have many advantages such as compact structure, lightweight, affordable, large bending force, and so on. However, the biggest problem with tendon-driven gloves is discomfort. This can be partly alleviated by proper tendon routing path and reasonable glove structure design.
Over the past decade, more and more rehabilitation gloves driven by soft bending actuators on the dorsum of fingers have been proposed to assist at-home rehabilitation. [24,25] These assistive gloves driven by soft actuators have infinite DOFs, meaning their design does not need to consider the misalignment problem. Compared with the tendon-driven robotic gloves, the bending motion of the assistive gloves based on such soft actuators is achieved by pushing fingers from the finger back instead of pulling from the finger front as in tendon driven. Pushing the finger from the back is more comfortable than the tendon-driven method. Furthermore, actuators placed behind fingers will not interfere with patients' finger flexion during grasping operations. The soft bending actuators for assistive gloves can be achieved by different principles like fluidic elastomer actuators (FEAs), inflatable soft bellow, and so on. [26][27][28] However, these actuators are too heavy and bulky. Thus, researchers have proposed fabric-based pneumatic actuators (FPAs) in the design of soft wearable assistive gloves. [29][30][31] The gloves driven by FPAs are more affordable, lighter, and easier to manufacture. But gloves based on FPAs are slow to respond due to the large inner chamber volume of the FPAs. In addition, these soft pneumatic actuators are driven by a bulky and complicated pneumatic system, and users must tolerate the noise and vibrations caused by the pump operation. [25] In practice, patients do not only care about the performance, comfort, and wearability of the robotic glove but also care about whether the glove is inconspicuous or not. For robotic gloves based on FEAs, FPAs, and soft bellows, the huge actuator size and noisy pump system make the user conspicuous. [26][27][28][29][30][31] Tendon-driven actuation method is an inconspicuous way to assist hand motions. However, finger extension assistance in previous tendon-driven glove research is achieved by passive springs or antagonistic structures on the dorsum of the hands. For the passive spring design, the extension torque passively increases with the finger flexion, which uncontrollably prevents finger flexion. [32] Cho's group proposed an ingenious dual slack enabling mechanism to achieve the antagonistic tendon-driven actuation, which requires a complex mechanical structure (1.14 kg). [18] In this article, a novel design with a single motor can provide both a tendon-driven actuator and a fabric-based pneumatic extension actuator (FPEA), defined as the double-acting soft actuator (DASA). The tendon is directly driven by a motor, and the FPEA pressure change is controlled by compressing bellow actuation (CBA). The main contributions are summarized as follows. 1) Tendon-driven finger flexion and pneumatic finger extension are driven by only a single motor. The finger extension driven by FPEA is safer compared with antagonistic tendon-driven actuations. 2) Even though compressed air is needed for finger extension, our proposed actuation is not tethered to a pump. 3) A new nested tendon routing path is proposed to avoid the trigger finger phenomenon, which can improve glove comfort. 4) Mathematical models are developed to represent the relationship between the finger flexion angle and the motor rotational angle, and this relationship is further evaluated by experimental results. In addition, we also investigate the effects of bending angle, pressure, and fabric material on the extension torque of the FPEA. And then, a finger flexion and extension cycle test was conducted 3000 times to prove the repeatability and reliability of the DASA design. 5) Finally, an inconspicuous robotic glove prototype (68 g) based on the DASA with a control box (835 g) is fabricated to demonstrate practical finger flexion and extension. To evaluate the performance of the assistive glove, we proposed and conducted a squeezing sponge experiment on a healthy subject monitored by an electromyography (EMG) system. The experimental result verifies that the robotic glove can restore the grip function of the subject's hand when the subject does not actively exert force, and it can enhance the subject's grip force when the subject's hand and the assistive glove are active simultaneously.

Mechanism of the DASA
For stroke patients, their stiff fingers are not only hard to bend but difficult to extend. To achieve both finger flexion and extension motion, we designed and proposed a novel design called the DASA, as shown in Figure 1. The glove can help patients to bend their fingers by tendon-driven actuation and extend their fingers by the FPEA. FPEA is made of two layers of TPU-coated fabric as shown in Figure 1a. The FPEA is fixed on the back of the finger, which can extend the finger when the air pressure in the FPEA increases. In the DASA design, the control of pressure in the FPEA is achieved by the CBA method, and the pumpless design avoids intolerant noise and vibrations from the normal pump system. In the tendon-driven part, the tendon routing path guide is made of Teflon tubes and arranged symmetrically on both sides of the finger across the back of the fingertips. Both tendon-driven actuation and FPEA are driven by one actuation system as shown in Figure 1. In Figure 1b, the motor output shaft connects two spools, one is for the tendon motion (in red color), and the other (in orange color) is for compressing the bellow. The tendon and cable are wound in opposite directions. That is, when the tendon loosens, the cable tightens onto the spool, and vice versa. When the cable is tightly wound on the spool, the bellow is compressed, as the finger full extension state in Figure 1b. The air inside the bellow is fully squeezed out and actuates the FPEA on the dorsum of the finger through an air tube, thus the finger will be straightened. In Figure 1c, the finger is in the full flexion state where the tendon is tightly wound to the spool and the cable is released, and the bellow returns to its uncompressed state due to its elasticity and pressure. In Figure 1d, the finger is in an intermediate stage where both the tendon and brown cable are partially wound on their respective spool. The finger bending angle is directly controlled by the motor.

Tendon Routing Path Design
Although the tendon-driven actuation method imitates the biological tendon in the human hand using a polyethylene (PE) wire, users still feel uncomfortable using the glove based on the tendon-driven method. The main reason for this discomfort is that the distance between the PE line and phalanges is much larger than the distance between the biological tendon and the phalanges because of the noninvasive medical procedure of the robotic glove. Constraining this distance to a smaller range will improve the glove's comfort. Tendons in human hands are constrained along phalanges and metacarpals through annular pulleys, cruciate pulleys, and palmar aponeurosis. Once one of these pulleys is broken, the routing path of the tendon will be changed, which will cause trigger finger or bowstring phenomenon like Figure 2a. Teflon tubes with less friction are chosen to imitate pulleys. The PE wire cannot always attach to the glove to avoid interference between Teflon tubes. To optimize the tendon routing path and constrain the distance between the PE line and the finger, the Teflon tube (inner diameter 1 mm; outer diameter 2 mm) on the proximal phalanx is nested on a larger Teflon (inner diameter 3 mm; outer diameter 4 mm) tube on the palm side, shown in Figure 2b. The tendon routing path is evenly distributed on both sides of the glove and runs around the dorsum of the fingertip to allow the fingers that can thus bear stress more evenly during bending.

Model of the Finger Motion with DASA
To evaluate the feasibility of the DASA design, this section presents the 3D kinematic model of the finger bending motion and extension torque model provided by the FPEA. In addition, a cycle time of 3000 is conducted to verify the reliability of the DASA.

Kinematic Model of the Finger Flexion
The flexion angle of the finger can be directly adjusted by controlling the motor rotation angle. Users can control the finger flexion angle to grip or pinch objects of different sizes. The kinematic model of the bending motion can be used to predict the motor rotation angle we need to flexion the finger.
The kinematic model of the bending motion is developed with assumptions as follows: 1) Teflon tubes placed on the intermediate and distal phalanges are evenly distributed on both sides of the finger. 2) The metacarpal and phalanges are rectangular, and finger joints are placed on the midpoint of the edges of each rectangle. [18] 3) The PE line is not stretchable. 4) The finger abduction and adduction motions are ignored because the glove we proposed can only help users to bend and extend their fingers.  www.advancedsciencenews.com www.advintellsyst.com In the kinematic modeling, as shown in Figure 3, the x-axis is aligned to the metacarpal's midline, the y-axis coincides with the boundary of the large Teflon tube on the palm side, and the direction of the z-axis is determined by the right-hand rule. θ jz is defined as the angle of each joint, and j ¼ 1, 2, 3 corresponds to the metacarpophalangeal (MCP) joint, proximal interphalangeal (PIP) joint, and distal interphalangeal (DIP) joint. R j is defined as the rotation matrix of each joint, and R jz is defined as the basic rotation matrix about z-axis. In this study, , because the finger only moves along z-axis according to assumption 4. The unit vector along the metacarpal is u 1 ¼ ½1 0 0 T . And the unit vector along each phalanx is In this model, Teflon tubes are fixed on the fabric glove. p k and p ' k are the endpoints of each Teflon tube. The distance from p 1 to the center plane of the metacarpal is defined as s 1 , which equals the sum of half thickness of the palm, the thickness of the fabric glove, and the outer circle radius of the large Teflon tube. In addition, s k (k = 2,3,4,5) is defined as the distance from p k to the center plane of the corresponding phalanx. Based on these parameters, the coordinate of the endpoint of each Teflon tube p k ðk ∈ Z; k ≤ 5Þ can be calculated as follow 8 > > > > > > < > > > > > > : where l is the length of the Teflon tube on the intermediate phalanx. a is the distance from the projection of p 2 on the central plane of the proximal phalanx to the MCP joint, b is the distance from the projection of p 3 on the central plane of the intermediate phalanx to the PIP joint, and c is the distance from the projection of p 5 on the central plane of the distal phalanx to the DIP joint. Because small Teflon tubes placed on the intermediate and distal phalanges are evenly distributed on both sides of the xy-plane, the coordinates of p 0 m ðm ∈ Z; 3 ≤ k ≤ 5Þ can be calculated by p 0 m ¼ p m À ½0 0 2d zm T , where d zm is the distance from the point p m to the xy-plane. The positions of these points decide the path of the PE line. Meanwhile, the change in the distance from p 1 to p 5 (k L initial À L final k) equals the length change of the PE line when the finger is bent or extended, where L initial is the distance from p 1 to p 5 in the finger full extension state and L flexion is this distance in any flexion state.
The length change of the PE line can be directly controlled by controlling the motor rotation angle φ as follow where r is the radius of the wire spool connecting with the PE line. In addition, the tendon routing path system is an underactuated adaptive system which means the finger flexion trajectory is not completely fixed. The experimental verification of the kinematic model of the bending motion is presented in Section 4.1.

Extension Torque Model of the Finger Extension
Referring to previous research by Conor Walsh, [33] we propose the extension torque model provided by the FPEA. When releasing the bellow, the FPEA is flexible enough to accommodate the finger flexion. After compressing the bellow, the FPEA air pressure increases, and the FEPA hardens to extend the finger. When the bellow is compressed by x, each compressible convolution of the bellow is compressed by x=ðn À mÞ, where n is the total convolution number and m is the incompressible convolution used  to connect other parts. [34] The bottom and top convolutions are incompressible in Figure 4, n ¼ 11, and m ¼ 3. Assuming the inner larger radius R b , inner smaller radius r b , and wall thickness t do not change in the compressing process. Each convolution is compressed by Moreover, the total volume change of the bellow is where x=ðn À mÞ < h À 2t; h is the height of the convolution in the uncompressed state.
Assuming that the volume of FPEA remains unchanged during inflation and deflation, its cross section is simplified to a circle. The radius of the FPEA cross section is w=π, where w is the FPEA inner chamber width at the flat state. The volume of the FPEA inner chamber V FPEA = π ðw=πÞ 2 l FPEA , and l FPEA is the length of the FPEA chamber. Based on the ideal gas law, the pressure after compressing the bellow can be estimated by where V AT is the inner volume of the air tube, and P 0 is the initial pressure when the bellow is uncompressed. In the 3D-printing bellow prototype, the design parameters in the uncompressed state are R b ¼ 10 mm, r b ¼ 7 mm, h ¼6 mm, and nÀm = 9 mm. The inner chamber volume of the bellow in the uncompressed state equals V B0 ¼ 15136.2 mm 3 , which will decrease to V Bf ¼ 8293.8 mm 3 after fully compressed. When the parameters of the FPEA are w ¼ 16 mm, l FPEA ¼ 100 mm, the inner radius of the air tube = 1 mm, and the air tube length = 600 m, the pressure amplification fac- The pressure amplification factor η can be adjusted by optimizing the bellow structure or using the air tube with a smaller inner radius. The extension torque of the FEPA is provided by the pressure acting upon two surfaces that are connected along one edge. [33] The force generated by the fluid across the actuator cross section equals PA cs . p is the pressure and A cs is the cross-sectional area shown in Figure 4. The extension torque generated by the FPEA can be calculated by where d is the force arm. When the FPEA is bent to a certain angle, the cross section changes from a circle to an ellipse, and the moment arm gradually decreases.

Bending Angle Test
We aim to clarify the relationship between the rotation angle of the motor φ and the bending angle θ total of the finger model actuated by the DASA (θ total ¼ θ 1 þ θ 2 þ θ 3 , θ 1 , θ 2 , θ 2 are the bending angles of the MCP joint, PIP joint, and DIP joint, respectively). In the tests, we took the index finger model for the bending angle test and analysis. A sample index finger model is shown in Figure 5a, and the length of the finger is 140 mm (metacarpal 55 mm, proximal phalanx 34 mm, intermediate phalanx 26 mm, distal phalanx 25 mm). The metacarpal was fixed, and each phalanx can rotate around a flexible joint. The joints are made of soft material (TPE-85A, eSUN, China), which simulates the actual human finger joint. The 3D-printing soft bellow design parameters are the same as the bellow used in the prototype mentioned in the model section. The radius of the wire spool connected to the tendon is 12 mm. Figure 6b displays the overall experimental setup. Markers made of textured paper are attached to each phalanx for the bending angle measurement of joints.
In this test, the motor rotated 475°for each group of tests. For every 25°, we measured the bending angle of the MCP joint θ 1 , PIP joint θ 2 , DIP joint θ 3 , and the finger model total bending angle θ total using the software Kinovea. Each group of tests was conducted three times to reduce the measuring errors. The experimental results are plotted by the average test values with error bars, as shown in Figure 5c. The total bending angle is mainly contributed by the MCP joint. Because the end of the tendon routing path is connected to the dorsal of the distal phalanx, at the beginning of the motion, the DIP joint bending angle is larger than the PIP joint bending angle. When the DIP joint is bent to a certain angle, the resistance of the DIP joint increases, and the bending speed of the PIP joint becomes faster. When the motor rotation angle is 475°, the total bending angle of the index finger is 173.  Figure 5d, we substitute the experiment results of the finger joint bending angles (θ 1 , θ 2 , θ 2 ) into the flexion kinematic model to get the input motor rotation angles needed to get these finger joint bending angles. The experimental result and the theoretical flexion kinematic model value have the same trend. However, the flexion kinematic model value is larger than the experimental value for a given motor rotational angle. The main reasons for this deviation are analyzed as follows. First, the model assumes that Teflon tubes are fixed at each phalanx. However, there is a relative sliding between the index finger model and the fabric glove due to the smooth and rigid surface of the finger model. Lastly, the physical glove has some inherent impedance against finger flexion.

Extension Torque Test
The extension torque generated by the FPEA was measured, as shown in Figure 6a. In this experiment, we investigate the influence of the pressure, fabric material, and actuator rotation angle on the extension force. The test rig consists of a torque sensor (DYJN-104 2 Nm, DAYSENSOR) and a precision rotation guide www.advancedsciencenews.com www.advintellsyst.com (RSP 125-L, Dongguan Shengling precision machinery). The rotation guide can accurately fix the FPEA at given angles. The edges of the FPEA are fixed on the test rig using clamps, and the FPEA rotates around a hinge to imitate the human finger joint. The FPEAs (chamber width w ¼ 18 mm, length = 100 mm) in this test are made of different thermoplastic polyurethane (TPU) composite fabrics (420 Denier TPU-coated nylon fabric and 30 Denier TPU-coated polyester fabric). For each pressure (40, 60, and 80 kPa), we inflated the FPEA first, and the FPEA was turned to 90°. For every 15°, we fixed the hinge using the rotation guide. And then, we measured and read the torque values at intervals of 3 s. Five torque values were recorded to reduce the measuring errors. The mean values with different fabrics and pressures are plotted in Figure 7b. The test results of the 30 Denier (30D) TPU-coated polyester fabric (Dongguan Rongsui Textile Products, China) show that the torque increases linearly with a rotation angle from 0°to 30°. However, when the rotation angle is larger than 30°, the slope decreases because the cross-section area decreases with the increase of the actuator angle, also called the creasing phenomenon. The reason for the slow increase is that the fabric becomes stiff, and the air pressure is not large enough to hold the fabric up to form a circular cross-section plane. The extension torques in the 90°are 0.1821, 0.2559, and 0.3018 Nm when the air pressures are 40, 60, and 80 kPa, respectively. According to Equation (6), if the pressure at the initial state is 29.2 kPa, the pressure will increase to 40 kPa after compressing bellow (pressure amplification factor η ¼ 1.37). We expect that the FPEA is flexible in the flexion process and stiff at the finger full extension state. Choosing an appropriate air pressure to keep flexible at the initial state and high pressure after compressing bellow is important in the DASA design. For 30D TPU-coated polyester fabric material, the pressure of 40 kpa is enough for normal people's finger extension, but it requires a larger pressure if the user is elderly with stiff fingers. In addition, unlike the 30D material, the 420D TPU-coated nylon fabric FPEA (Jiaxing City Yingcheng Textile Products, China) shows a different torque trend. At the same pressure, the 420D material provides larger torque because of its high hardness. In addition, after rotating the actuator 60°, the torque surges. The main reason is that the fabric is approaching its strain limitation. The 420D fabric FPEA still provides high torque when rotated up to 90°at low pressure or even without inflating situations. The 420D fabric FPEA still gains great torque (0.9234 N) when rotated up to 90°at low pressure. That is why fabrics with high hardness are not suitable for FPEA.

Cycle Test
Repeatability is an important index to evaluate the reliability of the actuator. The cycle test of the DASA was conducted 3000 times from 0°to 150°, demonstrating its excellent reliability, as shown in Figure 7. Before the test, the pressure precharged in the FPEA is 80 kPa at the finger full extension state. The cycle test platform is almost the same as the bending angle test in Figure 5b. The difference is that an additional cooling fan (The fan is not needed in practical applications as the user will  www.advancedsciencenews.com www.advintellsyst.com not keep doing finger flexion and extension for so many cycles in a short time) is set up to avoid motor overheating caused by the long-time operation. The cycle test was conducted 3000 times, and the maximum total bending angle θ total of the model finger with DASA was recorded every 25 times. The recorded data are plotted in Figure 7. In addition, to estimate the influence of the cycle times on the bending process, the bending motion curves at the process around 1000, 2000, and 3000 times are recorded to illustrate greater credibility. For the bending process stability verification, we test the response and recovery bending angles at 9 periods, as shown in Figure 7. For each period, the motor rotates from 0°to 450°and recovers from 450°to 0°. The total bending angle θ total of the model finger was recorded every 50°, and the bending angle was analyzed by Kinovea. Both the maximum bending angle cycle test and the bending process reliability cycle test show excellent stability in the DASA design.

Design and Evaluation of the Soft Robotic
Glove with DASAs

Design of the Soft Robotic Glove
Based on the DASA concept, we designed and manufactured a portable soft robotic glove, as shown in Figure 8. Notably, the FPEAs do not need to connect with an independent pneumatic system. The flexion and extension of each finger can be controlled by a single-motor rotation. Figure 8a presents the prototype of a soft robotic glove. Three DASAs are fixed on a fabric glove to drive the thumb, the index finger, and the middle finger.
Three fingers are actuated in this prototype which can meet the requirement of most grasping tasks. The lightweight soft glove (68 g) is connected to a compact control box (835 g, 180 Â 104 Â 73 mm) through three Polyurethane (PU) tubes for air delivery and three Teflon tubes to guide tendon wires. We designed and fabricated the FPEAs using a 30D polyester fabric coated with TPU film in this work. This composite fabric material has two layers, in which the polyester fabric layer provides anisotropic mechanical properties, and the TPU film is fixed on one side of the fabric to prevent air leakage. [30] Two layers of TPU-coated polyester fabric can be bonded together using a thermal sealing machine (FR-400 A, Blueberry, China). [29] In the thermal sealing process, the TPU surfaces of two composite fabrics are in contact to obtain better heat-sealing quality. The FPEA is thin (0.6 mm) and lightweight (1 g) in the uninflated state and the polyester fabric layer makes the robotic glove inconspicuous. When the internal pressure of FPEA is low during the finger flexion process, the FPEA is flexible, reducing the obstruction to finger flexion. As the internal pressure of FPEA increases by compressing bellow, the FPEA extends the finger better. The design parameters of the FPEA are length = 120 mm, width = 24 mm, and the distance between the fabric edge and its adjacent thermal welding traces is 4 mm to avoid leakage failure.
The tendon routing path on the glove can be seen in Figure 8b marked by a red line, also refer to Figure 3. The tendon applies force to the finger through the Teflon tube, which may deform the glove. [35] We chose a neoprene composite fabric coated with polyester fabric (Dongguan Rongsui Textile Products, China) on both sides to ensure that the glove has enough strength to resist  deformation. The thickness of the composite fabric is only 1 mm to keep the glove flexible.
The actuation system and control system are assembled in a control box shown in Figure 8c. The actuation system has three units, each controlling one finger. The soft bellows are fabricated by 3Dprinting technology (KP3 3D printer, Kingroon, China) using thermoplastic elastomers (TPEs) material (TPE-85A filament, eSUN, Chine). The bellow bottom is fixed on the frame, and the bellow top is connected to a cap that can uniformly bear pull force from three cables that distribute in a circumferential direction. To keep the bellow contract and elongate linearly, we installed 3d-printed guide rings in the middle of the bellow. Three linear shafts distribute in a circumferential direction for each CBA module, which can guide the cap and the guide ring to move linearly. [34] The control system consists of a Bluetooth transceiver module (HC-06, Guangzhou HC Information Technology, China), a microcontroller (Arduino mega 2560, Arduino, Italy), a signal converter module (FE-URT-1, Feetech, China), and two buttons. One of the buttons is used to control the battery to power the microcontroller, signal converter module, and servo motors (STS3215, Feetech, China). Another button controls the Bluetooth transceiver module on and off. The rotation of the servo motor can be controlled directly by Bluetooth serial port applications on smartphones. The control box can be placed in a crossbody bag so that it can be carried by patients easily.
In the prototype design, the finger bending speed is determined by the motor and the size of the wire spool 2. For the STS3215 motor at 7.4 V voltage input, the response time (from 0°to 170°) and recovery time (from 170°to 0°) corresponding to the finger flexion and extension are both 2 s as shown in Video S1, Supporting Information. There is a delay of several milliseconds between the motor rotation and the finger flexion when the finger starts to bend because the tendon is slightly loose at the finger full extension state to protect users. This delay is almost invisible to the naked eye. Because the tendon displacement for finger flexion is far larger than the cable displacement required by fully compressing bellow, the cables keep loose under the full finger flexion state. However, there is no delay between finger extension and motor rotation when the finger starts to extend, even though the bellow is not compressed immediately. The finger extension responds rapidly because keeping the finger at full flexion state requires tendon-driven force. Once the tendondriven force is released, the finger will immediately extend to a natural state. In addition, the sensitivity of the DASA is determined by the rotation resolution of the motor and the size of the wire spool 2. For the rotation degree resolution of the STS3215, this motor rotates 360°in 4096 pulses. The finger will bend 180°w hen the motor rotates 475°(diameter of the wire spool 2 equals 24 mm). The sensitivity S of the DASA can be roughly calculated by S ¼ 170°Â 360°=ð475°Â 4096Þ ¼ 0.0310°per pulse. Because of the adaptiveness and under-actuation characteristics of the DASA, precise bending angle control is impossible and meaningless without sensors. The calculation is intended to illustrate the good sensitivity of the DASA, which is enough for helping patients in ADL. Compared with pneumatic soft actuators with awkward sizes, around 13 mm in height, and bulky control boxes (5 kg), [26,36] the pumpless and inconspicuous DASA design makes the robotic glove looks like a normal glove without noise and vibrations generated by normal pump systems. Moreover, the flexible FPEA is safer compared with the antagonistic tendon-driven actuation controlled by a complex mechatronic system. [18] To further verify the practicability of the robotic glove with DASAs, an experimental test is introduced in Section 5.2.

Experimental Characterization of the Soft Robotic Glove
The development of robotic gloves does not only focus on hand function rehabilitation but also considers helping patients in ADL. For the elderly who cannot take care of themselves well, body cleaning is difficult and often requires the use of towels. Squeezing water out of towels is a significant routine in their daily life. Pioneer researchers proposed a twisting towel test to qualitatively estimate the robotic glove-assistive functions. [26] To quantitatively verify the assistive performance of the proposed robotic glove with DASA, we optimize the twisting towel test to a squeezing sponge test as shown in Figure 9a. The columniform sponges used in this test are made of polyvinyl alcohol (PVA) material with a length of 95 mm and a diameter of 55 mm. Although the robotic glove is waterproof, wearing a wet glove will be uncomfortable for the user. So, the sponge is wrapped in an open plastic bag to prevent the gloves from getting wet. During the test, the subject's active muscle activity is monitored by an EMG system with a sampling frequency of 1000 Hz (smacq USB 3130 acquisition card, SMACQ, China). The EMG system is developed and proven in our previous research. [37] As shown in Figure 9b, 5-channel EMG electrodes are placed on the forearm corresponding to the flexor pollicis longus, flexor digitorum superficialis, flexor digitorum profundus, extensor digitorum, and extensor carpi radialis brevis, respectively.
In the squeezing sponge test, a healthy adult volunteer was guided to squeeze the PVA sponges filled with water. The volunteer gripped the sponge under three conditions: 1) the volunteer wore the robot but the power of the actuator was off, as a baseline (actuator off ); 2) the robotic glove was activated to help the subject in squeezing the sponge, but the subject did not actively exert finger force during finger flexion (actuator on_passive); 3) the robotic glove was activated to help the subject, and the subject actively squeezed a sponge simultaneously (actuator on_active). For conditions 1 and 3, the subject presses the sponge using three fingers (thumb, index finger, and middle finger) as hard as possible. For each condition, the volunteer was required to squeeze the water out of the sponge into the bowl. The water weight was measured by an electric scale. For error alleviation, each condition was conducted three times to obtain the mean value, and the test order about three conditions was randomized. The volunteer relaxed for 15 s between each grasping which can avoid muscle fatigue to get convincing data.
The squeezing sponge test result is shown in Figure 9c. The volunteer can squeeze out 56.2 g of water with three fingers of the dominant hand under condition 1 (actuator off ). When the finger is only actuated by the glove (actuator on_passive), the volunteer can squeeze out 53.4 g of water. Compared with the weight of water output squeezed by hand only, the robotic glove can restore three fingers' 95.3% capability when the hand is completely relaxed during squeezing a sponge. Notably, the water output in condition 2 is more stable than that squeezed by the human hand because the glove does not exist fatigue, which can be verified by the error bars length. Under condition 3, the volunteer squeezed out 35% more water (75.8 g) than he did with his hand only. These results prove that the robotic glove with DASA has great hand-assistive functions in ADL.
The processed EMG signal of the squeezing sponge test is shown in Figure 9d, which is normalized to the maximum voluntary contraction (MVC) of the volunteer, [36] and the normalized data are processed by the Gaussian smoothing filter with smooth factor 500 in MATLAB. Figure 9d presents the EMG signal change of the corresponding electrode when the subject squeezed sponges three times under each of the three conditions. For all three flexors (0-2) under the actuator on_passive situation, the EMG amplitude is far below the other two conditions, which proves that the subject does not actively bend fingers when squeezing sponges. For two extensors (3,4), the bandwidths of EMG signals under condition 2 are distinctively smaller than in the other two conditions. In addition, the mean amplitudes of EMG signals corresponding to two extensors under condition 2 are smaller than in the other two conditions. Both show the FPEA could help the subject extend fingers with less force. In short, the glove can help the subject in dealing with some ADLs such as towel water squeezing even when the subject does not apply any force in finger flexion and extension as shown in experimental results from condition 2.
The experimental procedures were reviewed and approved by the Human Research Ethics Committee of the University with number EA1903040. We have the informed written consent of the participant.

Conclusion
This research proposes a novel actuator design for the soft robotic glove by combing two different actuation methods driven by a single motor, defined as the DASA. For the DASA design, the finger extension is achieved by the fabric-based extension actuator (FPEA), while the finger flexion is actuated by tendon driven. By means of the CBA, a single motor is used to actuate two actuation modes, which simplifies the mechatronic design. A nested structure tendon routing path is proposed to improve glove comfort. The kinematic flexion model of the tendon-driven actuator and the extension torque model of the FPEA are established to estimate the performance of the DASA, which is verified by the experimental characterization. Finally, the finger flexion and extension-assistive performance of the soft robotic glove are evaluated through squeezing sponge tasks with EMG monitored.
There are some potential development spaces for the current prototype design. First, even though the tendon-driven motion can partly protect users and avoid potential damage, users may still feel uncomfortable if the finger bending angle is far larger than they are required during grasping. Flexure and tactile sensors made of smart textile material can be integrated into the glove for closed-loop control to protect users. [6][7][8] Second, for patients with hand disabilities, it is difficult for them to put on or take off gloves by themselves. A fabric glove structure that is easy to wear and adapt to different palm sizes is required. Finally, the current smartphone control is not applicable to subjects with impaired function in both hands because one hand must be used to control the glove motion. In the future, a soft robotic glove with an EMG control system will be developed to realize feedback control by combining flexure and tactile sensors.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. www.advancedsciencenews.com www.advintellsyst.com