DEVELOPMENT OF A 7-DOF POWER ASSISTANT ROBOT

DAO Thanh Liem**,1, DOAN Ngoc Chi Nam**, Kyoung Kwan AHN*,2, Jihwan LEE**,

Hyung Gyu PARK**, and Soyoung LEE**

* University of Ulsan, School of Mechanical and Automotive engineering, S. Korea

93 Daehakro, Namgu, Ulsan, 680-749, S. Korea

2(E-mail: kkahn@ulsan.ac.kr)

** University of Ulsan, Grad. School of Mechanical and Automotive engineering, S. Korea

93 Daehakro, Namgu, Ulsan, 680-749, S. Korea 1(Email: daothanhliem@gmail.com)

ABSTRACT

Due to their high power output and reliability, robots can replace humans in most modern industrial tasks associated

with heavy loads. This paper presents the development of a so-called power assistant robot (PAR) for use in industrial

applications. The system is a 7-DOF redundant robot hand which bases on a new type of actuator, named as the

electro-hydraulic actuator (EHA). Due to its numerous advantages, including energy savings, less noise, and

compactness, the EHA is considered as a potential actuator for these types of systems. A 6D joystick is installed to the

robot tip as a user interacting device. By analyzing commands from a human, an inverse kinematic solution distributes

reference angles for all active joints. Then an intelligent control method is applied to perform low level closed loop

control for all joints. Finally, experiments were carried out to verify the applicability of the developed PAR.

KEY WORDS

Power assistant robot, 7-DOF redundant robot, 6D joystick, electro hydraulic actuator (EHA), and inverse kinematic

NOMENCLATURES

*x : Command vector for the robot tip

x : State vector for tip of robot hand *

i : Command angular of the thi joint

i : Angular of the thi joint

iE : Kinetic energy in the thi joint

u

i : Upper boundary of the thi joint

l

i : Lower boundary of the thi joint

iu : Control signal for low level control of the

thi joint

ie : tracking error of the thi joint

INTRODUCTION

Robots play crucial roles in modern industry, as

evidenced by their use in automatic manufacturing

systems, assembly tasks, welding, etc. While robot

intelligence has contributed to significant advances in

manufacturing, it is difficult to replace the flexibility

and high adaptability of humans, especially in difficult

operations. For this reason, there is a need to develop a

system that combines the power of robots and the

flexibility of humans in such a system which is

so-called a power assistant robot (PAR) [1]. The main

goal of the system is to strengthen human power and

provide support to accomplish tasks associated with

high loads, such as heavy weight lifting, grinding, etc.

This paper presents the development of a 7-DOF PAR

system for use in heavy industrial tasks, including load

lifting, grinding, moving, etc. Figure 1 shows the

Copyright © 2014 JFPS. ISBN 4-931070-10-8

Proceedings of the 9th JFPS International Symposiumon Fluid Power, Matsue, 2014

Oct. 28 - 31, 2014

732

3C1-2

structure of the proposed PAR system. In this paper, the

PAR system is a hydraulic powered anthropomorphic

exoskeleton system which is based on a new type of

pump controlled electro-hydraulics actuator. This is a

flexible design with a large working space that allows

user to operate easily and comfortably with loading

capacity up to 40kg. In order to perform HMI task, a

6-D joystick is chosen and installed at the robot hand tip

to obtain commands for 6 active degrees of freedom.

Since 6 command signals are obtained, an inverse

kinematic resolution with joint limitation avoiding

strategy will provides feasible reference control signals

to all 7 active joints. Then, low level tracking control for

all elemental joints are employed to make the robot

operate accurately under human control.

Figure 1. Working principle of PAR system

HARDWARE CONFIGURATION

Electrohydraulic actuator

MM

DC Motor

Figure 2. Schematic diagram of EHA

Pump controlled EHA is considered as a novel hybrid

actuator of various uses due to its advantages over

conventional hydraulic actuating system [2]. The

structure of this EHA is shown in figure 2. Here, the

state of hydraulic actuator is adjusted directly by

operation of the bi – directional pump. Obviously, the

pump–controlled EHA acts as a Power – Shift system,

which shifts the high speed from electric motor into the

high force at hydraulic cylinder. Thus, comparing with

the conventional hydraulic system, the pump–controlled

EHA creates a sleeker, cleaner way to produce hydraulic

power with higher energy efficiency. Because of its

advantages, the pump–controlled EHA has been

developed as commercial products, e.g., the mini

motion package from Kayaba Industry Co., the

intelligent hydraulic servo drive pack from Yuken

Kogyo Co.

Hydraulic CylinderOil Tank DC Motor

Bidirection Pump

Pilot-operated

Check Valves Fig 3. Components of MMP

In this paper, the EHA is developed using a DC motor, a

gear pump, a reservoir, and supplement valves system

of the Mini motion package [3] from Kayaba Industry to

drive a rotary actuator manufactured by Helac Corp. as

shown in figure 3. In order to provide safety for user, a

counter balance valve is added to each actuator in cases

of power loss (figure 4). Specifications of the EHA is

shown in Table 1.

Fig 4. Structure of hydraulic rotary actuator with

counter balance valve

Table1. MMP specification

1 Rated output 250W

2 Rated voltage 24VDC

3 Rated flow 0.9L/min

4 Rated pressure 6.4Mpa

5 Setting pressure of relief valve 7.1Mpa

6 Maximum pressure 13.7Mpa

7 Actuating torque 620 Nm (210bar)

Copyright © 2014 JFPS. ISBN 4-931070-10-8 733

Design of a 7 DOF robot hand

The considered 7-DOF robot manipulator in this paper

is an S-R-S (Spherical-Rotational-Spherical) robot

model, which is also known as anthropomorphic

manipulator. This structure is often used in humanoid

arm due to capability of generating human like arm

motion. The structure of 7-DOF manipulator is shown

in figure 5. Here, seven revolute joints are serially

arranged, and the two adjacent joint axes are place

perpendicularly.

Figure 5. Structure of the 7-DOF robot hand

The experimental apparatus of robot hand is described

in figure 6. In this configuration, seven EHAs are used

for seven joints of the robot hand. Each EHAs is driven

by an MD03 DC motor driver manufactured by

Devantech corp and angular state of each joint is

captured by a low noise potentiometer. Table 3 shows

the configuration of the developed robot hand while

Table 4 shows the engineering specification of the robot

hand.

Table 3: Denavit-Hatenberg table

Joint a d Alpha Theta

1 0 0.2 90 1

2 0 0.15 90 2

3 0 0.3 -90 3

4 0 0.3 90 4

5 0 0.1 -90 5

6 0 0.1 90 6

7 0 0.1 0 7

Figure 6. Experimental robot hand

Table2: Joints parameters

Joint Specification Working range

1 Torque 125 Nm

Working pressure 14Mpa -90o

90 o

2 Torque 620 Nm

Working pressure 21Mpa -90 o 90 o

3 Torque 410 Nm

Working pressure 14 Mpa -90 o 90 o

4 Torque 240 Nm

Working pressure 10Mpa -45 o 45 o

5 Torque 110 Nm

Working pressure 10Mpa -140 o 140 o

6 Torque 110 Nm

Working pressure 10Mpa -140 o 140 o

7 Torque 51Nm

Working pressure 10Mpa -45 o 45 o

CONTROL SYSTEM

Control system configuration

The control system is built on a personal computer in C#

environment with A/D and D/A PCI cards from NI (NI

PCI-6014, NI PCI-6713). The general control diagram

of redundant manipulator is shown in figure 7. As

shown in this figure, the control system includes of a

redundant inverse kinematic resolution for the 7-DOF

robot hand and seven low level control systems for all

elemental joints to perform tracking control. The aim of

the control system is to make the robot end effect track

as close as possible to the generated. As presented in

above section, the robot contains 7 rotational joints

which are driven by electric motors and their controller.

Besides, to generate the reference signal for elemental

actuator, inverse kinematic is applied, based on the

desired posture of robot end effect and the current states

of elemental joints, the constraints between the

rotational joints speed are deduced. It is known that the

robot end effect posture is described by 6 parameters in

3-D workspace, thus the 6 equivalent equations with 7

variables can be established. The feasible solution is

selected which minimize the consumption energy of the

robot hand [4]. Since the commands for all joints are

obtained, seven low level controllers will provide

tracking control for elemental joints of the robot hand.

Redundant

robot model

Joint 1

Joint 2

Joint 7

Desired

trajectory

Tip

Position

Tip

Orientation

Joints states

Low level

tracking

control for

all joints

Inverse

kinematic

resolution

Computer

base

Figure 7. Control diagram of 7-DOF manipulator

End point

1

2

3

4 5

Joint 1

6

Joint 2

Joint 4 Joint 5 Joint 6 Joint 7Joint 3

7

Fig. 1. 7-DOF Robot simplification structure

Copyright © 2014 JFPS. ISBN 4-931070-10-8 734

Processing input command from joystick

The joystick used in this application is a 6D joystick

which is based on the JR3 Multi force torque sensors

(figure 8). Six commands signals obtained from the

joystick are included with 3 force signals and 3 torque

signals. However, the torque and force signals of each

axis usually possess the cross-talk effect. Hence, an

effective processing method must be developed to solve

this crosstalk effect problem. In this application, a fuzzy

based compensator is developed to deal with this

crosstalk behavior. Consequently, 6 commands signals

for the PAR are obtained independently for most cases.

Connect to robot

6 axis load cell

Handle

Gripper

control button

Figure 8: Human robot interaction using load cell

Inverse kinematic resolution The robot command acquired by the joystick can be

illustrated as: , , , , ,J J J J J J

x y zC C C C C C (see figure 9);

the current robot state: 1 2 3 4 5 6 7

0 0 0 0 0 0 0, , , , , , ; and the

tip (end effector) state can be expressed in Euler angle

0 0 0 0 0 0, , , , ,x y z .

O

x

y

z

Cx

Cy

Cz

C

C

C

C

Figure 9. Joystick command in tip coordinates

The transformation matrix from the end effector to the

global coordinate which is calculated as follows:

0 0 1 2 3 4 5 6

7 1 2 3 4 5 6 70 1

Tip Tip

Origin OriginR t P tT t T T T T T T T

(1)

The final goal is to determine the displacement value of

joints 1 2 3 4 5 6 7, , , , , , which satisfies the

motion command

Let consider the joint shoulder group and wrist group as

the spherical joint. The PAR structure can be simplified

as in figure 10.

The joystick command C can be regarded as the

displacement of the tip in a small variation of time

(Including position and orientation). This displacement

can be expressed as the following matrix (The

orientation is presented in Roll-Pitch-Yaw angles):

Shoulder

Elbow

Wrist

Force

l1l2

l3

A

B

C

D

Fig. 10: Robot simplifying structure

0 0 0 1

x

y

z

J

C C C C C C C C C C C C

J

C C C C C C C C C C C C

J

C C C C C

c c c s s s c c s c s s C

s c s s s c c s s c c s Cd

s c s c c C

(2)

The new desired states of robot tip can be calculated as

follows:

1 11

0 1

0 1

Tip Tip

Origin Origin

Tip Tip

Origin Origin

R t P tT t

R t P td

(3)

where RWrist

Origin is the rotation matrix and PWrist

Origin is vector

position.

As mentioned in previous section, due to the specific

structure that the three joints intersect at a single point,

the three first joints and three last joints can be regarded

as spherical joints. As the posture of the robot tip is

defined, the position of the wrist joint (point C) is

calculated by following formula:

0

7 31 1 0 0 1T T

x y zC C C T t l (4)

where [0 0 –l3 1]’ is the position of the wrist in

coordinates (O7X7Y7)

It can be realized that the motion of 7-DOF manipulator

for the fixed tip posture does not change the wrist

position, while the writ orientation observed from the

global coordinates varies depending on the value of

shoulder and elbow joint.

Let consider the triangle ABC which has two known

length sides AB and BC. Hence, the value of elbow

angle can be uniquely calculated based on geometrical

relation:

2 2 2 2 2

1 2

4

1 2

cos2

x y zC C C l l

l l

(5)

where l1, l2 are the length of AB and BC, respectively.

Besides, based on the Denavit-Hatenberg rules, the

position of the wrist can be calculated in the global

coordinates as follows:

0 0 1 2 3

1 2 3 40 1

Wrist Wrist

Origin Origin

Wrist

R PT T T T T

(6)

where RWrist

Origin is the rotation matrix and PWrist

Origin is vector

position of the wrist in global coordinates (O0X0Y0).

Let consider the wrist position of the writs, it can be

expressed as the following formula:

Copyright © 2014 JFPS. ISBN 4-931070-10-8 735

1 1 2 3 4 2 4 1 3 1 2 3 1 4 2 1 1 2

2 1 2 3 4 1 1 2 2 4 1 3 2 3 1 4 1 2

3 1 2 3 4 1 2 2 2 4 3 2 4

, , ,

, , ,

, , ,

x

y

z

C f l s s s c c c c c s l c s

C f l s s l s c s c c s c s s

C f l c l c c c s s

(7)

Let consider the small displacement of C in one sample

time

1

1

1

x x x

y y y

z z z

C C t C t

C C t C t

C C t C t

(8)

We have the relation between the outputs and the

variables of equation (8):

1 2 3 4C , C , C ' , , , 'x y z J (9)

The differential translation motion of the joystick

coordinates can be written as: T

x y zC C C

s (10)

Let consider time factor, we have

1 2 2 40

limT T

x y zt

dC C C

t dt

s s=J& & & & & & & (11)

where J is Jacobian matrix and calculated as follows:

1 1 1 1

1 2 3 4

2 2 2 2

1 2 3 4

3 3 3 3

1 2 3 4

f f f f

J f f f f

f f f f

(12)

The elemental of Jacobian matrix can be calculated as

follows:

11 2 4 1 3 1 2 3 1 4 2 1 1 2

12 2 4 1 2 3 1 4 2 1 1 2

13 2 4 1 3 1 2 3

14 2 4 1 3 1 2 3 4 1 2

21 1 1 2 2 4 1 3 1 2 3 1 4 2

22 1 1 2 2 4 2 3 1 4 1 2

23 2 4

( ( - ) - ) -

(- )

( - )

( ( ) - )

- ( (- - ) - )

- ( - )

- (

J l s c s s c c s c s l s s

J l s c s c c c c l c c

J l s s c c c s

J l c s s c c c s c s

J l c s l s s s c c c c c s

J l s c l s s c s c s c

J l s

1 3 2 3 1

24 2 4 1 3 2 3 1 4 1 2

31 33 2 3 2 4

32 1 2 2 2 4 3 2 4 34 2 2 4 3 2 4

)

- ( ( - ) )

0 -

- (- - ) - (- - )

c c c s s

J l c c s c c s s s s

J J l s s s

J l s l s c c c s J l c s c s c

(13)

The value of the 4th angle joint is determined in

aforementioned step, thus the speed rate can be easily

obtained in one sample time t :

4 44

1t t

t

& (14)

Expanding equation (9), the linear set equations can be

established (15). As mentioned in previous, the wrist

position can be obtained independently with the wrist

orientation. In addition, the determinant of matrix in

(12) is equal to zero (The matrix has the rank less than

3). Therefore, the three equations in (15) are not

independent and we cannot determine the single

solution of the equation.

1 1 1 2 1 3 1 4

1 2 3 4

2 1 2 2 2 3 2 4

1 2 3 4

3 1 3 2 3 3 4

1 2 3 4

C

C

C

f f f x f

f f f y f

f f f z f

& & & &

& & & &

& & & &4

(15)

However, the relation between the joint in shoulder

group can be deduced. Due to J31=0, we choose the first

and the last equation in (15) to obtain the following

constraints:

1 3

2 3

A B

C D

& &

& & (16)

where the parameters A, B, C and D is calculated as:

12 4 4

4

1 4

4 32 11

4 4

411 33 13 32 33

11 32 32

1

; ;

J Cz f

A Cx fJ J

Cz fJ J J J J

B C DJ J J

&

&

&

(17)

From equation (16), the relation between the shoulder

angle joints are determined. We have two equations for

three variables, thus to compute the unique solution, one

more equation called cost function must be proposed.

7

1

*

0

i

i

l u

minimize E

subject to

x f

(18)

Based on this limitation, the boundary of3 can be

determined as follows:

min

3 3 3min max & & & (19)

Then cost function is covariate with 3&so the optimized

solution can be chosen base on equations (18) and (19).

Substituting value of 3& in (16), we get the solution of

1&and 2

&. Then the new angle values of the shoulder are

updated as follows:

1 1 1

2 2 2

3 3 3

1

1

1

t t t

t t t

t t t

&

&

&

(20)

After getting the new angle values of shoulder and

elbow, we can compute the rotational matrix which

locates in the wrist joint 0 1 2 3

1 2 3 4

Wrist

OriginR R R R R (21)

Besides, we have Tip Wrist Tip

Origin Origin WristR R R (22)

Copyright © 2014 JFPS. ISBN 4-931070-10-8 736

So we can computer the rotational matrix of the wrist

as:

1

Tip Wrist Tip

Wrist Origin OriginR R R

(23)

Based on the Denavit-Hatenberg table (table 1), the

formula of RTip

Wrist can be written as

5 6 7 5 7 5 6 7 5 7 5 6

6 7 5 5 7 5 7 6 5 7 5 6

7 6 6 7 6

c c c -s s c c s -s c c s

R c c s +c s c c +c s s s s

-c s -s s c

Tip

Wrist

(24)

Balancing the element of (23) and (24) we can deduce

5 6 7, ,

6

7

5

cos R 3,3

R 3,2tan

R 3,1

R 2,3tan

R 1,3

Tip

Wrist

Tip

Wrist

Tip

Wrist

Tip

Wrist

Tip

Wrist

a

a

a

(25)

Tracking control In this paper, fuzzy PID controllers are employed for

constructing low level control of the EHA system for all

joints of the PAR.

ReferenceTunable Fuzzy

PID controllerEHA system

de

+-

e

ucontrol

Figure 11. Control diagram of the each active joint

It is well-known that the design of fuzzy PID controller

bases on engineering experience. In this paper the

development of the fuzzy PID controller is inherited

from our previous work on EHA system [5].

EXPERIMENTAL EVALUATION

Performance of the low level controller

In order to evaluate the performance of the low level

controller, a set of sinusoidal signals and a set of manual

reference commands have been applied to perform

tracking control for all joints. Two examples of the

tracking responses have been shown in figure 7 and

figure 8.

As shown in these figure, the performance of the fuzzy

PID controller shows very good response which

promises an accurate tracking performance for the PAR.

Traveling of the tip pose

In order to evaluate the performance of the inverse

kinematic resolution, many evaluations have been

carried out.

Case 1: Suppose that the desired trajectory of the tip is

described by the following equation:

5 10 15 20 25 30 35 40

10

20

30

40

50

60

70

80

90

100

Jo

int

3 (

o)

Time (s)

Reference

PID

Fuzzy PID

Tuning fuzzy PID

Figure 12. Sinusoidal tracking response of the 3rd joint

0 5 10 15 20 25 30 35 40 45

0

10

20

30

40

50

60

70

80

90

An

gle

(J

oin

t 3

)

Time (s)

Reference

Response

Figure 13. Manual set tracking response of the 3rd joint

10 sin 2

8 665.7

10 cos 2 141.4

x t t t

y t t

z t t t

(26)

with the constant orientation matrix

0 1 0

R 0 0 1

1 0 0

Figure 14 shows the traveling performance of the

proposed PAR system.

-100

-80

-60

-40

-200

20

40

60

80

100

-220

-200

-180

-160

-140

-120

-100

-80

-60

580

6 0 0

6 2 0

6 4 0

6 6 0

6 8 0

-100

-80

-60

-40

-200

20

40

60

80

100580

6 0 0

6 2 0

6 4 0

6 6 0

6 8 0

Desired trajectory

Z (

mm

)

Y(mm)

Robot performance

X(mm)

Figure 14. Trajectory for the spherical tracking task

Copyright © 2014 JFPS. ISBN 4-931070-10-8 737

Case 2: This case is to illustrate the flexible

characteristic of robot in tracking an arbitrary trajectory.

The initial states of the end effector can be described as

follows:

Position: [0 550 -533]; Orientation:

0 1 0

R 1 0 0

0 0 1

With initial joint angles as:

1 2 3 4 590, 45, 0, 90, 0, 6 745, 0

And the destination states:

Position [-300 519 -453];

Orientation: 0.5 0 0.866

0.866 0 0.5

0 1 0

R

The trajectory is the interpolation between the discrete

points which locate arbitrarily on the robot path.

-600

-400

-200

0

200

400

-530

-520

-510

-500

-490

-480

-470

-460

515

520

5 2 5

5 3 0

5 3 5

5 4 0

5 4 5

5 5 0

5 5 5

-600

-400

-200

0

200

400

-530

-520

-510

-500

-490

-480

-470

-460

515

520

5 2 5

5 3 0

5 3 5

5 4 0

5 4 5

5 5 0

5 5 5

Desired trajectory

Z(m

m)

Y(mm)

Robot performance

X(mm)

Figure 15. Trajectory of the robot end effect

The results shown in figure 14 and figure 15 conferment

the effectiveness of the proposed controller in control

the robot trajectory. It also suggests that the redundant

inverse kinematic is effective in online computational

process.

In our experimental verifications several loading

conditions have been considered. In the present phase

the power assistant robot has been tested in general

operating conditions with load up to 30kg. For heavier

loading conditions, the second link and the third link

cannot provide enough power to perform power assisted

task.

CONCLUSION

In this paper, a power assistant robot was developed by

applying a novel electro-hydraulic system. An

interaction between the user and the PAR is achieved

through a 6-D joystick. An inverse kinematic resolution

based on joint limit avoidance and energy optimization

is employed to generate reference signals for all joints.

Low level controllers are designed bases on fuzzy PID

scheme for tracking control of all EHA systems.

Consequently, experimental evaluations have been

carried out to investigate the proposed PAR system.

Based on the experimental results, it can be concluded

that the developed PAR promises an effective

applications for use in several industrial fields, where

assistant system with high reliability are required.

REFERENCES

1. Hiroaki K. and Yoshiyuki S., Power assist system

HAL-3 for Gait disorder person, Lecture notes in

Computer Science, 2002, 2398, pp. 19-29.

2. Grabbel J. and Ivantysynova M., An investigation

of swash plate control concepts for displacement

controlled actuators, Int. J. Fluid Power, 2005, 6-2,

pp. 19–36.

3. Ltd. Kayaba Industry Co., Ltd., Mini-Motion

Package Catalog, available:

http://www.kybfluidpower.com/Mini_Motion_Pack

age.html

4. Patel R.V. and Shadpey F., Control of Redundant

Robot Manipulators - Theory and Experiments,

Springer Berlin Heidelberg Germany, 2005.

5. Ahn K.K., Doan N.C.N., Maolin J., Adaptive

Backstepping Control of an Electrohydraulic

Actuator, IEEE Trans. Mechatronics 2014, 19-3,

987 – 995.

Copyright © 2014 JFPS. ISBN 4-931070-10-8 738