See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271507708

A control technique for a 6-DOF master-slave robot manipulator system for

miniature tasks

Article · January 2010

CITATIONS READS

2 248

3 authors:

Kritsanun Malithong Viboon Sangveraphunsiri

Kasetsart University, Thailand, Nakhon Pathom, Kamphaeng Saen Campus Chulalongkorn University
9 PUBLICATIONS 34 CITATIONS 61 PUBLICATIONS 327 CITATIONS

SEE PROFILE SEE PROFILE

Supachai Vongbunyong
King Mongkut's University of Technology Thonburi
48 PUBLICATIONS 532 CITATIONS

SEE PROFILE

All content following this page was uploaded by Supachai Vongbunyong on 31 January 2015.

The user has requested enhancement of the downloaded file.

AIJSTPME (2010) 2(4): 47-55

A Control Technique for a 6-DOF Master-Slave Robot Manipulator System for

Miniature Tasks

Malithong K.
Department of Mechanical Engineering, Faculty of Engineering, Chulalongkorn University, Bangkok,
Thailand

Sangveraphunsiri V.
Department of Mechanical Engineering, Faculty of Engineering, Chulalongkorn University, Bangkok,
Thailand

Vongbunyong S.
Department of Mechanical Engineering, Faculty of Engineering, Chulalongkorn University, Bangkok,
Thailand

Abstract
This research covers the application of the virtual fixtures that is implemented on the 6-DOF master-slave
robot manipulator system for miniature tasks. A benefit of the control technique using virtual fixtures is to
increase performance of operating master-slave manipulators. In this paper, we discuss the implementation of
two categories of virtual fixtures: guidance virtual fixtures, which assist the operator in moving the
manipulator along desired paths or surfaces in the workspace, and virtual wall, which prevent the
manipulator from entering into forbidden regions of the workspace. The virtual fixtures are constructed from
the mathematics model in the control unit of the developed robotic system. The master arm, Phantom Omni®
Haptic Device, generates the forces reflect from the slave arm that interacts with the virtual fixtures. These
forces are felt by the operator. The direction and magnitude of these forces are associated with the desired
virtual wall. In general, virtual walls are modelled as a spring-mass system. This model is able to respond
effectively to sense of touch of an operator. In this research, the method of determining appropriated
parameters, such as spring constant and/or damp constant. These parameters based on the operators’ feeling-
of-touch and position error of the manipulator. Nevertheless, shape of the virtual walls can be created based
on a desired task. The strategy and procedure of constructing the walls are also presented in this article.

Keywords: Virtual wall, Master slave system, Miniature task manipulator

1 Introduction
Master-slave robot manipulator systems or such as difficult assembly tasks, inspection and
teleoperation have been well known as one of the manipulation tasks in dangerous environments, and
extension of manipulation capability to a remote medical applications. There are many robotic medical
location and a person’s sensing. They are used in applications, especially Minimally Invasive Surgery
many applications in order to extend the performance (MIS). For example, the daVinci® Surgical System
of each operation. The movements of the slave [1] is a notable MIS robotic system.
manipulator can be commanded more precisely and However, the teleoperation cannot be used
accurately by the robot’s AI and controller than effectively without the force reaction. Hence, haptic
human operator command of the desired movements. interfaces should be implemented. With the haptics,
In addition, these robots can operate in difficult tasks, the human operators can control the slave

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand 47

Malithong K. et.al.

manipulator more effectively. The feedback force,

generated from the haptic device (the master device),
represents the force that acts on the slave 0

manipulator. Therefore, the operators can feel the

force when the slave manipulators carry weights or
touch the real–world objects. 1
2a

The concept of force reaction is further extended to 5

the virtual fixture which can assist human operators 2b
4
to control the manipulator more accurately. Haptic 6

virtual fixtures are software-generated force and 2

position signals applied to operators via haptic

devices. Virtual fixtures help operators perform 3

master-slave manipulation tasks by limiting

movement into restricted regions and/or influencing
movement along desired paths. The haptic device Figure 2: RCRT-1 Manipulator
generates the force corresponding to the virtual
fixture math model. The movement of the operators
will be constrained to move only in the desired paths. 2 Mathematical model of the virtual fixture
The experiments and math modeling of virtual The concept of virtual fixtures is to specify the
fixtures are presented in several researches [2-4]. working area of the master device virtually. This will
With respect to the math model of the reacting force, help the operator to work in the specific area more
the virtual fixture can be categorized into two types: convenience. Human-machine systems with virtual
impedance and admittance [2]. fixtures can achieve safer and faster operation.
In this research, the virtual fixtures are implemented Figure 3 shows the circular working area. The virtual
on the 6-DOF master-slave robot manipulator system wall is defined by a function, f(x,y,z) = constant.
for miniature tasks. The Phantom Omni Haptic When the operator move the end-effector contact to
device (Figure 1) is used as the master device. The the virtual wall. The reaction force (F) will be
RCRT-1 Manipulator, a 6-DOF Slave Manipulator generated to against the operation motion.
for Miniature Tasks [5] (Figure 2) is used as the slave
manipulator with the six links labeled No.1 - 6. In
addition, impedance type virtual fixture is
implemented.

Figure 3: The circular working area

The shape and characteristic of virtual fixture is

Figure 1: Phantom Omni Haptic Device
determined from the mathematical model. With
respect to the impedance type, the math model is
equivalent to spring-mass-damper model. The
operator will feel the pushing force that interacts to

48 © King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

A Control Technique for a 6-DOF Master-Slave Robot Manipulator System for Miniature Tasks

their hands when the end-effector moves out of the virtual fixture is shown in Figure 5.  is thickness
preferred virtual path. Equivalent to the real of soft wall zone specified by math model K-
mechanical spring-mass-damper system that is constants.
attached on the friction-free and massless slider, the
force is generated only in the direction normal to the
slider [2]. Consequently, the magnitude of force is
the function of the deviated distance r (See Figure
5) and absolute movement speed. The deviated
distance, is determined from the distance between the
end-effector and the preferred path. On the other
hand, the end-effector can be moved freely along the
rail slider, representing the preferred path of the
virtual fixture. The math model in represented by
Figure 4, where, m is the mass of the Phantom’s
end-effector and Fuser is force applied by the operator
in order to move the end-effector.
Figure 5: Description of a virtual fixture

Rewrite Equation (1) as a linear combination of a

unit vector n , as shown in Equation (2):
F  K 1  dr  n  K 2 V  n (2)

where, r  n dr  , which dr is the corresponding

distance. r  n V  , V is the speed of the end-
effector. The unit vector n and dr can be determined
from Equation (4) – (6). The speed V  can be
Figure 4: Spring-mass-damper model obtained directly from the Phantom’s API and
libraries.
Assume a scalar function as virtual boundary (Figure
Basically, the magnitude and direction of the force 5), continuous function,
vector is directly related to the position, speed
  f x , y, z  (3)
(Phantom’s end effector), and spring-constant K 1 
and damper-constant K 2  , as shown in Equation The, compute F as follows:
(1), d 
dr  (4)
F  K 1 r  K 2 r  (1)   

where, r is the displacement between the position n (5)
of the end-effector and the fixture’s boundary 
perpendicular to the fixture’s surface. r is velocity
  
of the end-effector.   i j k (6)
x y z
2.1 Computation of the force vector Substitute Equations (4) – (5) into Equation (2), the
Direction of the force vector must be perpendicular force vector is:
to the fixture’s surface boundary in order to provide K 1d 
the most appropriate feeling to the operator. These F 2  K 2V
(7)
 
force vectors are determined from the gradient vector
field of the scalar function which represents the
where d  fboundary x i , yi , z i   boundary (8)
shape of the virtual fixture. The description of a

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand 49

Malithong K. et.al.

boundary  fboundary x i , yi , z i  i boundary (9) typically related to the distance dr , the force is
continuously changing along the non-preferred
Note that boundary = constant since  is a continuous direction. The force that determines the boundary
hardness is as shown in Figure 6.
function. This scalar function  , represents the shape
of virtual fixtures described in Section 3.
In general, the characterization of virtual fixtures is
based on the basic geometry or primitive volumes
(i.e. straight line, sphere, cylinder, cone, and
rectangular cube). This is a good approach for several
reasons. First of all, this design enforces well-defined
boundaries on the forbidden region, which makes the
usage and graphic display of the fixtures
comprehensible. Secondly, a volumetric
representation of virtual fixtures facilitates their
integration with the simulator or master-slave robotic
systems. Thirdly, a set of basic volumetric forbidden Figure 6: The interacting force in each region
region would make a flexible virtual fixture library
for operators to assemble forbidden region of more
geometrical complexity. The soft-wall thickness  can be computed from
the maximum force produced by the Phantom and
K 1 not dependent on the function of K 2 , as shown
2.2 Specify the K-constant
in Equation (10):
Physically, the K-constant represents how hard is the
constrain boundary. In case of the hard-fixture or   Fmax K 1 (10)
hard-wall (large K-constant), the interacting force
effectively regulate the movement within the
preferred path or boundary with no or little deviation 3 Shape of the virtual fixture
from this constraint is allowed during the operation. Virtual fixture can be categorized according to its
On the other hand, in case of the soft-fixture or soft- boundary. Two types of the fixture presented are 1)
wall (small K-constant), the operator can move more guidance virtual fixture and 2) virtual wall. A major
freely in the non-preferred directions. Theoretically, difference between these two fixtures is the region
the K-constant should be as large as possible in order that the end-effector can move without constrained
to completely control the end-effector to move within force. The details are described as follows.
the preferred path. However, the maximum force is
limited by the limitation of the master device (the
3.1 Guidance virtual fixtures (GVFs)
Phantom’s maximum force is 3.3 N). Nevertheless,
too large K-constant leads to large feedback force. It Guidance Virtual Fixtures are constructed as path
leads to overshoot when the end-effector had been lines. The operators can move the end-effector freely
pushed back into the preferred paths or boundary. In in the preferred-direction along the desired path lines.
practice, the appropriate value of K 1 and K 2 must The force vector is computed from Equation (7).
be adjusted in order to improve the operators’ Therefore, the pushing force pointing to the path is
interaction. The value of K 1 should be increased as generated in the forbidden regions around the desired
path. An example of the GVF is shown in Figure. 7.
large as possible until the response becomes
underdamp. While, the value of K 2 is adjusted to
eliminate the overshoot. In this research, the
preferred values are: K 1  1 and K 2  0.01 .

2.3 Forbidden region

The force is computed from Equation (2) – (6) are in
the forbidden region. As the force magnitude is

50 © King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

A Control Technique for a 6-DOF Master-Slave Robot Manipulator System for Miniature Tasks

The major benefit of virtual walls is to limit the

device to move only in the desired area. The walls
assist the operators to control the movement easily
and effectively in order to avoid prohibited area.
Moreover, the virtual environment can be created
from virtual walls. For example [2], many walls are
constructed and arranged in proper location. The
operators can feel the walls and operate the tasks
through the haptic device.
Figure 7: Guidance Virtual Fixture
4 Construct virtual fixtures using the Phantom
There are many applications of GVFs, for example in
cardiac surgery [6], the virtual fixture is constructed Shape of the virtual fixtures is defined from their
from the image data that provides the exact location boundary. For both GVFs and virtual walls, the
of the artery and other organs. boundary can be specified from the scalar function
boundary . In this research, the Phantom was used to
3.2 Virtual wall define the desired location of the virtual fixtures in
3D coordinate.
The preferred-region is bounded by the virtual
fixtures. These fixtures act like the walls that prevent
4.1 Basic geometry virtual fixtures
the operators moving the end-effector outside, while
can moves freely inside this region. An example of The location of virtual fixtures can be defined from
virtual wall is shown in Figure 8. one point if other features of the geometry (e.g.
radius, height, length, orientation etc) are specified or
pre-defined. This point can be set at the center point
of a virtual sphere, or a point on the center line of a
cylinder, or the center of a cube. For example,
motions of the end-effector are bounded inside these
geometries. The scalar functions of cylindrical and
sphere are Equations (12) and (13), respectively. As
examples, movement path of the end-effectors of the
Phantom and the RCRT-1 Manipulator in Cartesian
space are shown in Figures 9 and 10.
Figure 8: Virtual wall
cylinder  x 2  y 2  R 2 (12)
The desired boundary can be specified as the scalar spherical  x 2  y 2  z 2  R 2 (13)
function in Equation (3). However, the additional
boundary conditions must be provided to determine
which region is preferred or non-preferred. For 4.2 Shape-definable virtual fixtures
example, in Equation (11), the force vector is
computed if the end-effector is in the position Since the shape of objects in real world application is
various and complex, e.g. human’s organs, blood
x i , yi , z i  that corresponds to the condition vessels, and etc, virtual fixtures constructed based on
 x i , yi , z i   boundary . Apparently, this means the basic geometries are not capable to represent those
end-effector is in the forbidden-region and there is no objects effectively. Therefore, “shape-definable
force generated in the preferred-region. virtual fixture”, which is a virtual fixture that its
shape can be manually defined by the operators, is
 K 1d   proposed in this research.
  2 K 2V  ;  x i , yi , z i   boundary
F (11)

 0 ;  x i , yi , z i   boundary

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand 51

Malithong K. et.al.
p p ( )
As regard to the connecting straight lines. Each line
is reconstructed from a pair of defined points. The
linear equation and interacting force equation are
-120
used only between these two points. The connecting
-140 lines and their corresponding equations are
-160
constructed repeatedly for the whole virtual
Z

boundary’s range of operation. The description of the

-180 shape-definable virtual fixture is shown in Figure 11,
-200 where, Pt is the position of the end-effector at time t,
40 380
20 360 Pi is the desired point and Ft is the interacting force
0 340
-20 320
that act on the end-effector when it is moving in the
Y -40 300 X corresponding range x i 1, x i  .
Figure 9: A cylinder virtual fixture
p p ( )

-140

-150

-160
Z

-170

-180

-190 400
20
380
0
360
-20
Y -40 340 X

Figure 11: The shape-definable virtual fixture

Figure 10: A sphere virtual fixture

As shown previously this force is computed from

Here, the fixtures are considered only in 2D (XY- Equation (7) using scalar continuous function,  , as
plane) in order to reduce degree of complexity.
Hence, planes are defined as straight lines. the equation of the connecting line. Hence, this
Physically, the whole virtual wall boundary or function is derived from x i , yi , x i 1 and yi 1 :
preferred path is constructed from the set of 2D-
connecting lines. These connecting lines are defined line x t , yt   yt  mi x t  yi 1  mi x i 1  ;
(14)
from the set of defined points (Blue points in Figure x i 1  x t  x i
11) which located by the operators.
Typically, the operation can be divided into 2 key
stages. In the first stage, the operator needs to specify where
the number of defined point and define location of mi  yi 1  yi  x i 1  x i  ; i  0,1, 2,..., n
each one. The operator needs to move the end-
effector to the desired location of the defined point. Similarly, force vector in this range can be defined
Then, the desired points will be collected when the by Equation (15), below:
operator send a command to the program. Coordinate
Ft  F x t ,yt | x i , yi , x i 1, yi 1, K 1, K 2  (15)
(x, y) of the location corresponding to each defined
point will be stored in the program memory. This From the process flowchart in Figure 12, the virtual
process needs to be done repeatedly until every fixture is constructed after the entire defined points
defined point is defined. Afterwards, in the second are defined by the operators. As the end-effector is
stage, the virtual boundary (or preferred path) moved within the workspace, the appropriate
corresponding to the defined points will be function corresponding to each position is assigned
automatically constructed by the program. Force and the interacting force vector Ft is generated.
interaction will be generated along the boundaries.

52 © King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

A Control Technique for a 6-DOF Master-Slave Robot Manipulator System for Miniature Tasks

The corresponding range (xi-1, xi] of end-effector 5.2 The RCRT-1 manipulator
position Pt is determined by searching algorithm. The
This 6-DOF slave manipulator with the PID-gravity
searching procedure start with assigning i=1. Then
compensation is normally used as a motion controller
repeatedly increase i by 1 if (xt≥ xi -1) and (i <
(at 1,000 Hz). When used in independent operation
number of defined point – 1) are satisfied.
(without a master device), it has 0.514 mm precision
Eventually, this loop will be terminated if the
with 0.005 mm repeatability. For master-slave
corresponding range is found (xi-1 < xt ≤ xi) or xt is
operation, its overall capability is marginally reduced
out of the range (xt ≥ xi). Afterwards, this force
due to the unpredictable movements of human’s
command Ft (magnitude and direction) will be sent
control. Overall, the RMS (Root Mean Square)
to the Phantom through HDAPI (function hdset(…))
maximum position error is 0.791 mm, average
position error is 0.293 mm and maximum orientation
START error is 0.832 degree (fine motion mode) [8].

Define the defined points

(P0,P1,P2,...,Pi,…,Pn)
6 Experiment and result
In the case of the master-slave system for miniature
Locate the position of tasks, the operator can monitor the position and
the Phantom’s end-effector (Pt) movement of the slave’s end-effector through the
digital microscope [5]. The master-slave system test
bed is configured as shown in Figure 13.
Search for the corresponding range
_ xi )
( xi-1 < xt <

Compute the boundary of the virtual fixture

from the defined points Pi-1 and Pi

Compute the reaction force Ft

from the defined points K1,K2,Pi-1,Pi and Pt

Virtual Walls GVFs

Define shape

Assign Ft = 0
for the movement
in the preferred region

Send force (Ft) command to the

Figure 13: The master-slave system test bed
Phantom through HDAPI Library

STOP 6.1 Experiment method

Figure 12: Process flowchart of constructing the The performance of a shape-definable virtual fixture
shape-definable shape virtual fixtures was emphasized in this experiment; both GVFs and
virtual walls were tested. On the test bed, a digital
microscope was equipped on the slave-side in order
5 Performance of the master-slave system to monitor the movement of the RCRT-1’s end-
5.1 The phantom omni haptic device effector. The test operation was conducted according
to 2 stages described in section 4.2. First, the
This haptic device has 6-DOF having nominal operator defined location of every defined points
resolution of 0.055 mm, producing 3-direction force corresponding to the preferred boundary or desired
feedbacks. The maximum force is 3.3 N and the path. Afterwards, the operator controlled the
continuous force is 0.88 N [7].

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand 53

Malithong K. et.al.

movement of the end-effector via the master-slave -60

Phantom
system. In general, the movement should be occurred -70
RCRT-1
only in the preferred-region or the desired path. (see -80 Virtual fixture
-90
Figure 14). This desired boundary path were used for

Y
-100
both GVFs and virtual walls testing. The parameters
K 1 = 1 and K 2 = 0.01 were specified. Hence, soft- -110

-120
wall thickness is  = 3.3 mm. Afterwards, the -130
operator controlled the RCRT-1 to move freely -80 -60 -40 -20 0
X
20 40 60 80

within the preferred path or region.

In practice, the operator may intentionally avoid Figure 15: Movement in the virtual wall test
penetrating the end-effector into the forbidden -60
region. This factor leads to the bias in measurement -70 Phantom
of the performance of the GVFs and the virtual wall. -80
RCRT-1
Virtual Fixture
In order to avoid this bias, the operator’s visual -90
sensing was eliminated by shutting down the camera

Y
-100
which is used to monitor movement of the end-
-110
effector. Therefore, only one perception of the
-120
operator was force interaction corresponding to the
-130
virtual boundary and the preferred path. This force -80 -60 -40 -20 0 20 40 60 80
X
regulated the hand movement to move only in the
proper area. Figure 16: A Movement in the GVF test

6.2 Results
7 Conclusions
Data of the movement in both tests were captured as
illustrated in Figures 14 and 15. Some undesirable The impedance type virtual fixtures were able to
movements that penetrated into the forbidden region regulate the movements of the operator effectively.
were observed in both tests. Each deviation is as an The end-effector moved within the preferred
error determinable from the location of “desired bounded region but approximately 1.6 mm deviation
path” and “RCRT-1’s end-effector”. These RMS occurred in both virtual walls and GVFs testing.
errors are 1.646 mm and 1.552 mm for virtual wall However, with regard to the 3.3 mm thickness (soft-
and GVF respectively. wall), the movements were still constrained within
this zone and did not penetrate the hard-wall zone.
Moreover, the performance of the RCRT-1 partly
affected these overall errors.
In conclusion, the virtual fixture is expected to be
applied to a number of applications. For example,
regarding MIS, the surgeon can manually define
forbidden regions containing parts of important tissue
that is severely sensitive. Consequently, the surgeon
can sense this boundary via the interacting force and
avoid collisions. Moreover, GVF is possibly used as
a guide path in order to increase precision in sutures.
The performance of the virtual fixture, such as
precision and operator’s sensing, can be further
improved by fine adjustment of the virtual fixture
gain. In addition, improvement of the motion control
of the RCRT-1 is necessary.

Figure 14: A virtual fixture in the experiment

54 © King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

 A Control Technique for a 6-DOF Master-Slave Robot Manipulator System for Miniature Tasks

References Network of Thailand, Thammasart University,

Bangkok, Thailand
[1] Intuitive Surgical, Inc., da Vinci Surgical
[6] Park S., Howe R.D. and Torchiana D.F., 2001.
System, 2005, available online: http://www.
Virtual fixtures for robotic cardiac surgery,
intuitivesurgical.com/
lecture note in computer science vol. 2208, the
[2] Abbott J.J., Hager G.D. and Okamura A.M.,
4th international conference on medical image
2003. Steady-hand teleoperation with virtual
computing and computer-assisted intervention,
fixtures, 12th IEEE workshop on robot and
1419 -1420.
human interactive communication, pp. 145-151.
[7] SensAble Technologies Inc., PHANTOM Omni
[3] Rosenberg L.B., 1993. Virtual fixtures:
Haptic Device, 2008, Available online:
Perceptual tools for telerobotic manipulation,
http://www.sensable.com/haptic-phantom-
Virtual reality annual international symposium
omni.htm
IEEE, 76-82.
[8] Vongbunyong S., 2008. A 6-DOF slave
[4] Abbott J.J., Marayong P. and Okamura A.M.,
manipulator for miniature tasks, M.Eng. thesis,
2007. Haptic virtual fixtures for robot-assisted
Department of Mechanical Engineering,
manipulation, The international symposium of
Chulalongkorn University
robotics research (ISRR), 28: 49-64.
[9] Ngamvilaikorn T., 2001. Design and
[5] Sangveraphunsiri V. and Vongbunyong S.,
development of a master arm for force
2008. A 6-DOF master slave system for
reflecting system, M.Eng. thesis, Department of
miniature tasks, paper presented in The 22nd
Mechanical Engineering, Chulalongkorn
Conference of Mechanical Engineering
University

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand 55

View publication stats