ICCAS2004 August 25-27, The Shangri-La Hotel, Bangkok, THAILAND
1. INTRODUCTION
The people with lower-limb disability perform rehabilitation to prevent diseases caused by their bedridden state or to maintain their mental and physical health. In their usual life or during rehabilitation, they use some kinds of walking stand or assist tools. But because most those things require user to have some physical strength, it is hard to use for the people with weak muscle. In our last research, we developed the prototype of the walking assist system that could provide easy and joyful walking or rehabilitation to the people with lower-limb disability with less muscle strength.
The prototype had original closed links structure and employed four servo motors, two degrees of freedom on each leg. With the original links structure, user could perform rehabilitation without attention of falling down during walking.
Because of the simplicity and the robustness as input device, we adapted four limit switches as HMI of the prototype. Limit switches were attached on foot and waist. When user wanted to walk, he could easily trigger motions just by switching on.
But because the walking patterns were pre-defined and just triggered by limit switches, once one motion was started, user could not change speed or stride. User could easily interface with the walking assist system, but could not actively control the system with limit switches. In this research, we propose the advanced HMI for the walking assist system by using two kinds of input devices.
EMG(electromyograph) has been used to control artificial limb, and there are many researches about it [1] [2] [3].
Because it is readily affected by the condition of the skin, EMG is attached to the skin directly using some special gel.
Nowadays stiffness sensors are more efficiently used as HMI to overcome those restrictions of EMG. Komiya measured muscle stiffness by measuring reacted force [4]. Yamamoto adapted muscle stiffness sensor to amplifying power [5]. And
Moromugi developed muscle stiffness sensor that could be more practical under normal condition [6]. Because it is hard to be affected by the skin condition and can be attached even through clothes. We decided to adapt Moromugi’s Stiffness sensor, as the HMI of the walking assist system.
2. SYSTEM CONFIGURATION 2.1 Mechanical structure
In this chapter, we propose new mechanism (Ver.2) of the walking system for the lower-limb disability. Fig.1, Fig.2 show the present mechanism (Ver.1), and Fig.3 shows the proposed mechanism of the walking assist system. Basically both mechanisms have same links structure that makes foot to maintain parallel to the ground.
ٻ ٻ
Fig. 1 Walking Assist System: Ver.1
Development of advanced walking assist system employing stiffness sensor
Seok-Hwan Kim*, Moromugi Shunji** and Takakazu Ishimatsu***
*Dept. of Mechanical system, Nagasaki University., Nagasaki, Japan (Tel : +81-95-819-2511; E-mail: shkim@welcome.mech.nagasaki-u.ac.jp)
** Dept. of Mechanical system, Nagasaki University., Nagasaki, Japan (Tel : +81-95-819-2511; E-mail:shunji@welcome.mech.nagasaki-u.ac.jp)
** Dept. of Mechanical system, Nagasaki University., Nagasaki, Japan (Tel : +81-95-819-2508; E-mail: ishi@welcome.mech.nagasaki-u.ac.jp)
Abstract: Many walking stands, and assisting tools have been developed for the people with low-limb disability to prevent diseases from bedridden state and to help them walk again. But many of those equipments require user to have some physical strength or balancing ability. In our last research, we developed walking assist system for the people with lower-limb disability.
With the system, user can be assisted by actuators, and do not have to worry about falling down. The system adapted the unique closed links structure with four servomotors, three PICs as controller, and four limit switches as HMI (human man interface). We confirmed the adaptability of the system by the experiment. In this research, Muscle Stiffness Sensor was tested as the advanced HMI for walking assist system, and confirmed the adaptability by the experiment. As Muscle Stiffness Sensor can attain the muscle activity, user can interface with any device he want to control. Experimental result with Muscle Stiffness sonsor showed that user could easily control the walking assist system as his will, just by changing his muscle strength.
Keywords: Walking assist system, rehabilitation, lower-limb disability, HMI(human man interface), EMG(electromyograph)
1638
At Ver.1, all motors are assembled around waist to reduce the load to motors, and two links are added to translate power to shank. Because of this structure, the center of gravity goes up and total weight is increased. And two rear links make it difficult to sit down and up, and take large space. And each two joint of the closed links are parallel to ground, also make it difficult to sit down and up. From the point of precise position control, the reducer with three steps of spur gear causes some backlash.
To improve such problems of Ver.1, Ver.2 structure that has following properties is proposed, where four flat DC motors are used to minimize useless space, and four harmonic drives are used to minimize backlash. A motor for shank is attached to the knee, to go down the center of gravity. So two rear links can be removed and total weight is decreased. And by positioning each two joints of the closed links declined to ground, joint size can be minimized. And a gas spring is employed to knee to assist knee motor. In standing posture, this gas spring assist knee motor, but for opposite direction, it obstructs knee motor.
ٻ Fig. 2 Man with walking Assist System: Ver.1 When man rotates his shank back, gas spring can be obstacle to motor. But at that time, the loads to motor are only shank weight and gas spring’s obstructing force. Therefore, load to the knee motor can be reduced. And because gas spring does damping function, it can reduce shock during the walking motion.
Fig. 3 Proposed Mechanism: Ver.2G
2.2 HMI (Human Machine Interface)
It is important to detect user’s intent precisely to revive real walking motion. The purpose of this research is to help the aged people or people with lower-limb disability to walk again or to take rehabilitation with joy and safety. To accomplish this purpose, HMI is important to control the system safely and actively.
What kind of HMI is good for the walking assist system?
At last research, we used limit switches to trigger each step of walking motion. Limit switch is simple and robust, so it is a good interface to trigger start or stop motion, but hard to control speed or walking pattern. For the active and precise control, HMI of the walking assist system must detect more complex signal.
Up to now, EMG has been widely used to detect muscle’s state. When user strengthens his muscle, EMG detects its voltage change. This method is widely used to control artificial limbs, but easily affected by noise or skin’s condition change. Nowadays, some kinds of stiffness sensing systems are used to detect muscle stiffness change, and Dr. Moromugi of Nagasaki Univ. developed Button type’s Muscle Stiffness Sensor that can separate muscle stiffness (force) from the pushing force on the outer part of the sensor and. Fig.4 shows the structure of Muscle Stiffness Sensor.
Fig. 4 Structure of Muscle Stiffness Sensor
ٻ ٻ
Fig. 5 Muscle Stiffness Sensor 2.3 Walking assist system with MSS
Fig.6 shows the control flow chart of the walking assist system. MSS and limit switches detect user’s intent to walk, and main PIC send command to each PIC that really makes command pulses for motor control. HMI always checks MSS and limit sensors, and send its data to main PIC. Limit sensors are used to start or stop each step of walking motion, and MSS to change walking speed or walking pattern. With two kinds of sensing systems, user can change walking motion (speed or walking pattern) at any time he wants.
Skin Button
Disk Pressure Sensor
1639
Fig.6 Walking Assist System with Muscle Stiffness Sensor
3. EXPERIMENT
Experiment was executed to verify adaptability of Muscle Stiffness Sensor. Both left and right legs are symmetrical structure, so we experimented with one side of the system.
Two muscle stiffness sensors were used to control two motors of a leg of the walking assist system.
3.1 Calibrating Muscle Stiffness Sensor
Dr.Moromugi’s Muscle Stiffness Sensor is composed of two pressure sensors that transmit muscle stiffness to voltage change. Because output voltage of the Muscle Stiffness Sensor is very small, OP-Amp is used to amplify. To get the property graph of the sensor, we imposed four loads (0, 1, 2, 3 Kg) on the sensor. And through experiments, we calibrated proportional gains of two pressure sensors of a Muscle Stiffness Sensor. Table.1 shows the property data of the two muscle sensors.
.
Gain
Button1 Entire1 Button2 Entire2 Load
[Kgf]
0.36 0.32 0.35 0.35
0ٻ 0 0 0 0
1ٻ 0.11 0.11 0.07 0.16
2ٻ 0.3 0.3 0.24 0.36
3ٻ 0.5 0.5 0.48 0.48
ٻ ٻ sensor output Voltage = table/gain [V]
Table. 1 Data of Muscle Stiffness Sensors With Table.1 we derived four inverse approximate graphs with the least square method. When main PIC gets data from Muscle Stiffness Sensor, we can calculate loads by using inverse approximate equations and gains of Table.1. Fig.7 shows four inverse approximate graphs and equations.
ybx,ybx: load on button of the sensor 1, 2 ytx,ytx: entire load on the sensor 1, 2
0 0 .5 1 1 .5 2 2 .5 3 3 .5
0 0 .1 0 .2 0 .3 0 .4 0 .5 0 .6
senso r o utp ut
load[kg]
Butto n1 Entire1 Butto n2 Entire2
x x
yb1 5.1079 28.501 (1) x
x
yt1 5.1079 28.501 (2) x
x
yb2 11.057 211.504 (3) x
x
yt2 1.7379 25.289 (4) Fig. 7 Inverse Approximating Relationship
3.2 Control with Muscle Stiffness Sensors
In the main PIC of the controller, we can calculate load on Muscle Stiffness Sensors by using inverse approximate equations and gains of Table.1. If
yb is load on button, yt is load on total sensor system,
t
b y
y / will change according to muscle stiffness state. When user strengthens his muscle,
t
b y
y / also increases. With this property of Muscle Stiffness Sensor, user can control torque or rotating angle of the walking assist system.
In this re search, we used two Muscle Stiffness Sensors to control rotating angle of two motors of one leg. One is attached to the muscle of thigh, and the other to the muscle of shank. Two muscles are deliberately chosen, so that two measuring points to effect each other as less as possible. Fig.6 shows attach points.
㪤㫆㫋㫆㫉㩷㪉㩷㪽㫆㫉㩷㫊㪿㪸㫅㫂
㪤㫆㫋㫆㫉㩷㪈㩷㪽㫆㫉㩷㪫㪿㫀㪾㪿
㪪㫀㪻㪼 㪝㫉㫆㫅㫋
㪤㫌㫊㪺㫃㪼㩷㪪㫋㫀㪽㪽㫅㪼㫊㫊㩷㪪㪼㫅㫊㫆㫉
㪪㪼㫅㫊㫆㫉㪈 㪪㪼㫅㫊㫆㫉㪉
Fig. 8 Sensing Points on the Leg 3.3 Result
Because muscle stiffness always changes, we measure its maximum and minimum strength at the beginning of each experiment. Selected two muscles were not independent to each other, so we used sensitivity factors to reduce relations.
The main PIC controller calculated A/D data, and made three states commands [go, stay, back]. Other two PICs executed the commands from the main PIC.
Fig.9 shows normalizing process of Muscle Stiffness Sensor.
1640
- 0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0 0.5 1 1.5 2 2.5 3
time(sec)
voltage x gain (V
FButton1 FEntire1 FButton2 FEntire2
Fig.9 Normalization of Muscle Stiffness Data Fig.10, Fig.11 show torques and speeds of two motors are changed according to changes of Muscle Stiffness. When thigh or shank reaches at pre-defined range limits, even though Muscle Stiffness data is enough high or low, joint angle and torque does not change.
㪄㪇㪅㪏 㪄㪇㪅㪍 㪄㪇㪅㪋 㪄㪇㪅㪉 㪇 㪇㪅㪉 㪇㪅㪋 㪇㪅㪍 㪇㪅㪏
㪐 㪐㪅㪉 㪐㪅㪋 㪐㪅㪍 㪐㪅㪏 㪈㪇
㪫㫀㫄㪼㩷㩿㫊㪼㪺㪀
㪰㪹㪆㪰㪼㪲㪴㪃㩷㪭㫆㫃㫋㪸㪾㪼㩷㫏㩷㪾㪸㫀㫅㪲㪭㪴
㫐㪹㪈㪆㫐㪼㪈 㪤㪈㪪 㪤㪈㪫
Fig.10 Motor Control for Thigh
- 0. 8 - 0. 6 - 0. 4 - 0. 2 0 0. 2 0. 4 0. 6 0. 8
9 9. 2 9. 4 9. 6 9. 8 10
Time ( sec)
Yb/Ye[], Voltage x gain[V]
yb2/ ye2 M2S M2T
Fig.11 Motor Control for Shank
4. CONCLUSION
We proposed new mechanism for the walking assist system for the lower-limb disability, and adapted Dr. Moromugi’s Muscle Stiffness Sensor system as HMI. We did the experiment to verify adaptability of Muscle Stiffness Sensor system to the walking assist system. Because both left and right legs were symmetrical, only one leg with two motors was experimented with two Muscle Stiffness Sensors. Experiment showed that user could control two motors with two Muscle Stiffness Sensors simultaneously, even though two muscles effected each other during some motions. The experiment result showed that Muscle Stiffness Sensor could the good HMI for the walking assist system for the lower-limb
disability.
REFERENCE
[1] Saridis, N., and Gootee, T., "EMG Pattern Analysis and Classification for a Prosthetic Arm," IEEE Trans.
Biomedical Engineering, BME-29(6), pp.403-412, 1982 [2] Doerschuck, P.C., Guftafson, D.E., and A. S. Willsky,
“Upper extremity limb function discrimination using EMG signal analysis,” IEEE Trans. Biomedical Engineering, BME-30(1), pp.18-28, 1983
[3] Graupe, D., Salahi, J., Zhang, D., "Stochastic Analysis of Myoelectric Temporal Signatures for Multifunctional Single-Site Activation of Prostheses and Orthoses,"
Journal of Biomedical Engineering, Vol. 7, pp.18-29, 1985
[4] Komiya, H., Maeda, J. and Takemiya, T., “New functional measurement of muscle stiffness in human,”
Advances in Exercise and Sports Physiology, 2(1), pp.
31-38, 1996.
[5] Yamamoto, K., Hyodo, K. and Matsuo, T., “Powered Suit for Assisting Nurse Labor Employing Muscle Sensor and Sliding Rotary Actuator,” Proc. 5th International Symposium on Fluid Control, Measurement and Visualization, Hayama, Japan, vol.1, pp.497-501, Sep.1997.
[6] Moromugi, Okamoto, “Device for Assisting Graspin g Function(2nd Report: Maneuverability Evaluation)”, ICCAS2003, Medical Robots, pp.2666-2669,Oct.2003.
ٻ
