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

net/publication/250918031

Robotic Arm for Remote Surgery

Article · July 2013

Source: arXiv

CITATION READS

1 212

3 authors, including:

Steven Conrad Dinger Adam Pantanowitz

University of the Witwatersrand University of the Witwatersrand
3 PUBLICATIONS   19 CITATIONS    27 PUBLICATIONS   135 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Estimation of Body Mass Index from Photographs View project

All content following this page was uploaded by Steven Conrad Dinger on 19 June 2014.

The user has requested enhancement of the downloaded file.

 Robotic Arm for Remote Surgery
Steven Dinger∗, John Dickens and Adam Pantanowitz†
arXiv:1307.5641v1 [cs.RO] 22 Jul 2013

Abstract
Recent advances in telecommunications have enabled surgeons to op-
erate remotely on patients with the use of robotics. The investigation
and testing of remote surgery using a robotic arm is presented. The
robotic arm is designed to have four degrees of freedom that track the
surgeon’s x, y, z positions and the rotation angle of the forearm θ.
The system comprises two main subsystems viz. the detecting and
actuating systems. The detection system uses infrared light-emitting
diodes, a retroreflective bracelet and two infrared cameras which as a
whole determine the coordinates of the surgeon’s forearm. The actua-
tion system, or robotic arm, is based on a lead screw mechanism which
can obtain a maximum speed of 0.28 m.s−1 with a 1.5 ◦ .step−1 for the
end-effector. The infrared detection and encoder resolutions are below
0.6 mm.pixel−1 and 0.4 mm respectively, which ensures the robotic
arm can operate precisely. The surgeon is able to monitor the patient
with the use of a graphical user interface on the display computer. The
lead screw system is modelled and compared to experimentation re-
sults. The system is controlled using a simple proportional-integrator
(PI) control scheme which is implemented on a dSpace control unit.
The control design results in a rise time of less than 0.5 s, a steady-
state error of less than 1 mm and settling time of less than 1.4 s.
The system accumulates, over an extended period of time, an error of
approximately 4 mm due to inertial effects of the robotic arm. The
results show promising system performance characteristics for a rela-
tively inexpensive solution to a relatively advanced application.

Keywords: Remote surgery, PI controller, lead screw, optical encoder,

infrared light emitting diode.
∗
S. Dinger is with the Biomedical Engineering Research Group, School of Elec-
trical & Information Engineering, University of Witwatersrand, Johannesburg, e-mail:
[email protected].
†
J. Dickens and A. Pantanowitz are with the Biomedical Engineering Research Group,
School of Electrical & Information Engineering, University of Witwatersrand, Johannes-
burg.

1
1 Background
Remote surgery or telesurgery enables a surgeon to operate on a patient who
is at a different geographical location. The use of robotics in surgery has
improved the accuracy and capabilities of surgeon’s significantly in recent
years [1]. The major drawback, however, with current systems is that the
surgeon is usually required to control the robotic device in an uncomfortable
position with the device usually vulnerable to human tremors and operator
fatigue [2].
The concept of remote surgery is explored and tested with the use of a
robotic arm that is able to mimic the movement of a surgeon’s forearm. The
robotic arm is designed to have four degrees of freedom viz. the x, y, z
positions and rotation angle θ of the forearm. The simplest requirement of
the robotic arm is to track the movements of a surgeon’s forearm precisely
and to operate in an identical workspace to that of a typical surgeon.
The three types of human movement detection systems are the inside-in,
outside-in and inside-out detection systems [3]. The use of an accelerometer
to detect three-dimensional movement is a feasible technology and repre-
sents an inside-in type system [4]. The use of potentiometers, piezo-resistive
flex and cable extension represent examples of outside-in type systems [3],
however they suffer from drift.
A spherical design of the robotic arm overcomes many of the problems
faced in robotics, such as reducing the size and weight of the robotic instru-
ment [5]. The reachable workspace, however, for a spherical robotic arm is
limited to a sector of a sphere, and is therefore not desirable in terms of
reproducing the complete operating workspace of the surgeon.
The telesurgery robotic arm comprises two major non-collocated subsys-
tems, viz. the detection and actuation system. The coordination between the
two systems can be achieved through the use of current telecommunications
technology [1].
This paper specifies the system requirements, constraints and assump-
tions that are used to effectively implement such a system. The layout of the
system is presented and covers the two main subsystems, viz. the detection
system and the actuation system. The lead screw system is modelled and
used to compare to the experimentation results. The controller used for the
robotic arm is also discussed. The simulation and experimentation results are
provided and analysed in order to determine the performance of the system.
The system limitations are listed including recommendations and suggestions
for future work. A conclusive summary pertaining to the performance and
fundamental aspects of the system is given.

2
2 System Specifications
The forearm of a surgeon is considered to have four degrees of freedom, which
include the x, y, z positions and rotation of the forearm denoted as θ. A
satisfactory design, therefore, includes the ability of the robotic arm to mimic
and track the surgeon’s forearm for all four degrees of freedom.
The delay of the control signal of the system must be less than 500 ms for
the surgeon to successfully compensate for the latency during the operating
procedure [6]. Visual feedback is required in order for the surgeon to monitor
the patient and correct for any error that accumulates during the procedure.
The tracking error is considered more important as opposed to the steady-
state error, since the steady-state error can be easily corrected by the surgeon
through the visual feedback system. The system must be able to be calibrated
to remove any cumulative error that occurs during operation.
A constraint on the amount of wrist pitch and forearm flexion of the
surgeon exists mainly due to the outside-in type of detection system em-
ployed [3]. The outside-in detection system uses artificial sources placed on
the body with the sensors situated externally.

2.1 Assumptions
The assumptions used to design, implement and operate the system success-
fully are:

• Pitch of the surgeon’s hand is negligible during an operation.

• A maximum speed of 0.25 m.s−1 for each axis is deemed satisfactory.

• The workspace of the surgeon is devoid of infrared noise.

The assumption that the hand pitch is negligible is necessary, since the detec-
tion sources are placed on the hand as described in Section 3.1. A maximum
speed of 0.25 m.s−1 is reasonable, since a surgeon’s forearm should remain
in the same anatomical region throughout the operation, with the hand per-
forming most of the fast and precise movements. A speed of 0.25 m.s−1 ,
which was measured experimentally, is also relatively large for a typical hu-
man forearm to move. The workspace should be illuminated with fluorescent
lighting in order to reduce infrared noise.

3
3 System Architecture
The system comprises two main subsystems viz. the detection and actua-
tion systems. The block diagram that illustrates how the two systems are
coordinated and linked as shown in Figure 1.

reflection
IR LED Wii Stepper Stepper
array remote 1 drive IC θ
Surgeon’s forearm
movement IR LED 1
Wii
remote 2 H-bridge DC motors
IR LED 2 circuit (x, y, z)
Detection System
visual feedback

bluetooth Encoders and

Control dSpace push switches
PC control
unit
RS232 position feedback
Display Data
PC projector

Control & Disp. System Actuating System

Figure 1: System block diagram.

3.1 Detection System

The movement of the surgeon’s forearm is detected using a non-invasive
outside-in detection system, which consists of placing the sensors off, and
artificial sources on, the person. The artificial sources used are two infrared
(940 nm) light emitting diodes (LEDs) and a retroreflective bracelet.
The two infrared LEDs are located in the region of the proximal pha-
langes and are used to measure the z-axis position of the surgeon’s arm,
including the forearm rotation angle θ, as shown in Figure 2. Two Nintendo
Wii remotes are used to capture and process the data obtained from the
infrared sources and reflections. The Wii remote infrared (940 nm) camera
has a resolution of 1024 × 768, which at the heights and distances shown in
Figure 2, provides a resolution of 0.586 mm.pixel−1 for x and z axes, and
0.576 mm.pixel−1 for the y axis. A retroreflective wrist band is used as op-
posed to infrared LEDs in order for the Wii remote to continue tracking the

4
forearm under rotation. A 132-LED array is used to produce the infrared
light source for the reflective wrist band [7].

Wii remote 1
Infrared LED array

x-axis
Stand 0.6 m y-axis
0.34 m
1.27 m z-coordinate Infrared
Bracelet LED 2

Infrared θ 0.45 m
LED 1

z-axis
0.839 m Wii
remote 2

Figure 2: Illustration of the infrared detection system.

3.2 Actuation system

The actuation system is illustrated in Figure 3, and includes the optical en-
coders that measure the position of the carrier blocks and calibration switches
located on each axis. The robotic arm follows a simple lead-screw guide-rail
design, with the advantages of the design being:

• No energy consumption at rest.

• Stable and rigid.

• Lightweight and modular.

• Simple to construct and control.

The disadvantages of the design include a low mechanical efficiency due to

excessive amounts of dynamic friction [8]. The non-linear static friction in
the system causes a deadzone effect, whereby the control signal has no effect

5
 Push-button End-stop

12 V , 0.2 A
stepper motor
Vesconite carrier block
End-effector
Lead-screw
Push-button

Guide-rail
Encoder
12 V , 9 A Platform
DC motor

Supply

Motor bracket

Figure 3: Illustration of the mechanical robotic arm.

on the system within a certain armature voltage range. The fine standard
pitch used for the lead screws is 1.5 mm for the x and z axes, and 1.25 mm
for the y axis. The size of pitch requires large angular speeds (17000 rpm) to
generate modest linear speeds. The large angular speeds cause mechanical
vibrations, an increase in friction (heat losses) and significant audible noise.

4 Control
4.1 System Interfacing
The detected infrared information is processed and conditioned as inputs for
the dSpace 1104 control unit. The desired coordinates are written to dSpace
using the commands and functions supplied by the Matlab library called
mlib.
The dSpace unit interfaces with the actuation system and communicates
with the control computer. The model of the controller is built in the Mat-
lab Simulink environment, including the necessary routing such as the in-
puts from the encoders and outputs to the logic and pulse-width modula-

6
tion (PWM) channels.
The 1.5 ◦ .step−1 geared stepper motor has no feedback and is rotated
using open-loop, single-phase excitation [9][10]. A step frequency of 125 Hz
is used, which results in an angular speed of 31.25 rpm.

4.2 System Modeling

The model of the system is presented and is used to obtain the gain values
for the chosen proportional-integral (PI) controller. The assumptions used
to model the system are listed below:

• Inelastic steel lead screw.

• Motor inductance is negligible.

• Static friction coefficient for lubricated metals.

• The lead screw moment of inertia is larger than that of the motor shaft.

• Mechanical power loss in the motor is negligible.

The assumption that elastic torsion of the threaded steel rod does not occur
is validated since the rods are short in the lengths (less than 1 m). The
mass accelerated by the motors is much larger than the inertia introduced
by the inductance of the motor. The mass of the lead screw is much larger
than that of the motor shaft, as shown in equation 6, which validates the
assumption that the lead screw moment of inertia is greater than that of the
motor shaft. The mechanical windage and frictional losses in the motor are
assumed to be negligible in order to approximate the back-EMF constant Kb
to the armature torque constant Ka , and is shown in detail by Sen [9].
The dynamics of the vertical (z-axis) lead screw are obtained with the
gravitational acceleration (g = 9.8 m/s2 ) treated as a constant disturbance
on the system. The system model diagram, consisting of the DC motor and
lead screw, is shown in Figure 4. The system model diagram is used to derive
the system equations by using Newton’s laws of motion, with the relationship
between motor torque and load weight provided by Hollander and Sugar [8].
The static friction coefficient is assumed to have a value of 0.06, which is the
value for lubricated metals [8].

7
The definitions of the parameters and quantities used to describe the lead
screw and motor system are given as:
m - Load mass (kg)
D - Dynamic friction coefficient (N.s/m)
µ - Static friction coefficient
r - Lead screw radius (m)
L - Lead screw length (m)
l - Pitch (m) (m/rev)
K - Pitch constant (m/rad)
τmotor - Motor torque (N.m)
τinertia - Lead screw inertial torque (N.m)
τmech - Resultant mechanical torque (N.m)
ωm - Motor angular velocity (rad/s)
Ra - Armature resistance (Ω)
Jl - Lead screw moment of inertia (kg.m2 )

+
Ia
z
+

mg m mz̈ D ż
+
Va E = Kb ω m

τinertia = J1 ω̇m
+
τmotor = Ka Ia

Figure 4: System model diagram.

8
The equation that describes the dynamics of the motor is obtained by using
Kirchoff’s voltage law, as shown in equation 1.

Va = Ia Ra + Kb ωm (1)
sinα + µcosα
τmech = rFw
cosα − µsinα
 
l
α = arctan
2πr
l = 1.25 mm
r = 4 mm
∴ α = 2.847 ◦
⇒ τmech = 0.4403 × 10−3 × Fw
(2)

Let:

γ = 0.4403 × 10−3
(3)

Define:

τmech = τmotor − τinertia

The resultant force Fw is the sum of the opposing forces acting on the mass
for motion in the positive z-direction. The equation that describes the rela-
tionship between the torques and opposing forces is shown in equation 5.

Fw = D ż + mg + mz̈ (4)
τmotor − τinertia = γ (D ż + mg + mz̈) (5)
ż = Kωm

The mass accelerated by the vertical lead screw, m, is approximately 1.3 kg.
The armature resistance of the brushed DC motor is provided by the data
sheet with a value of 0.205 Ω [11]. The armature torque constant Ka is
approximately 10.25 × 10−3 N.m/A [11]. The back-EMF constant Kb is the
same as the armature constant with a value of 10.25 × 10−3 V.s/rad. The
pitch per radian of rotation K for the standard 8 mm diameter lead screw is
calculated as 1.25×10
−3
2π
, which is equal to 0.1989 × 10−3 m/rad.

9
 dωm
τinertia = Jl
dt
Jl
τinertia = z̈
K
1
Jl = Mr 2
2
M = πr 2 Lρsteel
ρsteel = 7850 kg/m3
L = 0.54 m
∴M = 0.213 kg (6)
⇒ Jl = 1.71 × 10−6 kg.m2
Jl
∴ = 8.57 × 10−3 kg.m.rad
K
The two separate systems, motor and lead screw, are combined into one
system with the appropriate substitutions shown in equation 7.
τmotor ż
Va = Ra + Kb (7)
Ka K
Ra γ ż Ra
Va = (D ż + mg + mz̈) + Kb + τinertia
Ka K Ka
   
Ra Jl Ra mγ DRa γ Kb Ra γmg
∴ Va = + z̈ + + ż + (8)
Ka K Ka Ka K Ka

The dynamic friction coefficient D is determined from the experimentation

results by estimating the constant maximum velocity (0.262 m/s), which
occurs at maximum voltage of 13.8 V , and substituting the two values into
equation 5 and equation 7. The dynamic friction value is calculated to be
approximately 81 N.s/m, 80 N.s/m and 575.6 N.s/m for the z, x and y-axis
lead screws respectively. The dynamic friction coefficient for the y-axis lead
screw is significantly larger than the other two lead screws since it is driven
by a different motor that has a fan attached to the motor shaft [12]. The
simplified state-space representation of the system is obtained by substituting
the listed variables and constants into equation 8.

10
 z = x1 , Va = u 1
ż = x2 = x˙1 , g = u2
z̈ = x˙2 = x¨1 , y1 = x1
 
DRa γ Kb
K1 = Ka
+ K
, y2 = x2
 
Ra Jl Ra mγ Ra γm
K2 = Ka K
+ Ka
, K3 = Ka

The state equations are given as:

       
x˙1 0 1 x1 0 0 u1
= + 1
x˙2 0 K2−K1
x2 K2
−K3
K2 u2
   
y1   x1
= 1 1
y2 x2

The non-linear friction in the system causes a deadzone effect to occur

whereby the input control voltage is not able to move the mass until it is
greater than a certain value. The value of the input voltage at which move-
ment occurs was found through experimentation to be approximately 5 V ,
3 V and 3.4 V for the z, x and y-axis lead screws respectively. The deadzone
effect is included in the Simulink s-function, which is used to model the sys-
tem. The derived state equations hold for the other two axes by equating the
gravitational acceleration input to zero and substituting in the appropriate
masses, motor constants and other adjusted parameters.

4.3 Controller design

A classic controller is used for the robotic arm and comprises a propor-
tional gain and integrator, with gain constants of Ki = 150 V.m−1 .s−1 and
Kp = 55 V.m−1 . The gain constants were determined from the model and
refined through experimentation. The integrator is limited in order to pre-
vent integrator-wind up [13]. All of the axes are controlled with identical
controllers.
A derivative term was not used since it was found from experimentation
that it amplified noise and increased the motor armature current signifi-
cantly (∆Ia = 3 A). The controller was designed to optimize the actuation
system in terms of power consumption, tracking error and rise time. The
reduction of power consumption is achieved by switching the PWM off when
the position error is within a 1 mm error band for a duration of 5 s. The
trade-off, however, is that there exists some steady-state and overshoot error
in the step response.

11
5 System Performance and Analysis
The system performance was measured using sinusoid, ramp and step inputs.
The rise time and tracking error are obtained from the step response and
ramp response respectively. The steady-state error and settling time are also
obtained from the step response of the system. The deadzone time is the
duration that the system remains stationary due to non-linear friction, and
is most evident in the sinusoidal response. The reliability of the system was
measured by observing the accumulated error for a specified time limit. These
results were compared to simulations based on the Newtonian models. The
various system responses for the z-axis lead screw are shown in Figures 5, 6
and 7. The results for the other two axes are similar.
The overall results of the system performance in terms of responsiveness
and resolutions are shown in Table 1. The encoder resolution is determined
by dividing the lead screw pitch by the number of opaque sectors (4 dark
segments per axis encoder).

12
 X: 0.63
0.12 Y: 0.11
X: 2.48 X: 4
Y: 0.10 Y: 0.10
0.10
z-position (m)

0.08
Desired step
0.06 Experimentation
Simulation

0.04

0.02

0
0 1 2 3 4 5
t (s)

Figure 5: Step response of the z-axis lead screw.

Table 1: Experimentation results of system performance.

Parameter x-axis y-axis z-axis

Steady-State Error (mm) 0.9 0.313 0.9
Rise Time (s) 0.35 0.28 0.41
Overshoot (%) 16 16.3 11.9
2 % Settling Time (s) 1.31 0.95 1.21
Deadzone Time (s) 0.33 0.35 0.33
Encoder Resolution (mm) 0.375 0.313 0.313
Resolution (mm.pixel−1 ) 0.586 0.586 0.576

13
 X: 2.18
0.16 Y: 0.151
0.14 X: 3.62 X: 4.55
Y: 0.151 Y:0.153
0.12
z-position (m)

0.10
Desired ramp
0.08 Experimentation
Simulation
0.06
0.04
0.02
0
-0.02
0 1 2 3 4 5
t (s)

Figure 6: Ramp response of the z-axis lead screw.

The results show that the system is able to perform satisfactorily according
to the success criteria stated in Section 2. It is noted that the rise time of
y-axis lead screw is the fastest, which is due to the fact that the motor has
a higher rated speed compared to the other two DC motors [12][11].
The error accumulation in the system was found to be a function of speed
and sudden changes in motor direction. The cumulative error was therefore
quantified using two different waveforms: a low-frequency sinusoid and a
higher harmonic triangular waveform. The sinusoid and triangular waveforms
were each applied separately for a duration 68 s and 41 s respectively. The
offset error was recorded at the end of each test with the average results
shown in Table 2.
The cumulative error is due to the inertia of the masses opposing sudden
changes in velocity, and since the motor direction switches electronically,
results in an increment or decrement of the position. The error therefore
accumulates when there are sudden changes in direction at a reasonable
speed. An error is also associated with the incremental measurement made
by the low-cost optical encoders, which can be corrected by using alternative
encoders. The error accumulation is largest as expected for the triangular
waveform, as shown in Table 2.

14
 Desired sinusoid(0.25 Hz)
0.15 X: 1 Experimentation
Y: 0.102 Simulation
0.10

0.05
z-position (m)

-0.05

-0.10
X: 3.34
-0.15 Y: -0.102

-0.20
0 1 2 3 4 5
t (s)

Figure 7: Sinusoidal response of the z-axis lead screw.

5.1 System limitations

The telesurgery robotic arm system has a number of limitations viz.:

• Average maximum speed of 0.28 m/s.

• Reasonable mechanical vibrations and resonance.

• Mechanical stability varies as a function of z-position.

• Maximum instantaneous power consumption of 360 W .

• Detection is vulnerable to infrared noise.

• Inertial effects cause error accumulation.

• Controlled system exhibits overshoot and steady-state error.

The majority of the limitations are due to the mechanical aspect of the system
and can be improved with further design and development.

15
 Table 2: Experimentation results of system error accumulation.

Parameter Axis
x y z
Waveform (m) 0.1 × sin(2π(0.25)t)
Offset Error (mm) 0.938 3.73 2.5
Displacement (m) 6.8 6.8 6.8
Error (%) 0.0138 0.0549 0.0368
Waveform (m) 0.1 × tri(2π(0.375)t)
Offset Error (mm) 11.6 7.34 24.1
Displacement (m) 4.4 4.3 4.4
Error (%) 0.264 0.171 0.548

6 Conclusion
The objective of remote surgery is tested through the design and implemen-
tation of a relatively inexpensive robotic arm. The robotic arm consists of
the detection system, which is implemented using the Nintendo Wii remote
technology, and the actuation system which is implemented using a lead
screw design. The robotic arm is able to track the surgeon’s forearm with
less than 1 mm steady-state error and with a rise time of less than 0.5 s.
The non-linear static friction, however, creates a deadzone effect that delays
the motion of the robotic arm by 0.35 s.
The detection and encoder resolutions are both submillimeter, which en-
sures that the robotic arm can move and operate precisely during surgery.
The error accumulated in the system is below 4 mm when the robotic arm
is operated smoothly and at slower speeds. The end-effector of the robotic
arm could be further developed by including the surgeon’s precise finger and
hand movements measured using more sensors.

16
Acknowledgment
The authors would like to thank Professor David Rubin and Mr Harold Fel-
lows at the University of Witwatersrand for their guidance, advice and sup-
port.

References
[1] G. Bekey, R. Ambrose, V. Kumar, A. Sanderson, B. Wilcox, and
Y. Zheng, “International assessment of research and development in
robotics.” World Technology Evaluation Center, Inc, Tech. Rep., 2006.

[2] D. Holt, A. Z. J. Abramson, and R. Somogyi, “Robotic surgery and tele-

surgery: Basic principles and description of a novel concept.” University
of Toronto Medical Journal, vol. 82, no. 1, pp. 52–54, December 2004.

[3] A. Mulder, “Hand centered studies of human movement project human

movement tracking technology.” Simon Fraser University, Tech. Rep.,
1994.

[4] Y. Tao and H. Hu, “3D arm motion tracking for home-based rehabilita-
tion.” Proceedings of the 3rd Cambridge Workshop on universal access
and assistive technology, Tech. Rep., April 2006.

[5] J. Rosen, M. Lum, and D. Trimble., “Spherical mechanism analysis of a

surgical robot for minimally invasive surgery analytical and experimental
approaches.” University of Washington, Seattle, Tech. Rep., January
2005.

[6] B. Challacombe and P. Dasgupta, “Telemedicine-the future of surgery.”

The Journal of Surgery, vol. 1, no. 1, pp. 15–17, 2003.

[7] A. Pantanowitz., “Surgery environment assistant and automated control

using infrared light tracking.” University of Witwatersrand, Johannes-
burg, Tech. Rep., 2008.

[8] K. W. Hollander and T. G. Sugar, “Design of lightweight lead screw ac-

tuators for wearable robotic applications.” Journal of Mechanical Design
by the Arizona State University, vol. 128, no. 1, pp. 644–648, 2006.

[9] P. C. Sen, Principles of Electric Machines and Power Electronics.,

2nd ed. John Wiley and Sons, 1997.

17
 [10] SLA7024M, Datasheet, “Two-Phase Stepper Motor Unipolar Driver
IC,” http://www.datasheetcatalog.com, Last accessed 11 October 2009.

[11] FLRS-770PM-7038, Datasheet, “DC brush motor.”

http://www.mantech.co.za, Last accessed 26 September 2009.

[12] JRS-550PM-7024, Datasheet, “Fulling DC brush motor.”

http://www.mantech.co.za, Last accessed 26 September 2009.

[13] R. S. Burns, Advanced Control Engineering., 1st ed. Butterworth-

Heinemann, 2001, ch. 4, p. 91.

18

View publication stats