STUDY OF A 4DOF UPPE

-LIMB POWE

-ASSIST

EXOSKELETON WITH PERCEPTION-ASSIST

Second Stage of Power-Assist

Kazuo Kiguchi and Manoj Liyanage

Graduate School of Science and Engineering, Saga University, 1 Honjomachi, Saga, Japan

Keywords: Power-Assist, Perception-Assist, Exoskeletons, Robots, EMG.

Abstract: As a second stage of the research on power-assist exoskeleton systems, this paper presents a new concept of

an upper-limb power-assist exoskeleton that can assist physically weak persons in performing their daily

activities. The proposed exoskeleton assists not only the motion of the user but also the perception of the

user by using sensors. In the proposed power-assist method, the assisted user’s motion can be modified

based on the environmental information obtained by the sensors if problems are found in the user’s motion.

The effectiveness of the proposed concept is evaluated by performing experiments.

1 INTRODUCTION

Decrease in birthrate and increase of percentage of

aged people are progressing in several countries. In

these societies, the shortage of nursing people has

become a serious problem. Many robotic systems

such as power-assist robots (Kiguchi, et al., 2001-

2007; Rosen, et al., 2001; Kawamoto and Sankai,

2005; Naruse, et al., 2004; Sasaki, et al., 2004;

Guizzo and Goldstein, 2005; Vukobratovic, 1975) have

been proposed to cope with this problem. We have

proposed power-assist exoskeletons to assist the

upper-limb motion of physically weak persons such

as disabled, injured, and/or elderly persons since the

upper-limb motion is important for daily activities.

The conventional exoskeletons only assist the

motion of the user (Kiguchi, et al., 2001-2007). As a

second stage of the research on power-assist

exoskeleton robot systems, this paper proposes a

new concept of an upper-limb power-assist

exoskeleton in order to assist physically weak

persons in performing their daily activities.

In the conventional power-assist exoskeletons,

the motion of the user is supposed to be assisted in

accordance with the user’s motion intention. The

skin surface electromyogram (EMG) is often used to

detect the user’s motion intention (Fukuda, et al.,

2003) since it directly reflects the user’s muscle

activity. Therefore, information of the EMG signals

and/or force sensors is often used to predict the

user’s motion intention in the conventional power-

assist exoskeletons (Kiguchi, et al., 2001-2007).

However, in the case of physically weak persons, the

perception ability is also poor sometimes. For

example, such persons sometimes unknowingly trip

over small obstacles because of their poor perception

ability. Therefore, it is important to assist the

sensing ability of physically weak persons by using

sensors on robotic exoskeletons. Any sensor such as

ultrasonic sensors, infrared sensors, and/or CCD

sensors can be a candidate for the sensors for the

perception-assist.

In the proposed power-assist method, the assisted

user’s motion can be modified based on the

environment information obtained by the sensors if

the exoskeleton detects some problems in the user’s

motion. For example, if the exoskeleton notices that

the user’s hand is going to collide with an obstacle,

an additional assist force including the power-assist

force is provided to the user’s motion in order to

avoid the collision between the user and the obstacle.

On the other hand, the exoskeleton attempts to guide

the user’s hand to an object when the exoskeleton

notices that the user is going to grasp that object, but

his hand is not moving along the correct trajectory.

If the modified motion by the exoskeleton is

different from the user’s intended motion, the

exoskeleton changes its strategy in accordance with

the user’s motion intention. When motion

modification is not required (i.e., when there is no

164

Kiguchi K. and Liyanage M. (2008).

STUDY OF A 4DOF UPPER-LIMB POWER-ASSIST EXOSKELETON WITH PERCEPTION-ASSIST - Second Stage of Power-Assist.

In Proceedings of the First International Conference on Biomedical Electronics and Devices, pages 164-169

DOI: 10.5220/0001050801640169

 SciTePress

problem in the user’s motion), the control method is

the same as the conventional EMG-based power-

assist method (Kiguchi, et al., 2007).

In this study, the proposed power-assist method

is applied to a 4DOF upper-limb power-assist

exoskeleton. The exoskeleton assists shoulder

flexion/ extension, shoulder horizontal flexion/

extension, elbow flexion/ extension, and forearm

supination/pronation motion.

2 UPPER-LIMB POWER-ASSIST

EXOSKELETON

In order to assist 4DOF upper-limb motion, a power-

assist exoskeleton (Fig. 1) was developed (Kiguchi,

et al., 2007). It mainly consists of a shoulder motion

support part, an elbow motion support part, and a

forearm motion support part. The exoskeleton

system can be installed on a mobile wheel chair.

Figure 1: 4DOF upper-limb power-assist exoskeleton.

Usually, the movable range for the human

shoulder is 180º in flexion, 60º in extension, 180º in

abduction, 75º in adduction, 100-110º in internal

rotation, and 80-90º in external rotation. The limit on

the range of forearm pronation-supination motion is

50-80º in pronation and 80-90º in supination, and

that on the elbow flexion-extension motion is 145º in

flexion and -5º in extension. Considering the

minimally required motion in everyday life and the

safety of the user, the shoulder motion of the 4DOF

exoskeleton is limited to 0º in extension and

adduction, 90º in flexion, and 90º in abduction. The

limit on its forearm motion is decided as 50º in

pronation and 80º in supination, and that on the

elbow motion is decided as 120º in flexion and 0º in

extension.

3 POWER-ASSIST WITH

PERCEPTION-ASSIST

In the conventional power-assist robot systems, the

user’s motion intention is estimated in real-time;

subsequently, the estimated motion is assisted by the

power-assist robot systems (Kiguchi, et al., 2001-

2007). However, the perception ability is also

deteriorated sometimes in the case of physically

weak persons. Therefore, there is a possibility of

colliding with an obstacle, tripping over a small

obstacle, or failing to grasp an object even though

the motion is assisted according to the user’s

intention. In this study, the perception of the

environment is also assisted by the exoskeleton.

(a) Human behavior cycle.

(b) Conventional power-assist cycle.

Figure 2: Power-assist cycles.

The concepts of the human behavior cycle, the

conventional power-assist cycle, and the proposed

power-assist cycle are depicted in Fig. 2 in order to

show the difference among them.

In the proposed power-assist process, when the

user’s motion is not interacting with the environment,

the power assist is the same as the conventional

power assist. Moreover, when the user interacts with

the environment properly, the power-assist method

continues to be the same as the conventional power-

assist method (Fig. 3(a)). However, if the

exoskeleton infers that the user is attempting to grab

an object and the user has miscalculated the position

of the object, motion modification is carried out to

ensure the correct hand trajectory to the object, as

STUDY OF A 4DOF UPPER-LIMB POWER-ASSIST EXOSKELETON WITH PERCEPTION-ASSIST - Second Stage

of Power-Assist

165

shown in Fig. 3(b). If the modified motion is correct

(i.e., the decision of the exoskeleton is correct), then

the ordinal power assist is performed after the

motion modification, as shown in Fig. 3(a).

However, if the modified motion is wrong (i.e., the

decision of the exoskeleton is wrong), the user

attempts to reject it (sometimes unconsciously) as

shown in Fig. 3(c). This rejection can automatically

be detected by the exoskeleton by monitoring the

user’s EMG signals. Sub sequently, the exoskeleton

changes its strategy, and another motion

modification is carried out in order to avoid a

collision with the object.

(a) Ordinal power-assist.

(b) Motion modification.

Figure 3: Motion modification.

4 CONTROL METHOD

4.1 EMG

The EMG signals are used as the main input signals

in order to control the exoskeleton in accordance

with the user’s motion intention. Since it is difficult

to use the raw EMG signal as input information for

the controller, the root mean square (RMS) value of

the signal is calculated to extract the feature from the

signal. The equation for the RMS value is written as:

∑

RMS

(1)

where v

is the voltage value for the i

sampling and

N is the number of samples in a segment. In this

study, the number of the sample is set to be 100 and

the sampling time is 500 μsec.

When a certain motion is performed, the EMG

signals of the related muscles show a unique pattern.

Therefore, since the magnitude of the RMS of the

EMG signal indicates the activity level of the

muscles, the upper-limb motion of the user can be

predicted by monitoring the EMG signals of certain

muscles of the user.

In order to predict the 4DOF motion, the EMG

signals from the related muscles of 12 locations are

measured in this study (Kiguchi, et al., 2001-2007).

4.2 EMG-Based Control

The basic architecture of the controller is depicted in

Fig. 4. The controller basically consists of a power-

assist part and a perception-assist part. The power-

assist part consists of three stages (first stage: input

signal selection stage; second stage: posture region

selection stage; and third stage: neuro-fuzzy control

stage). This power-assist part is basically the same

as the conventional EMG-based controller (Kiguchi,

et al., 2007). In the first stage of the power-assist

part, the EMG-based control or the wrist-sensor-

based control is applied in accordance to the muscle

activity levels of the user. In the second stage of the

power-assist part, proper neuro-fuzzy controllers are

selected according to the shoulder and elbow angle

regions. In the third stage of the power-assist part,

the torque required for each joint motion assist is

calculated by using the selected neuro-fuzzy

controllers.

Figure 4: Controller architecture.

In the perception-assist part, the sensor

information and the estimated user’s motion

intention (the output of the neuro-fuzzy controllers)

are used to decide the motion to be modified in this

part. Sensors such as ultrasonic sensors, infrared

sensors, and/or CCD sensors can be used to detect

the objects in the environment. In this study, an

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

166

ultrasonic sensor [FW-H10R, Keyence] is applied to

detect the objects.

The force vector of the assisting motion at the

user’s hand can be calculated on the basis of the

estimated torque (the output of the neuro-fuzzy

controllers) from the EMG signals in the third stage

of the power-assist part. Since the estimated force

vector contains noise and an estimated error, it is

averaged with the estimated force vector data in the

past. The relationship between the force vector at the

user’s hand and the joint torque vector of the user’s

upper-limb is written as:

−

(2)

where F is the force vector at the user’s hand

(averaged by using the past data), τ is the joint

torque vector of the user’s upper-limb, and J is the

Jacobian matrix. The estimated force vector of the

user directly indicates the user’s motion intention.

4.3 Perception-Assist with Ultrasonic

Sensor

When a user is moving his/her arm toward an object

in order to grab it, the trajectory of the hand (tip of

the arm) is the almost a straight line toward the

object (Flash and Hogan, 1985). Therefore, the

change in the distance of the tip of arm and the

reduction in the distance between the tip of arm and

the object are supposed to be the same. The change

in the distance of the tip of arm is calculated by

using the kinematics of the exoskeleton. The

reduction in the distance between the hand and the

object is calculated by using the ultrasonic sensor.

When the arm is moving toward the object, these

two values come close to each other but vary in a

particular range. This range is determined based on

the experimental results. It is important to select this

range to be as narrow as possible in order to identify

the trajectory of the arm more accurately.

If the exoskeleton identifies that the trajectory

required for the user is different from the current

user’s trajectory, then the exoskeleton attempts to

modify the trajectory by applying an additional force

at the tip of the arm. If the estimated force vector at

the user’s hand is changing to the modified

trajectory, the exoskeleton assumes that the motion

modification strategy as correct. If the user’s hand is

moving along the correct trajectory to grab the

object, no motion modification is provided to the

user’s motion. However, if the estimation force

vector at the user’s hand is changing to a direction

opposite to the motion modification direction, the

exoskeleton assumes that the motion modification

strategy is not correct and attempts to determine

another possibility. If the exoskeleton can not

determine another possibility, then it simply

performs the conventional power-assist.

5 EXPERIMENT

The experimental set-up is shown in Fig. 5. As

shown, a plastic bottle was used as the object. In the

experiment, upper-limb motion was performed

toward the object. Two interface boards (RIF–171–1

and JIF–171–1) are used to process the A/D

operations of potentiometer signals, force sensor

signals, EMG signals, and ultrasonic sensor signals

and also to process the D/A operations required to

send the calculated torque commands back to the

motor drivers to control the motors. The measured

EMG signals are amplified by the EMG amplifier

before sending them to the interface board. The

motor torque commands are calculated in the PC and

then sent to the four motor drivers to operate four

motors. The output of the ultrasonic sensor is sent to

the RIF–171–1 interface board and then processed

with the same frequency as the other signals (i.e.,

2,000 Hz frequency).

Figure 5: Experimental setup.

Three kinds of experiments were performed on

the same subject in order to evaluate the

effectiveness of the perception-assist. In the first

experiment, the subject attempted to move the hand

toward the object to grab it along the correct hand

trajectory. In the second experiment, the subject also

attempted to move the hand toward the object to

grab it, but deliberately along a wrong trajectory. In

the third experiment, the subject attempted to move

the hand forward to avoid colliding with the object.

In this experiment, a wrong trajectory (i.e., the

trajectory for which the hand collides with the

object) is generated deliberately.

Figure 6 shows the results of the first experiment.

The hand trajectory determined by the exoskeleton,

the distance to the object measured by the ultrasonic

STUDY OF A 4DOF UPPER-LIMB POWER-ASSIST EXOSKELETON WITH PERCEPTION-ASSIST - Second Stage

of Power-Assist

167

sensor, the change in the position of the sensor, the

raw EMG signal of the deltoid – anterior part, and

the RMS value are shown in Fig. 6 (a). The

estimated force vector at the user’s hand (calculated

from the output of the neuro-fuzzy controllers), the

combined force vector (modified force vector), the

assisted force vector (additional force for the motion

modification), and the distance to the object

measured by the ultrasonic sensor are shown in Fig.

6 (b). These experimental results show that the

exoskeleton effectively performs the power-assist

(conventional power-assist) on the basis of the user’s

motion intention, when there are no problems in the

user’s motion.

(a) EMG signal.

(b) Motion modification.

Figure 6: Experimental results of the first experiment.

Figure 7 shows the results of the second

experiment. During the interval (a) in Fig. 7, the

motion modification was performed to change the

trajectory of the user’s hand to the correct trajectory

to proceed toward the object since the exoskeleton

determined that the hand trajectory of the user is

different from the estimated one. Since the decision

of the exoskeleton was correct, the ordinal power-

assist (power-assist without any motion

modification) was performed until the user grasped

the object (the interval (b)) after that.

Figure 8 shows the results of the third

experiment. During the interval (a) in Fig. 8, the

motion modification was performed to change the

trajectory of the user’s hand to a trajectory leading to

the object since the exoskeleton determined that the

trajectory of the user’s hand was different from the

estimated one. During the interval (b) in Fig. 8, the

exoskeleton determined that its strategy was wrong

and changed it to modify the trajectory of the user’s

hand to avoid a collision with the object. Since the

second decision of the exoskeleton was correct, the

ordinal power assist (power assist without any

motion modification) was performed.

(a) EMG signal.

(b) Motion modification.

Figure 7: Experimental results of the second experiment.

These experimental results show the

effectiveness of the proposed power-assist method

with perception-assist.

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

168

(a) EMG signal.

(b) Motion modification.

Figure 8: EMG of elbow biceps (short head) with power

and perception assist.

6 CONCLUSIONS

A new concept of a power-assist exoskeleton that

assists not only the motion but also the perception of

the user by using sensors is proposed. In the

proposed power-assist method, the user motion is

modified by the exoskeleton if it is necessary,

although the conventional power-assist robot never

modifies the user motion. The effectiveness of the

proposed power-assist exoskeleton was verified by

performing experiments.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the support

provided for this research by Japan Society of

Promotion of Science (JSPS) Grant-in-Aid for

REFERENCES

Flash, T., Hogan, N., 1985. The coordination of Arm

Movements: An Experimental Confirmed

Mathematical Model, Journal of Neuroscience, vol.5,

pp.1688-1703.

Fukuda, O., Tsuji, T., Kaneko, M., Otsuka , A., 2003. A

Human-Assisting Manipulator Teleoperated by EMG

Signals and Arm Motions, IEEE Trans. on Robotics

and Automation, vol. 19, no. 2, pp.210-222.

Guizzo, E., Goldstein, H., 2005. The Rise of the Body

Bots”, IEEE Spectrum, vol.42, no.10, pp.42-48.

Kawamoto, H., Sankai, Y., 2005. Power Assist Method

Based on Phase Sequence and Muscle Force Condition

for HAL, Advanced Robotics, vol.19, no.7, pp.717-

734.

Kiguchi, K., Esaki, R., Fukuda, T., 2005. Development of

a Wearable Exoskeleton for Daily Forearm Motion

Assist, Advanced Robotics, vol.19, no.7, pp.751-771.

Kiguchi, K., Imada, Y., Liyanage, M., 2007. EMG-Based

Neuro-Fuzzy Control of a 4DOF Upper-Limb Power-

Assist Exoskeleton, Proc. of 29

Annual International

Conf. of the IEEE Engineering in Medicine and

Biology Society.

Kiguchi, K., Iwami, K., Yasuda, M., Watanabe, K.,

Fukuda, T., 2003. An Exoskeletal Robot for Human

Shoulder Joint Motion Assist, IEEE/ASME Trans. on

Mechatronics, vol.8, no.1, pp.125-135.

Kiguchi, K., Kariya, S., Watanabe, K., Izumi, K., Fukuda,

T., 2001. An Exoskeletal Robot for Human Elbow

Motion Support – Sensor Fusion, Adaptation, and

Control, IEEE Trans. on Systems, Man, and

Cybernetics, Part B, vol.31, no.3, pp.353-361.

Kiguchi, K., Tanaka, T., Fukuda, T., 2004. Neuro-Fuzzy

Control of a Robotic Exoskeleton with EMG Signals,

IEEE Trans. on Fuzzy Systems, vol.12, no.4, pp.481-

490.

Kiguchi, K., Yamaguchi, T., Sasaki, M., 2006.

Development of a 4DOF Exoskeleton Robot for

Upper-limb Motion Assist, Proc. of 2006 ASME/JSME

Joint Conf. on Micromechatronics for Information and

Precision Equipment, S10_03.

Naruse, K., Kawai, S., Yokoi, H., Kakazu, Y., 2004.

Design of Wearable Power-Assist Device for Lower

Back Support, Journal of Robotics and Mechatronics,

vol.16, no.5, pp.489-496.

Rosen, J., Brand, M., Fuchs, M., Arcan, M., 2001. A

Myosignal-Based Powered Exoskeleton System, IEEE

Trans. on System Man and Cybernetics, part A, vol.

31, no. 3, pp. 210 - 222.

Sasaki, D., Noritsugu, T., Takaiwa, M., 2004.

Development of Wearable Power-Assist Device for

Lower Back Support, Journal of Robotics and

Mechatronics, vol.16, no.5, pp.497-503.

Vukobratovic, M., 1975. Legged Locomotion Robots and

Anthropomorphic Mechanisms, Mihailo Pupin

Institute, Belgrade.

STUDY OF A 4DOF UPPER-LIMB POWER-ASSIST EXOSKELETON WITH PERCEPTION-ASSIST - Second Stage

of Power-Assist

169