IROS2019国际学术会议论文集2482

Abstract Constraint-induced movement therapy (CIMT) is an effective recovery protocol for patients with stroke. However, not all the stroke patients can participate in CIMT. Devices that assist the movement of the affected side could help increase the number of stroke patients who can participate in CIMT like protocol. In this paper, a soft cable-driven wrist-wearable robot called Exo-Wrist, which assists the wrist of a paretic arm in performing the dart-throwing motion (DTM) with an active forearm anchor, is proposed. To increase the force transmission efficiency while preventing medical issues due to long-term pressure, a corset active anchor, which compresses the body only when anchoring is needed, is developed. A moving pulley mechanism and a three-layered structure increase the functionality of the anchor. Considering the natural movement of wrist extension and the frequently-used wrist motions in daily life, the wrist is assisted along the DTM direction, and the method to find a personal DTM is introduced. A 3D-printed custom wrist armlet is developed to hold the tendon path and exert a torque along the DTM direction. Together with a hand assistive device, Exo-Wrist could be used as a tool to help stroke patients with difficulty in wrist extension to participate in task- oriented rehabilitation protocols such as CIMT. Keywords Rehabilitation robotics, Soft robotics, Robot motion. I. INTRODUCTION EOPLE with hemiplegia after stroke lose in whole or in part the functionality of the affected side. Hence, the major goal of stroke recovery protocol is to recover the function of the affected limb 1. Patients with hemiparesis undergo various rehabilitation interventions, and among them, constraint-induced movement therapy (CIMT) is regarded as one of the most effective rehabilitation protocols for patients with hemiparesis 2. The major goal of CIMT is constraining the non-affected side of the upper limb and forcing the use of affected side to perform task-oriented activities with intensive practice 3. Nevertheless, a majority of patient with stroke are not eligible for CIMT due to harsh requirements of CIMT protocolsonly 6.5% of patients with stroke are eligible for CIMT 4. Among them, a few impaired patients(around 10% in 5)cannot benefit from CIMT because they have very little residual upper limb motion on their affected side, especially relevant to the voluntary extension and flexion of the hand due to their flexion synergy of wrist and fingers 2. Therefore, in order to increase the number of patients with hemiparesis to benefit from CIMT therapy, especially patients who cannot participate in CIMT due to their flexion synergy, one alternative solution is to develop devices which can help hand and wrist movements while the patients intend to move. This paper is supported by the National Research Foundation of Korea Grant funded by the Korean Government (MSIT) under Grant NRF- 2016R1A5A1938472. This would allow them to participate in therapy which follows CIMT protocols but with robotic assistance. For designing such robots, we selected several factors which should be considered when designing a wrist assistive device to help patients participate in CIMT with robotic assistance protocols based on traditional requirements in CIMT 2-4. First, developers should consider the minimum requirements of hand and wrist movements, 20 extension of the 1-2 fingers and wrist from the original protocol 3. Second, since task- oriented motor tasks are also conducted during CIMT, robots should assist necessary movements in activities of daily living (ADLs), especially tenodesis grasp (passive finger flexion in response to wrist extension) for finger movement or dart- throwing motion (DTM) for the wrist 68. Thirdly, considering that the CIMT is conducted not only in a clinical setting but also in the patients home environment, the robots are preferred to be portable, compact, and lightweight. Fourthly, as CIMT is a rehabilitation method that uses the entire upper extremity, it would not be adequate to severely hinder the movements of other upper limb joints. Lastly, pressure on the skin and muscle tissues over long durations should be avoided considering the long-duration of the therapy. Previous studies have suggested various hand-wearable robots that can fulfill these requirements, including tenodesis assistance robots 913. Robots for wrist rehabilitations or assistance also have been developed so far. Several robots have been made rigid for high transmission efficiency and controllability 14, 15. To better adapt to humans and their ADLs, some researchers have developed mobile soft wearable robots 1618. However, few wrist robots are considering all the above mentioned factors, and some of them are not immediately compatible with other wearable robots due to their size or rigidity For making the robot mobile, the robot needs anchors that affix it to the human body and help transfer an assisting force to the wrist by resisting the reaction force of the actuator, especially for soft wearable robots. In soft wearable robots, the reaction force usually occurs along the shear direction, parallel to the bone. Most of the soft robots realize anchoring with geometric locking using the surrounding bone shape or with frictional force by tightly tying the structures 9, 10, 19. In wrist robots, it is desirable to use anchoring on the forearm to prevent wire tension from affecting other joints. Geometric locking cannot be used in it because the geometric shape of the forearm cannot resist the shear force exerted by the actuator. Therefore, most of the soft wrist wearable robots simply tighten the robots to the body to secure their positions. Researchers are with Biorobotics Laboratory, Department of Mechanical Engineering, Seoul National University, Seoul, 08826, Republic of Korea.(corresponding author to provide e-mail: kjchosnu.ac.kr) Hyungmin Choi, Brian Byunghyun Kang, Bong-Keun Jung, and Kyu-Jin Cho, Member, IEEE Exo-Wrist: A Soft Tendon-Driven Wrist-Wearable Robot with Active Anchor for Dart-Throwing Motion in Hemiplegic Patients P IEEE Robotics and Automation Letters (RAL) paper presented at the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Macau, China, November 4-8, 2019 Copyright 2019 IEEE However, considering that CIMT lasts for a long period of time (90% of waking hours) 2, this anchoring method could cause medical issues from long-term pressure if an anchor strongly tightens the body and remains that way for a long time. Hence, it is necessary to develop a new type of anchor that can mitigate the pressure issues while fixing the robot to the body. In this paper, we introduce Exo-Wrist, a novel soft tendon- driven wrist-assist robot. Exo-Wrist supports the wrist in the dart-throwing motion (DTM) direction and has an active anchor (Fig. 1(a). Exo-Wrist is portable, compact, and lightweight enough to be used in daily life and supports wrist movements adequately fit to ADLs with a simple command. Moreover, it can support hand ability by tenodesis grasp. Additionally, it can directly coordinate with other wearable robots, especially hand-assist gloves because of its compactness and softness. Therefore, Exo-Wrist can support patients with wrist impairment, as well as the ones with other upper extremity impairments. The main contributions of the paper are as follows: 1. Corset-active-anchor-based soft tendon-driven wearable robots: For efficient force transmission, a forearm anchor should be able to resist the pushing force from the actuator, but the part should not tighten the body for a long time, to prevent medical issues due to long-term pressure exertion. Therefore, we propose the active forearm anchor, which compresses the forearm only when a resistance force is needed. To generate high compression with a small actuator, a corset based moving pulley mechanism actuated by a tendon is used 2. DTM direction wrist assistance: The robot is designed to assist the wrist in the DTM direction, which is the natural wrist extension and principal wrist movement during ADLs 7, 8. Because the DTM is slightly different from person to person, we propose a noninvasive and ergonomic method to find the DTM of each person. Moreover, in soft wearable robots, it is hard to hold tendon paths in the desired way, due to the compliance and deformability of soft structures. Therefore, we developed a 3D-printed custom wrist armlet, which can maintain the tendon. II. DESIGN A. Hand glove Figure 1(a) gives an overview of the Exo-Wrist. Hand glove is the part that attaches the tendon to the human body, to transmit tension to the target joints. At the glove, a wire enters the dorsal side of the hand from the wrist, winds around the hand below the MCP, and goes out to the wrist ring near the inserted point. The path is similar to the Exo-Glove Poly II 9, which designs the path to exert the tension in the direction normal to the surface for stability. The main body of the glove is a fingerless golf glove with a wire path formed by embedded Teflon tubes in silicone. The wire in-and-out location is designed to make the tendon path collinear with the path on the moment-arm support structure on the wrist ring. B. Wrist ring Wrist ring is composed of two parts: oval-shape rigid armlet and flexible moment-arm supporting structure. The ring determines the position of the actuation wire on the wrist, so it is important for the ring to put the wire on the DTM plane of each person. Because the DTM plane varies from person to person, the armlet of the wrist ring was customized based on three personal parameters: wrist width (w), wrist height (h), and tilted angle of the DTM plane (_d) (Fig. 1(b). _d determines the location of the wire path, and w and h determine the shape of the armlet. By this customization process, the wire near the carpal joint can be located on the personalized DTM plane, and the torque generated by the wire tension would make the wrist move in the DTM. Moreover, the customized size will not allow the ring to rotate, which means that the wire will be held in the desired position. The rigid armlet is produced by a 3D printer (Connex 30, Stratasys, USA), and silicone pads are attached to its inner part for comfort wearing. _d is calculated from wrist kinematics, which will be introduced in Section 3. The moment-arm supporting structure is designed as a series of trapezoids made of flexible silicone rubber attached to the rigid connection part (Fig. 1(c). Because the moment arm of the suspended tendon-driven system is the vertical distance between the joint and the wire, this structure can Fig. 1. (a) Illustration of the robot (b) Front view of a wrist ring. The shape of the armlet is oval with a flat top line, which is customized to each participants distal forearm shape. Moment arm-supporting structure, where tendon passes through, is designed to locate its middle to be above the line tilted by in a vertical line. (c) Right-side view of moment arm-supporting structure. Due to its flexibility, it can adapt its shape to the movement of the wrist. (d) Front view of the forearm active anchor. It consists of three layers, with a sheath and an actuating wire inserted in it (e) Forearm anchor shape before (left) and after (right) actuation (f) Conceptual figure of collaborative work of Exo-Wrist with Exo-Glove Poly 9 increase the efficiency of torque transmission to the wrist by increasing the moment arm of the wire. The structure is integrated with the armlet using bolts at the location of the wire path of the armlet (_d). Since the supporting structure is located on the wrist, it should be able to change its shape with wrist movement. Flexibility of the material and shape of the structure help the structure to adapt to the shape of the wrist. Moreover, a low-friction Teflon sheet at the bottom of the structure makes it slide back and forth C. Forearm active anchor The forearm active anchor compresses the forearm to generate a large frictional force when the robot assists the wrist movement, and releases when the robot does not exert any assisting force. An active forearm anchor is a truncated cone shape structure consisting of three layers (Fig. 1(d), (e). The bottom layer is in contact with the skin and is made of silicone, which is safe in physical contact with skin and has a high friction coefficient. The high friction coefficient and large contact area with the skin make it possible to generate sufficient frictional force against the pushing force of the actuator with less pressure. Spring sheaths from the actuator are inserted into this layer. Embedded Teflon tubes form wire paths collinear with those sheaths. The top layer is an active layer that compresses and releases the anchor by sliding above the middle layer. Pulleys on both ends of the top layer are connected to the anchoring wire (Dyneema), which is pulled and released by a small motor at the anchor. Similar to the corset, the anchoring wire is routed using a moving pulley mechanism. This mechanism increases the tightening force between the ends of the top layer by several times compared to the wire tension at the end of the wire. In other words, a small motor that can generate up to a fraction of the required tightening force is enough to generate the required compression force. A PET sheet is used as the base of the top layer because though compliant, it is stronger along the tensile direction and has a small friction coefficient. Between the bottom and top layers, a Teflon sheeta compliant sheet with a very low friction coefficientcovers the bottom layer. Due to its low friction coefficient, the top layer can slide freely with respect to the bottom layer, which increases the efficiency of the anchor. The reason for using multiple layers instead of a single layer in the anchor is to avoid direct contact between the anchoring wire and the skin to prevent abrasion between them, and to increase the anchoring ability by satisfying the conflicting friction conditions: high friction is needed between the anchor and the skin, but low friction is needed when sliding. When the robot starts to assist wrist movements, anchoring is needed. Thus, an active anchor compresses the forearm before the force transmission wire is actuated by the pulling anchoring wire using a small motor. Once the wrist movement is assisted and the tension on the force transmission wire becomes zero, anchoring is no longer required. Hence, the anchoring wire is released from the anchoring motor, relieving the compression. Then, the subjects can feel comfortable and avoid problems due to long-term pressure. D. Other parts The force transmission wire is pulled by a small pulley combined with a rotary motor (IG-32GM 03Type, D thus, 3 mm was selected as the maximum allowable anchor displacement. For all three participants, the maximum anchor displacements were within the allowable range for an anchoring wire tension of 11 N and above. Furthermore, at 11 N, the average pressure applied by the anchor is endurable for short durations 27. Therefore, a tension of 11 N was selected as the actuation level of the anchor, which corresponds to 45% of the power of the selected actuator. The purpose of the feasibility test of the active anchor is to validate the role of the active anchor during the operation of the Exo-Wrist. After they wore the customized robots, the subjects were instructed to sit on chairs, put their forearms on the jig (Fig. 4(b), with their elbows straight and their forearms pronated, and firmly hold a 3D-printed pulling rod (UPrint SE PLUS, Stratasys, USA). The pulling rod was connected with a loadcell (333FDX, KTOYO, Korea), to measure the wrist torque. This torque was measured by multiplying the force measured by the loadcell and the length between the rod and the wrist joint center. When the test began, each participant was instructed to exert the maximum possible wrist extension force. In each trial, they reached their maximum in 5 seconds, and a break of 30 seconds between the trials was given to the participants. The robot assisted the wrist extension in a total of 12 trials (support trials, ST) but did not provide any support to the participants in the other six trials (non-support trials, NST). Among the STs, forearm anchor was activated in six trials (activated anchor support trials, AAST) but not in the other six trials (deactivated anchor support trials, DAST). Robots supported the wrist movement only when commanded by the participant to support the extension. The intention (command) of the participant was detected by an EMG sensor. When the robot was commanded not to support, the wire tension was maintained at 0 N. The anchor compressed the forearm during the activated anchor support trials and released the forearm during the rest. The order of the trials was randomized. Table I lists the results of the test. Values in the “activated” row of each participant indicates the difference between the average of the maximum wrist extension torque during NSTs and the average of the maximum wrist extension torque during AASTs and deviations. Values in the “deactivated” row are similar, but the difference is that it is calculated between the NSTs and the DASTs. As can be seen, when activated, a robot can assist more than 0.5 Nm. When the anchor is not activated, the robot cannot generate an average assistance torque of more than 0.2 Nm. By using the Wilcoxon rank sum test, statistically significant differences between the measured wrist extension torques of the NSTs and AASTs, and of the AASTs and DASTs were found for all the participants (p0.1). C. Motion tracking The purpose of the second experiment is to validate whether the robot supports wrist movement in the target DTM