1Graduate School of Engineering, Kagawa University, Takamatsu, Japan; 2Department of Intelligent Mechanical Systems Engineering, Kagawa University, Takamatsu, Japan; 3The Institute of Advanced Biomedical Engineering System, School of Life Science and Technology, Key Laboratory of Conver-gence Medical Engineering System and Healthcare Technology, the Ministry of Industry and Infor-mation Technology, Beijing Institute of Technology, Beijing, China
Abstract: Robot-aided rehabilitation training allows patients to receive a more effective and stable re-habilitation process. Exoskeleton devices are superior to the endpoint manipulators and cable suspen-sion devices on the aspect that they can train and measure the angle and torque on each joint of im-paired limbs. For robotic devices, physical safety should be guaranteed since the robot-assisted train-ing relies on high human-robot interaction especially for exoskeletons. Traditional robotic devices mainly introduce the stiffness actuator, while the high levels of kinetic energy of robots will induce unsafe. For guaranteeing the safety of pa-tients, compliant actuator such as the series elastic actuator (SEA) and variable stiffness actuator (VSA) design has been involved into these devices. The added compliance can make robots intrinsically safe and realize the energy-efficient ac-tuation. The VSA used a variable stiffness elastic component instead of a constant stiffness elastic component, and VSAs is deemed to a kind of SEAs. A closed-loop interaction control method was used for SEAs to generate low impedance. By comparison, the VSA realizes adaptable compliance properties with inherent mechanical design. Thus, for SEAs, an addi-tional mechanism is needed to adjust the output stiffness. In this paper, two kinds of compliant exoskeleton devices de-signed with the SEA and VSA respectively are introduced. The mechanical design and control method for each device are introduced; especially the design for guaranteeing patients’ safety.
Keywords: Compliant actuator, home-based rehabilitation device, series elastic actuator, variable stiffness actuator, mechanical design, control method.
1. INTRODUCTION
Approximately 795,000 people suffer from a new or re-current stroke every year according to the statistics [1].Stroke can lead to impaired motor control of the upper and lower limbs with significant impairment of activities of daily living (ADL). Traditional therapy is labor-intensive and it requires one-to-one therapist-patient interaction [2]. There-fore, it becomes onerous to perform highly intensive treat-ment for all patients. Given that, treatment based on the ro-botic technology has been investigated by many research groups and can be separated into three types: end-effector[3], cable suspension [4], and exoskeleton [5]. Among them, the exoskeleton device proposes a solution to the problem of control and measurement of angle and torque on each joint of impaired limbs [6]. For instance, the ARMin robot which is one of typical exoskeleton devices can support the entire arm for rehabilitation with patient-cooperative control strate-gy during activities of daily living (ADL) training tasks [7]. For robotic training devices, it is necessary to guarantee enough torque performances to perform training while
*Address correspondence to this author at the Graduate School of Engineer-ing, Takamatsu, Japan; Tel/Fax: +81-087-864-2356, +81-087-864-2369;E-mail: [email protected].
allowing for a safe patient-robot interaction, especially spasmodic happens [8]. For the conventional robot with “stiffness actuator”, the high levels of kinetic energy of ro-bots will induce unsafe; especially the motion is associated with the fast speed. The SEA designed with a new actuator principle has been introduced into the compliant robot de-sign. The SEA is a high weight-to-torque ratio actuator which can decrease the weight while guaranteeing sufficient force performance. The added compliance makes robots in-trinsically safe and realizes the energy-efficient actuation. The compliance of the SEA realizes low impedance with an additional control loop and the compliance property is fixed [9]. As an improvement, VSAs added variable stiffness elas-tic components instead of constant ones. VSAs are capable of varying the apparent output stiffness independently of the actuator output position. The adjustable compliant behaviorof VSAs is an inherent hardware property [10, 11], thus a mechanism for varying the stiffness is needed. The VSA is also a trade-off between the compliancy and accuracy [12]. For instance, Lenzi et al. proposed a variable impedance powered elbow exoskeleton device named “NEUROExos”.The device used an antagonistic actuation system providing a software-controllable passive compliance [10]. Wang et al. also proposed the active variable stiffness elastic actuator (AVSER), and the compliance was changed by shortening
Mechanical Design and Control Method for SEA and VSA-based Exoskeleton Neuroscience and Biomedical Engineering, 2014, Vol. 2, No. 3 143
the effective length of the leaf spring, thus the transmission ratio between the actuated load and the force from the inter-nal elastic elements can be adjusted [13]. In this paper, two kinds of exoskeleton devices with the principles of SEAs and VSAs respectively are introduced.Both of them were designed according to therapist’s advices. The mechanical design and control method for each device are introduced, and the method for ensuring patients’ safety is addressed. The remainder part of this paper is organized as follows. In the section 2, the working principles of SEAs and VSAs are briefly introduced. In the section 3, the mechanical design and control method of the SEA-based exoskeletondevice are focused on. In the section 4, the VSA-based exo-skeleton device is introduced. The conclusion will be given at last.
2. WORKING PRINCIPLE OF SEAs AND VSAs
The working principles of the SEA and VSA are shown in (Fig. 1) [12, 13]. Usually, a gearbox is added to the motor which can realize a high weight-to-torque ratio performance and decrease totally weight of devices. However, an una-voidable problem with large reduction ratio gearheads is the non-backdrivability resulting from high-reflected inertia and friction. To obtain variable impedances or a compliant joint, elastic elements were added between the motor shaft and output of the device. The dynamic of the load in each actua-tor can be written as:
m��=k(x1-x2) (1)where the m is the mass of the load, x1 is the equilibrium controlled by the motor, x2 is the displacement of the output, and k is the coefficient of the elastic element. It can be found that VSAs are differing from SEAs with a controllable stiff-ness k [12]. The variable stiffness elastic components make the VSA an inherent hardware property compliant behavior. Thus, a mechanism for adjusting the stiffness is needed. Bycomparison, the SEA realizes a closed-loop interaction con-trol method to generate low impedance with a constant stiff-ness elastic component.
(a) Working principle of the SEA.
(b) Working principle of the VSA.
Fig. (1). Working Principles of the SEA and VSA [12, 13].
3. SEA-BASED EXOKELETON DEVICE AND THE CONTROL METHOD
3.1. Mechanical Design
In our previous study, Z. Song et al. designed a human upper-limb exoskeleton rehabilitation device (ULERD) as
shown in (Fig. 2) [14]. The motivation of the ULERD is to provide passive and active training to patients with motor dysfunction to recover the motor function of upper limb. The working principle of this device is relative to the principle of the SEA [14]. The exoskeleton device realizes an ergonomic physical human-robot interface that is convenient to wear and comfortable to operate. The device can provide the train-ing for both elbow and wrist joints and it is easily worn by caregivers or patients themselves. In this paper, we main focus on the elbow joint, thus the wrist part is ignored.
Fig. (2). Human upper limb exoskeleton rehabilitation device (ULERD).
The detailed structure of the elbow joint is shown in (Fig.3). The two passive Degree of Freedom (DoF) mechanisms (rotation and translation) in the elbow joint allow constant alignment between the user’s elbow and robot axes. Other-wise, joint misalignment between the robot and human joints would introduce unwanted translational forces. Patients may feel not comfortable when the translational force happens.
Fig. (3). CAD drawing of the elbow joint structure.
3.2. Torque Limiter for Physical Safety of Patients
Security issues of patients are important caused by robot-ic devices. A torque limiter mechanism was designed as shown in (Fig. 4). The friction between the axle sleeve of the elbow motor and cable driving part is controlled by adjusting the screw. Therefore, the cable driving part will separatefrom the motor shaft when overloaded. When external torque is less than the threshold preset, the movement can be trans-
����� ����� � ���
����� ����� � ���
� ��
�
� ��
�
144 Neuroscience and Biomedical Engineering, 2014, Vol. 2, No. 3 Zhang et al.
mitted to the cable driving part through a stainless steel wire with a diameter of 0.5mm. The torque threshold of the torque-limiter mechanism can be adjusted to different pa-tients. The design can avoid the suddenly increased force (e.g. spasm).
Fig. (4). Torque limiter in the elbow joint.
3.3. Control Method for the ULERD with Non- back-drivability
To decrease the totally weight of device, the BLDC mo-tor with high power density was used, and the frame of thedevice was fabricated with aluminum. After adding a gear-box with a high reduction ratio (231:1), the device can out-put enough torque performance to perform rehabilitation tasks (with Max. continuous torque 14.2mNm), especially for passive trainings. However, the intrinsic stiffness was increased after adding the reduction gear and a reflected iner-tia will be added to the elbow joint. Consider that the com-pliant actuation is claimed to be one of the main require-ments for a rehabilitation device [15], the added reflected inertia will influence the compliance. The influence can be partly mitigated by using a direct-drive motor; however, the weight of the device will be increased for generating the same power performance. Therefore, the elastic elements were added between the motor shaft and output of the de-vice. Thus, the mechanical design is similar with the Series Elastic Actuator (SEA) [13]. The SEAs have several ad-vantages, including high shock tolerance, low reflected iner-tia, high energy-storage capacity, and accurate and stable force control [13]. After introducing the SEA into the exoskeleton device, the compliance is actually realized with a closed-loop inter-action control strategy. By controlling the deflection of elas-tic elements in the SEA, the output impedance can be con-trolled. For generating the desired virtual impedance, the control algorithm can be written as Equation (2),
Xmd�-��= KV�+BV�Ks
(2)
where Xmd is the desired rotation angle of the motor, � is the rotation angle of the device, Kv and Bv are the parameters for virtual impedance model, and Ks is relative to the elastic el-ement. It can be found that for the mechanism of SEA, the stiffness of the actuator is constant. A typical proportional-
integral-derivative (PID) algorithm was used for controlling the deflection accurately. The detailed PID algorithm is writ-ten as Equation (3) and (4),
v t =Kpe t +Ki e(t)dt�t0 +Kd
ddt
e(t) (3)
e t =Xmd t -�(t) (4)
where the Kp, Ki and Kd are three parameters of the PID con-troller, e is controlled error. These parameters can be exper-imental determined. The active training can also be realized with the closed-loop interaction control [16]. The generated variable impedance can help the motor recovery of stroke patients. We carried out an experiment by controlling the deflec-tion between the forearm and device as a constant value for evaluating the proposed control method. During the experi-ment, a subject was invited to wear the device to perform the flexion and extension motion with a speed about 1Hz while the device provides a constant assisting force. The experi-mental result is plotted in (Fig. 5). The blue line is the rota-tion angle of the device recorded from the encoder. And the red line represents the rotation angle of forearm recordedfrom an inertia sensor mounted on the forearm. The result shows that the deflection can be controlled as a constant val-ue with the proposed control method. Moreover, the device can also give a resisting force during the active training.
4. VSA-BASED COMPLIANT EXOSKELETON DE-VICE AND THE CONTROL METHOD
4.1. Mechanism Design
The designed prototype of the VSA-based exoskeleton device is shown in (Fig. 6). The stiffness control part is in-dependently from the position control part. A forearm sup-porter is a commercial product which is easy to wear and the motion range can be adjusted manually. In the position control part, a torque-limiter mechanism is designed for guaranteeing the safety of patients. The torque limiter can provide patients more protection. The power for the flexion/extension motion is provided by a Maxon motor (12V 60W RE-30 dc-motor) combined with a planetary gearhead (Maxon GP32C). Power is transmitted via a steel-wire rope with a diameter of 1.0 mm to the turntable. The rotation of the turntable will cause the rotation of the device, thus the patients’ forearm will move together. And one side of the turntable is connected to a bearing holder with bearing (BEM-6005ZZ; MIYOSHI, Japan) with a low friction. The rotation motion of the device can be measured with a con-tact-less Hall-IC angle sensor (CP-20HB; Midori Precisions Co., Ltd., Japan). For obtaining a smooth control signal, a low-pass Butterworth filter with a cutoff frequency of 5 Hz was used further. The stiffness control part is fixed to a bearing while rotat-ing with the motion of position control part. For the stiffness control part, the variable stiffness behavior is realized with a variable transmission ratio between the elastic elements and the actuator output. A Maxon motor (6V 8W EC-max-16 dc-motor) combined with a planetary gearhead (Maxon GP16C planetary gearbox) was applied into a hypocycloid mecha-nism as shown in (Fig. 7). The hypocycloid gear mechanism
Mechanical Design and Control Method for SEA and VSA-based Exoskeleton Neuroscience and Biomedical Engineering, 2014, Vol. 2, No. 3 145
was designed with a special diameter, thus the rotation mo-tion can be changed to a straight-line motion of the pivot [17].
Fig. (6). VSA-based exoskeleton device design.
Fig. (7). Stiffness adjustment part with the Hypocycloid mecha-nism.
In more details, the rotation of the motor can be changed into a straight-line motion of the pivot point and the ratio
between the elastic elements and the output can be adjusted as shown in (Fig. 8). Thus, the stiffness of the device can be adjusted by the moving distance of the pivot. The stiffness of the device can be calculated by (5),
K= F�
(5)
where the K represents the output stiffness, F is the applied force and�� is the deflection of the output. The output stiff-ness can be further calculated by (6),
K=2kl2 (l-�1)2
l12 cos(2(�-�1)) (6)
where the k is the elastic constant of the spring, l is the whole length that the pivot can move, l1 is the distance that the piv-ot moves (Fig. 8), and the �1 is the rotation angle of the frame (Fig. 7) [17].
Fig. (8). Variable ratio between the elastic elements and the output.
The idea that brings the hypocycloid mechanism into the stiffness adjustable mechanism was first presented by Groothuis SS et al. [17]. Groothuis SS et al. also proved that by moving the pivot point along the lever arm can minimize the involved forces during the change of stiffness. A similar mechanism is also used in HypoSEA [18]. We directly fixed the arm supporter to the output of the stiffness adjustable mechanism while the rotating center of the device is accord-ance with that of the arm supporter. Compared with the SEA-based exoskeleton device, the device becomes more complex. However, an easier position controller is enough, to realize the compliance by just control the movement of the pivot point. This device is also designed for the requirements of home-based rehabilitation training; the device is easy to be fixed to a table before the training. And both the passive and active training can be performed with the same device.The variable stiffness is also adapted to special level of im-pairment of patients by moving the pivot point.
Fig. (5). Constant output impedance.
0 100 200 300 400 500 600 700 800 900 1000-10
0
10
20
30
40
50
60
70
Sampling numbers
Ang
le (d
eg)
ExoskeletonInertial sensor
Applied force
Pivot point
Elastic elements
146 Neuroscience and Biomedical Engineering, 2014, Vol. 2, No. 3 Zhang et al.
4.2. Torque Limiter for Physical Safety of Patients
A torque limiter clutch mechanism is also designed in the VSA-based exoskeleton device as shown in (Fig. 9). Theclutch mechanism is superior to that of the ULERD by usingthe ball rollers. When the interaction torque between the pa-tients and the device is larger than a preset threshold, the torque limiter will be released, thus the device will rotate regardless of the position of the motor. In detailed, the cou-pling is used to connect the axis of motor and torque limiter part. The clamp spring restricts the translation motion. The mechanism is similar with a clutch that the compressed force from the coil spring can force four ball rollers insert into the fillister. When the interaction force becomes larger than the preset threshold, the ball roller will be apart from the fillister, so that the safety of patients could be ensured.
4.3. Evaluation of the Compliance of the Device with Dif-ferent Stiffness
A position controller (PID controller) was used for mov-ing the pivot point to control the output stiffness. The com-pliance variable device is necessary for adapting to a specific level of impairment of patient. Therefore, the compliancewith variable stiffness should be evaluated experimentally. A force sensor (MINI 4/20; BL AUTOTEC, LTD.) was attached to the output link at 91 mm from the center of rota-tion. We manually rotate the frame clockwise and an output deflection was caused. The rotation angle of the frame was measured with the angle sensor as shown in (Fig. 10). An inertial sensor (MTx sensor, Xsens, Enschede, the Nether-lands) was installed on the output part for measuring its rota-tion angle. The measured force with force sensor is relative to the deflection between the frame and the output. We se-lected three parameter (0mm, 4mm, 8mm) which represent distances that the pivot moves from the rotation center and the experimental result is shown in (Fig. 11). It can be found that the measured force increased linearly with the deflection angle, defined as the deviations from its own equilibrium
position. Therefore, for the experimental result, with the in-crease of the selected parameter, the passive joint stiffness was increased. During the rehabilitation training, a suitable stiffness can be selected for a better recovery of impairment of patients.
Fig. (10). Compliance test with different stiffness.
Fig. (11). Relationship between the output deflection and force.
CONCLUSION
Compliant actuator is one of the main requirements for a rehabilitation device. Both the SEA and VSA have been ap-plied to many rehabilitation devices. In this paper, two kinds of exoskeleton devices which realize the compliance with the SEA and VSA respectively are introduced.
The SEA-based exoskeleton device has a relatively sim-ple structure and the compliance was realized with a closed-loop interaction control method. The device can output low impedance with the proposed control method. A needed im-pedance output can also be obtained by setting the virtual impedance. During the mechanical design, there are two pas-sive DoFs designed for solving the misalignment between the human joint and device joints during the training. A torque limiter was designed for ensuring the patients’ safety.
��������������
�����
�������������������
����
�
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
Forc
e (N
)
Deflection (deg)
�
�
0 mm4 mm8 mm
Mechanical Design and Control Method for SEA and VSA-based Exoskeleton Neuroscience and Biomedical Engineering, 2014, Vol. 2, No. 3 147
Compared with it, the VSA-based exoskeleton device can realize inherent hardware compliance. The variable stiffness behavior was realized with a variable transmission ratio be-tween the elastic elements and the actuator output. The hy-pocycloid mechanism was used to move the pivot point along the lever arm for changing the transmission ratio. Es-pecially, for guaranteeing the safety of patients, a torque-limiter mechanism was designed. The performance that the device can change the stiffness was evaluated. With a con-stant stiffness, the output impedance (force) is relative to the deflection of elastic elements.
Overall, the VSA-based device does not need a complex control loop to realize the compliance; however, an addition-al mechanism for adjusting stiffness is needed. For obtaining a better recovery of impairment limb, a suitable stiffness can be selected. By comparison, the structure of SEA-based exo-skeleton device is relative simple; however, the output stiff-ness of the SEA-based exoskeleton device is constant. The adequate and inadequate of two kinds of devices for rehabili-tation training will be studied in our future work with more clinical evidences.
CONFLICT OF INTEREST
The author(s) confirm that this article content has no con-flict of interest.
ACKNOWLEDGEMENTS
This research is partly supported by National High Tech. Research and Development Program of China (No.2015AA043202), JSPS KAKENHI Grant Number 15K2120 and Kagawa University Characteristic Prior Research Fund 2014.
REFERENCES[1] American Heart Association. Heart disease and stroke statistics
2010 update: a report from the American Heart Association. Circu-lation 2010; 121: e46-e215.
[2] Barreca S, Wolf SL, Fasoli S, Bohannon R. Treatment interven-tions for the paretic upper limb of stroke survivors: a critical re-view. Neurorehabil Neural Repair 2003; 17(4): 220: 226.
[3] Song Z, Guo S, Fu Y. Development of an Upper Extremity Motor Function Rehabilitation System and an Assessment System. Internat J Mechatronics Automation 2001; 1(4): 19-28.
[4] Loureiro R, Amirabdollahian F, Topping M, Dressen B, Harwin W. Upper Limb Robot Mediated Stroke Therapy – GENTLE/s Ap-
proach. Autonomous Robot 2003; 15(1): 35:51. [5] Zhang S, Guo S, Pang M, Qu M. Training Model-based Master-
slave Rehabilitation Training Strategy using the Phantom Premium and Exoskeleton Device. Proceedings of the 2014 ICME Interna-tional Conference on Complex Medical Engineering; 2014 June 26-29; 105:110.
[6] Stienen A, Hekman E, van der Helm F, van der Kooij H. Self-Aligning Exoskeleton Axes Through Decoupling of Joint Rotations and Transla-tions. IEEE Transactions on Robotics 2009; 25(3): 628: 633.
[7] Nef T, Mihelj M, Colombo G, Riener R. ARmin - robot for rehabil-itation of the upper extremities. Proceedings of 2006 IEEE Interna-tional Conference on Robotics and Automation; 2006 May 15-19; Orlando, American, pp. 3152: 3157.
[8] Tsagarakis NG, Caldwell DG. Development and Control of a ‘Soft-Actuated’ Exoskeleton for Use in Physiotherapy and Training. Au-tonomous Robots 2003; 15(1): 21-33.
[9] Oblak J, Cikajlo I, Matjacic Z. Universal haptic drive: A robot for arm and wrist rehabilitation. IEEE Transactions on Neural Systems and Rehabilitation Engineering 2010; 18(3): 293-302.
[10] Vitiello N, Lenzi T, Roccella S, Rossi SMMD, Cattin E, Giovac-chini F, et al. NEUROExos: A powered elbow exoskeleton for physical rehabilitation. IEEE Transactions on Robotics 2013: 29(1): 220:235.
[11] Groothuis SS, Rusticelli G, Zucchelli A, Stramigioli S, Carloni R. The Variable Stiffness Actuator vsaUT-II: Mechanical Design, Modeling, and Identification. IEEE/ASME Transactions on Mecha-tronics 2014; 19(2): 589:597.
[13] Williamson MM. Series Elastic Actuator. A.I. Technical Report 1995; 1524.
[14] Song Z, Guo S, Pang M, Zhang S, Xiao N, Gao B, Shi L. Imple-mentation of Resistive Training Using an Upper-Limb Exoskeleton Rehabilitation Device for Elbow Joint. J Medical Biol Eng (JMBE) 2014; 34(2):188:196.
[15] Morasso P, Casadio M, Giannoni P, Masis L, Sanguineti V, Squeri V, Vergaro E. Desirable features of a “humanoid” robot-therapist. Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2009 Sept.;2418:2421.
[16] Song Z, Guo S, Xiao N, Gao B, Shi L. Implementation of Human-Machine Synchronization Control for Active Rehabilitation Using an Inertia Sensor. Sensors 2012; 12(12):16046:16059.
[17] Groothuis SS, Rusticelli G, Zucchelli A, Stramigioli S, Carloni R. The vsaUT-II: A Novel Rotational Variable Stiffness Actuator. Proceeding of the 2012 IEEE International Conference on Robotics and Automation (ICRA); 2012 May: 3355:3360.
[18] Thorson I, Caldwell D. A nonlinear series elastic actuator for high-ly dynamic motions. Proceeding of the 2011 IEEE/RSJ Internation-al Conference on Intelligent Robots and Systems (IROS), 2011 Sept: 390:394.
Received: March 03, 2015 Revised: May 04, 2015 Accepted: May 14, 2015
