Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

Contents lists available at ScienceDirect

Robotics and Computer–Integrated Manufacturing

journal homepage: www.elsevier.com/locate/rcim

Design of a force-controlled end-effector with low-inertia effect for robotic

polishing using macro-mini robot approach
Abd El Khalick Mohammad a,b,∗, Jie Hong a, Danwei Wang a,∗
a
School of Electrical and Electronics Engineering, Nanyang Technological University, 50 Nanyang Avenue, 639798, Singapore
b
Mechanical Engineering Department, Faculty of Engineering, Assiut University, Assiut 71515 Egypt

a r t i c l e i n f o a b s t r a c t

Keywords: In this paper, the novel design of a force-controlled end-effector for automated polishing processes is presented.
Force-control The proposed end-effector is to be integrated into a macro-mini robot polishing cell. The macro robot (in this
Low-inertia study, it is a six-axis industrial robot) is used to position the mini robot (the proposed end-effector) according
Robotic polishing and macro-mini robot
to the workpiece profile while the mini robot controls the polishing force. Th end-effector has a polishing head
that can be extended and retracted by a linear hollow voice coil actuator to provide tool compliance. The main
advantage of the proposed design is that it allows this motion without extending or retracting the polishing motor
nor spindle, which reduces the inertial effects that may results in undesired vibrations. By integrating a force
sensor, the polishing force is measured and fed back to the controller to regulate it according to the polishing
pre-planned requirements. The effectiveness of the proposed device to track a certain desired force with step
changes under different feed rates has been examined through polishing experiments. The results demonstrate
the effectiveness of the presented device to reduce the vibration and achieve remarkable force tracking.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction markable positioning accuracy and splendid ability to simultaneously

Polishing process is considered as one of the essential final machin-
ing processes in various precision industries including die and mould
manufacturing, airfoils, camshaft, crankshaft and sculpture. It is used
to remove surface and subsurface damages and improve its roughness
[1–3]. However, the polishing processes of these parts are primarily
conducted manually, which is not only time-consuming and exposes la-
borers to high noise levels and metal dust environments, but also it is
difficult to maintain a stable polishing operation for long time [4]. For
example, to manufacture a mold or die, the time spent on the polish-
ing process accounts for 37 − 50% of the total manufacturing time [5].
In addition, labors conducting manual polishing for long time may get
”vibration white finger” or other musculoskeletal diseases [6]. Further-
more, to obtain quantitative and qualitative processing, some compa-
nies may have difficulty in recruiting and training sufficient numbers of
highly skilled manual workers [7].
In order to address the above mentioned limitations, some industries
are strongly motivated to seek and implement alternative solutions in
their manufacturing processes such as computer numerically controlled
(CNC) machines and industrial robots[8–10]. CNC machines have re-
adjust the trajectory, posture, and force during polishing tasks [11].
However, the limited available working space of these machines usu-
ally leads to process one part in multiple stages and restricts the size
of the workpiece. In addition, special fixtures and techniques to pro-
duce the surfaces with complex shapes are needed [12]. Apart from the
working space, polishing processes do not require a high positioning ac-
curacy otherwise human operators could not be able to perform them
nevertheless it requires accurate force control.
In recent years, robotic machining and finishing have attracted many
researchers due to its advantages compared with CNC machines, such
as low cost, higher flexibility and greater capability of integration with
actuators, sensor and different end-effectors [13]. Furthermore, indus-
trial robots deal with various types of workpieces, such as large-sized or
complex workpiece without any need to special fixtures. Hence, indus-
trial robots become an effective and economical solution for material
removing process from geometrically complex workpieces regardless of
the workpiece size [14].
In order to achieve a human-like polishing, the polishing tool should
have some flexibility which generally includes passive compliance or ac-

∗
Corresponding authors at: School of Electrical and Electronics Engineering, Nanyang Technological University, 50 Nanyang Avenue, 639798, Singapore. Tel.: +6582051952.
E-mail addresses: abdelkhalick_m@yahoo.com, abdelkhalick_m@aun.edu.eg (A.E.K. Mohammad), edwwang@ntu.edu.sg (D. Wang).

http://dx.doi.org/10.1016/j.rcim.2017.05.011
Received 9 November 2016; Received in revised form 19 March 2017; Accepted 26 May 2017
Available online 1 June 2017
0736-5845/© 2017 Elsevier Ltd. All rights reserved.
A.E.K. Mohammad et al. Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

tive compliance control. In passive compliance control, by using a pas-

sive mechanical element such as springs, the contact force between the
workpiece and the polishing tool is converted to a natural obedience de-
formation. On the other hand, the active compliance control, also known
as force control, employs a closed control loop to regulate the contact
force between the polishing head and the workpiece [4,15,16].
The later approach is categorized according to hardware into two
techniques; the first techniques is through-the-arm’s force control sys-
tem and the second one is around-the-arm by external active end-
effector. Although through-the-arm technique requires minimal hard-
ware, the slow response of the robot arm due to its high inertia limits the
bandwidth of the system. Alternatively, in around-the-arm technique,
the industrial robot arm, also called macro robot, traces a nominal path
around the workpiece while the force adjustments are performed by the
end-effector, also called mini robot [17]. Because only the mini robot
(usually it is a light actuator with a high bandwidth control loop) con-
trols the contact force, it allows a higher bandwidth and accuracy to be
achieved [18].
Researchers have developed several approaches from hardware
and software perspectives to improve the macro-mini robot. Bone
and Elbestawi [18] designed a unit (mini robot) that allows robot-
independent positioning along two orthogonal axes with high accu-
racy and bandwidth. They employ a combination of DC servo motors
with ball screws for the mini robot. Gloria et al. [19] integrated a two-
Fig. 1. Conventional structure of macro-mini robot system for polishing tasks.
degree of freedom mini-manipulator into a parallel kinematic mecha-
nism. The mini-manipulator allows normal force control to maintain
constant chamfer depth and it controls the tangential position of the
spindle along the feed direction. Because of the limited range of the
mini-manipulator, it may reach its workspace limit while the macro fol-
lows the nominal path without compensating for the positional discrep-
ancy if the workpiece has a large geometrical error. In order to solve this
problem, Arifin [20] proposed a motion control framework for macro-
mini manipulator such that the position control of the macro presents
in the same axis as the force control of the mini. Furthermore, Liao et
al. [21] designed an active axial-compliant force end-effector that con-
sists of three pneumatic cylinders. The cylinders are evenly distributed
and constrained to move only in the direction of the tool axis to allow
the spindle to be held in the center of the moving platform. Wu et al.
[22] proposed an adaptive neural network compensator to the mini’s
control system to eliminate the dynamic coupling effect coming from
the macro system in real time. Ma and Yang [23] proposed a three-
legged prismatic-prismatic-spherical parallel manipulator for a dexter-
ous 3-DOF force-controlled end-effector module. Lew [24] developed
a flexible micro/macro-manipulator can make smooth contact with a
rigid surface without instability. The controller is combined with the
force damping (FD) controller [25] and the inertial force active damp-
ing (IFAD) controller [26]. The FD controller regulates the contact force
while the IFAD controller dissipates the impact. Roveda et al. [27] devel-
Fig. 2. Comparison between the conventional and the proposed structure of macro-mini
oped a strategy that allows to track a desired force while compensating
robot system.
for the robot base dynamics through the estimation of the interacting
environment stiffness and the robot base state. Furthermore, Lew et al.
[28] proposed an active damping controller for a manipulator mounted
on a compliant base.
Although the above mentioned designs improve the performance of design of the mini robot to reduce the moving inertia by the mini robot.
the macro-mini robot approach, a common drawback is existing in all Fig. 2 schematically shows the proposed modification to the conven-
polishing devices those integrated in macro-mini robot systems. The tional macro-min robotic polishing system. As we can see from the fig-
drawback is moving of the polishing motor, spindle and polishing head ure, the proposed design allows to control the polishing force between
using the mini robot as shown in Fig. 1. Generally, the loads of these the polishing head and workpiece without moving the polishing mo-
items are considerable specially the polishing motor. In addition, the tor nor the spindle. The effectiveness of the proposed design is verified
holding force of the mini robot is expected to be relatively small due to through real polishing experiments.
its size limitation. Hence, if these loads are carried by the mini robot, The rest of this paper is organized as follows: Section 2 presents the
the holding force of the mini robot needs to be increased and conse- mechanical design, system modeling and controller design for the pro-
quently its physical size and inertia should be increased. Because the posed mini robot. The hardware and software of the experimental setup
macro robot carries the mini robot, a large macro robot is needed if is presented in Section 3. Discussion of the experimental results and ver-
the mini becomes heavier and thus the total cost of the system is in- ification for the proposed system are presented in Section 4. Finally, the
creased. In order to address this problem, the authors propose a novel conclusions and future works are given in Section 5.

55
A.E.K. Mohammad et al.
Fig. 3. Sectional view of the proposed end-effector shows its internal structure by INVENTOR software.
2. Mechanical design, modeling and control of the proposed
end-effector
2.1. Mechanical design
Fig. 3 shows a sectional view of the proposed force-controlled end-
effector. The polishing motor (7) provides the rotational motion of the
polishing pad (15) through a mechanical transmission as follows: The
polishing spindle (1) is supported by the bearing (2) to carry the radial
loads and the axial loads are supported by another bearing located in
the polishing motor housing. The polishing spindle (1) consists of three
sections; the upper section is connected to the polishing motor (7); the
middle section transmits the rotational motion of the upper section to
the lower section; and the lower section is a spline shaft that contains a
raceway grooves. The spline nut (14) consists of an outer cylinder with
raceway grooves and ball elements that is designed and manufactured to
achieve a reliable smooth motion. Hence, the ball spline (lower section
of the polishing spindle (1) and spline nut (14)) can transmit torque
and allows for a relative linear motion between the spline shaft and
the spline nut simultaneously. The spline nut (14) is connected to the
polishing head (13) which is connected to the polishing pad (15). Hence,
the power of the polishing motor (7) is transmitted to the polishing pad
(15) to conduct the mechanical polishing.
During robot interacting with the external environment, a fine con-
trol of the interaction force is needed to avoid, or at least to limit, the
force overshoots during the workpiece approaching and task execution
[29,30]. Generally, there are two known approaches of force control
used in automated surface finishing process. The first approach, trough-
the-arm force control, applies force using the position of all robot joints.
However, low stiffness, small actuation bandwidth and low positioning
resolution of a large industrial robot are seen to degrade the perfor-
mance of this approach. On the other hand, the second method, around-
the-arm force control, uses the robot for positioning only, and applies a
controlled force through an auxiliary compliant end-effector which is an
additional axis of motion [14,31–33]. In the proposed setup, the indus-
trial robot carries the proposed end-effector and responsible to control
its position. The end-effector is responsible to control the force in the
axial direction.
56
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
To control the force in the present design, a linear hollow voice coil
motor (HVCM) is used as shown in Fig. 3. The HVCM consists of two
separate parts; the magnetic housing (4) and the coil (5). The magnetic
housing (4) is stationary and it is attached to its housing (3) which is
connected to the polishing motor frame. To ensure the rigidity of the de-
sign and prevent the vibrations during operation, the coil (5), which is
moving linearly relative to the magnetic housing, is attached to a con-
necting plate (10) which is bolted to the sliding rod (9) of the linear
guide (8). The linear motion of the connecting plate (10) is transmitted
to the inner side of the bearing (6) via the force transmission plate (12)
and force sensor (11). We selected this position for the force sensor to al-
low accurate force monitoring since it is very close to the polishing pad.
The bearing (6) is an angular contact thrust bearing. Hence, it carries
the linear motion of the force transmission plate (12) to the polishing
head (13) while allowing relative rotational motion between the polish-
ing head (13) and the connecting plate (10). Hence, the polishing force
can be controlled by adjusting the position of the polishing pad through
the linear HVCM.
The main advantage of this design is that the polishing motor and
spindle are kept stationary unlike the conventional design. These two
parts have relatively high loads which require high-force linear actuator
(that may enlarge the end-effector weight and accordingly requires a
high-payload robot arm). In addition, moving of this two items may
cause high noise and vibration of the linear actuator which degrade
the control performance. Alternatively, in the presented design, only
the polishing head, which is a relatively light load, is allowed to move
linearly to achieve the desired polishing force.
2.2. System modeling
The voice coil motor generates a proportional force to the current
that flows through the motor coil. This force is approximately constant
in the specified working range of the motor. In this study, the voice
coil motor is driven by a servo amplifier running in the current control
mode. Hence, the control voltage signal is proportional to the motor
current and as a result proportional to the motor force. The relationship
between the motor force and the control voltage signal is as follows:
𝑓𝑚 = 𝐾𝑚 𝑢. (1)
A.E.K. Mohammad et al. Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

Fig. 6. Polishing of two surfaces under different polishing parameters.

back voltage is
𝑣 = 𝐾𝑠 𝑓 = 𝐾𝑠 (𝑓𝑚 + 𝑓𝑝𝑟𝑒 ). (3)
where v and Ks are the feedback voltage and amplifier gains, respec-
tively.

2.3. Controller design

Fig. 4. The experimental setup includes an ABB IRB2600 industrial robot mounted by
the proposed end-effector. Assume that the desired contact/polishing force to be tracked by the
end-effector is given as fd and the actual force f is measured. The force
tracking error ef is defined as follows:
𝑒𝑓 = 𝑓𝑑 − 𝑓 . (4)
Substituting Eqs. (1) and (2) into Eq. (4), we have
𝑒𝑓 = 𝑓𝑑 − (𝐾𝑚 𝑢 + 𝑓𝑝𝑟𝑒 ). (5)
In order to reduce the tracking error results from modeling error and
disturbance, we propose the following controller in this study:
( )
𝑡 𝑑𝑒𝑓
𝑢 = 𝑓𝑑 − 𝑓𝑝𝑟𝑒 + 𝐾𝑝 𝑒𝑓 + 𝐾𝑖 𝑒𝑓 𝑑𝑡 + 𝐾d ∕𝐾𝑚 . (6)
∫0 𝑑𝑡
where Kp , Ki and Kd are the proportional, integral and derivative gains,
respectively. Substituting Eq. (6) into Eq. (5), we have
𝑡 𝑑𝑒𝑓
(1 + 𝐾𝑝 )𝑒𝑓 + 𝐾𝑖 𝑒𝑓 𝑑𝑡 + 𝐾d = 0. (7)
∫0 𝑑𝑡
Taking the time derivative of the above equation, we have
𝑑 2 𝑒𝑓 𝑑𝑒𝑓
𝐾d + (1 + 𝐾𝑝 ) + 𝐾𝑖 𝑒𝑓 = 0. (8)
𝑑𝑡2 𝑑𝑡
By appropriate selection of Kp , Ki and Kd , Eq. (5) becomes asymptot-
ically stable and hence the force tracking error ef converges to zero
asymptotically. The ideal derivative in Eq. (6) has a very high gain for
high frequency signals results from force measurements. Hence, it may
generate large variations of the control signal. In order to reduce this ef-
fect, we replace the ideal derivative by a filtered derivative as follows:
Fig. 5. The hardware (lower part including RAPCON platform, donut-shape Forsentek
force sensor, amplifier, PC, HVCM and its drive) and software (upper part) used to imple- 𝑑𝑒𝑓 𝑁𝑆
ment the force control of the proposed end-effector. 𝐾d → 𝐾d .
𝑑𝑡 𝑆 +𝑁
where S denotes to def /dt. It can be seen that the ideal derivative is
where fm , Km and u are the force generated by the motor, motor with filtered using a first order system with a time constant of 1/N.
amplifier gain and control signal, respectively. The contact/polishing
force f is a sum of the motor force and the pre-load from the polishing 3. Experimental setup
head inertia fpre as follows:
3.1. Macro-mini robot setup
𝑓 = 𝑓𝑚 + 𝑓𝑝𝑟𝑒 . (2)
The contact/polishing force is measured by a force sensor and am- The experimental system in this study consists of a macro robot arm
plified by an amplifier to be fed back to the controller. Hence, the feed and a mini robot (the proposed end-effector). The macro robot is an ABB

57
A.E.K. Mohammad et al.
Fig. 7. The experimental results of Exp. No. I for force tracking under different feed rates without polishing for the proposed and conventional systems: (a) Feed rate of 5 mm/s, (b)
Feed rate of 10 mm/s and (c) Feed rate of 20 mm/s.
IRB 2600/20/1.65 six-axis industrial robot arm as shown in Fig. 4. The
robot arm has a handling capacity of 20 kg and 1.65 m reach and is
controlled by ABB IRC5 M2004 industrial controller. In addition, the
robot arm is integrated with a force control package which includes an
SRI six-degree of freedom force/torque sensor with a capacity of 165 N
(in the polishing direction) and a software for force control.
The mini robot includes a ball spline (model No. NB SSPFS10-1-200)
of 10 mm shaft diameter and 21 mm outer housing diameter. The bidi-
rectional thrust bearing (model No. BEAS 030062-2RS1) has a stiffness
of 870 N/μm and inner and outer diameters of 30 and 62 mm, re-
spectively. In addition, the linear guide (model No. SLHB16UU1x110)
is 36 × 45 × 54 mm with a sliding rod of 16 mm diameter. The polishing
head of the end-effector consists of a polishing pad of 22 mm diameter
and it is driven by an AC motor with a rated input power of 850 W and
six-level speed (2800–11000 r.p.m.).
3.2. Hardware implementation
Fig. 5 (lower part) shows schematically the hardware used to con-
trol the contact/polishing force in the proposed end-effector. The system
consists of a linear HVCM (model No. LVCM-070-038-01) with a stroke
of 12.7 mm and a maximum constant power of 45 W. The motor gen-
erates a force proportional to the current that flows through the motor
coil (24.8 N/A). This force is almost constant in the specified stroke
range of the motor. The HVCM is driven by a 25A8 PWM servo drive
running in the current/torque control mode (i.e. the input voltage sig-
nal is proportional to the motor current). The gain of current control
loop is set to be 4.0 A/V in this study. If the motor is energized, a force
is generated between the end-effector and the workpiece. This force is
58
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
measured by a donut-shape Forsentek force sensor which is connected
to a voltage amplifier to amplify the output signal. The gain of the load
cell and the amplifier is adjusted to be 0.03 V/N.
The force signal is fed back to the controller which is implemented
on a real-time RAPid CONtrol (RAPCON) prototyping platform for Mat-
lab/Simulink. The RAPCON platform offers a seamless interface be-
tween the HVCM, force sensor and Matlab/Simulink for implementation
of hardware-in-the-loop real-time control system with a sampling rate
up to 15.2 kHz. The force sensor is connected to the analog input (A7)
of RAPCON platform (0 to 5 V input range) with 12 bit resolution. In
addition, the servo drive of the HVCM is connected to the analog out-
puts (B0 and B1) (0 - 5 V output range) with 12 bit resolution. It should
be noted that two analog outputs are used as a differential reference
input to the servo drive using the inverting gain in order to achieve
the operating range of -5 to 5V. RAPCON platform is connected to a
PC’s serial port through a serial crossover cable. The developed MAT-
LAB code, SIMULINK model and RAPCON are uniquely integrated to
achieve real-time operation in Matlab under Windows as given in the
following subsection.
3.3. Controller software
The controller is implemented using the SIMULINK model shown in
Fig. 4 (upper part). A sum of step inputs are used to generate the desired
force to be tracked by the end-effector. The force signal is received from
the analog input A7 and then it is filtered using a low-pass filter with
a stop-band frequency of f1 (frequency units: normalized from 0 to 1),
stop-band attenuation of 60 dB and a passband ripple of 1 to remove the
high frequency noise. The voltage signal is converted to a force signal
A.E.K. Mohammad et al.
Fig. 8. The control signal of the linear HVCM in Exp. No. I: (a) Feed rate of 5 mm/s, (b) Feed rate of 10 mm/s and (c) Feed rate of 20 mm/s.
and compared with the desired force. A PID controller is employed to
regulate the desired force with a feed-forward controller by adjusting the
motor current through the analog output ports B1 and B2. An upper and
lower limits are used to prevent any undesired control signal to reach
the servo drive which may cause saturation of the motor. It should be
noted that the sampling frequencies for all input and outputs ports of
RAPCON board are set to 9.0 KHz in this study.
4. Experimental results
During polishing of surface with uneven desired material removal,
the polishing force should be regulated according to the required mate-
rial removal along the surface. The robotic polishing systems should be
able to track the desired force as well as switch from a certain desired
force to another one without any overshoot. We have verified this ex-
perimentally as shown in Fig. 6 where two surfaces are polished under
different polishing force, rotational speed and fed rate. We can see that
the material removal is different for both surfaces. Hence, the main ad-
59
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
vantage of force control in polishing process is to control the material
removal and roughness improvement.
In order to verify the effectiveness of the proposed system to deal
with such challenges, we conducted several experiments. In all exper-
iments, the macro robot carries the mini robot to be in contact with a
workpiece of stainless steel (50 × 300 × 5 mm). The macro robot is pro-
gramed to move along the workpiece with different feed rates as given
in the forthcoming explanation.
Exp. No. I: Force tracking under different feed rates during con-
tact
The objective of these experiments is to verify the performance of the
proposed system to track a desired contact force on the surface of the
workpiece. The performance of the proposed system is compared with
the performance of the force control system provided in the industrial
robot arm from ABB. In order to guarantee a fair comparison of both
systems, the actual contact/polishing force is measured by the donut-
shape Forsentek force sensor which integrated in the ed-effector and
A.E.K. Mohammad et al.
Fig. 9. The experimental results of Exp. No. II for force tracking under different feed rates with polishing for the proposed and conventional systems: (a) Feed rate of 5 mm/s, (b) Feed
rate of 10 mm/s and (c) Feed rate of 20 mm/s.
saved to the PC through RAPCON board. In addition, a low-pass filter
with a stop-band frequency of f2 , stop-band attenuation of 60 dB and
a passband ripple of 1 dB is used for both systems. This filter is used
for all next experiments and the values for f2 are given correspondingly.
Furthermore, the linear guide (part No. 8 in Fig. 2) is lucked and the
HVCM is turned off during the experiments of the conventional system.
As shown in Fig. 7, both systems are desired to track a force of
20 N followed by a step-up to 30 N followed by a step-down to 20 N.
This experiment is repeated under three different feed rates (5, 10 and
20 mm/s) for both systems. For the first feed rate, it can be seen that
the proposed system achieves smaller tracking error and the oscillation
is reduced. In addition, the overshoot during the step-up and step-down
is smaller than the conventional system. The interpretation for this per-
formance is that the proposed system responses faster than the conven-
tional system due to its reduced inertia effect. By increasing the feed
rate, the tracking of the conventional system becomes worse and the
settling time becomes longer. On the other hand, the proposed system
keeps the good tracking, small overshoot and faster settling time. In the
third feed rate, the proposed system demonstrates its effectiveness re-
gardless of the feed rate of the macro robot unlike the performance of
the conventional system which degrades as the feed rate is increased.
Fig. 8 shows that the control signal of the linear HVCM is smooth with-
out any saturation which provides a smooth motion of the motor without
any vibrations.
Exp. No. II: Force tracking under different feed rates during pol-
ishing
In this part of experiments, we turned the polishing motor on and a
real polishing is conducted on the workpiece surface. The polishing pad
is of a 100ϑ grit size and the polishing motor is rotating at a rate of 2800
60
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
r.p.m. Here, we use the same desired force and feed rates those used in
the previous experiments. Fig. 9 shows the tracking performance of the
proposed and the conventional systems. The conventional system has a
bigger overshoot after the polishing motor is turned on. On the other
hand, the proposed system adapts the desired force in a faster manner
than the conventional due to its reduced inertia effects. It has smaller
settling time, smaller overshoot and smaller tracking error compared to
the conventional system for different feed rates. The stable and smooth
control signals for the motor are shown in Fig. 10.
Exp. No. III: Low and high force tracking during contact
These experiments are carried out to verify the ability of the pro-
posed system to track low desired force as well as high desired force
with the same high performance. In these experiments, we changed the
desired force to be 8 N followed by a step-up to 12 N followed by a
step-down to 8 N and 40 N followed by a step-up to 50 N followed by
a step-down to 40 N as shown in Fig. 11. One can see that the perfor-
mance of the conventional system degraded if the desired force is low
(the steady state error is 5 N approximately). On the other hand, the
macro-mini system keeps high tracking performance regardless of the
desired force. Furthermore, the conventional system force profile con-
tains low frequency deviations result from the high inertia of the macro
robot. This inertia prevents the robot to reacts faster against disturbance.
On the other hand, the proposed system reacts faster to compensate the
disturbance. The control signal of the HVCM are shown in Fig. 12 for
these experiments.
Exp. No. IV: Low and high force tracking during polishing
The previous experiments are repeated under real polishing in this
part. It can be seen that the proposed system tracks the desired force
more accurately regardless of the desired force without big overshoot or
A.E.K. Mohammad et al. Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

Fig. 10. The control signal of the linear HVCM in Exp. No. II: (a) Feed rate of 5 mm/s, (b) Feed rate of 10 mm/s and (c) Feed rate of 20 mm/s.

Table 1 mini system is achieved from the modified mechanical design by reduc-
Controller Kp , Ki and Kd parameters and the stop-band frequen-
ing the moving loads by the mini robot, if an advanced controller is
cies f1 and f2 used in Exp. No. I - IV.
applied, certainly, the performance will further improved. In order to
Exp. No. Kp Ki Kd f1 f2 improve the performance by advanced controller, we presented a new
I 0.00005 0.15 -0.003 0.05 0.05 composite nonlinear feedback (CNF) controller.
II 0.00100 0.50 0.000 0.60 0.05 It is well known that the polishing process has a complex nature
III 0.00100 0.75 0.000 0.60 0.05 because it involves rubbing, ploughing and cutting simultaneously. In
IV 0.00100 0.75 0.000 0.60 0.05
order to achieve a good force tracking regarding of the system distur-
low frequency error as in the conventional system as shown in Figs. 13 bance such as uncertainties or modeling error, in this controller, the
and 14. The controller parameters (i.e. proportional, integral and deriva- linear feedback control signal is replaced by a nonlinear feedback that
tive gain), filter coefficient f1 for the control loop and filter coefficient is developed by using composite nonlinear feedback method as follows:
f2 for result comparison are given in Table 1.
( )
𝑡 𝑑𝑒𝑓
4.1. Composite nonlinear feedback controller 𝑢= 𝑓𝑑 + (𝐾𝑝 + Ψ)𝑒𝑓 + 𝐾𝑖 𝑒𝑓 𝑑𝑡 + 𝐾d ∕𝐾𝑚 ,
∫0 𝑑𝑡
However, the main objective of this research paper is to show that the −𝛽 [( )]
Ψ= exp(−(1 − 𝜁 2 )) − exp(−1) . (9)
significant improvement in the force control performance of the macro- 1 − exp(−1)

61
A.E.K. Mohammad et al. Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

Fig. 11. The experimental results of Exp. No. III for low and high force tracking without polishing for the proposed and conventional systems: (a) 8-12-8 N force tracking and (c)
40-50-40 N force tracking.

Fig. 12. The control signal of the linear HVCM in Exp. No. III: (a) 8-12-8 N force tracking and (c) 40-50-40 N force tracking.

where Ψ is a non-positive nonlinear function that depends on the output
and it is employed to change the damping ratio of the system. However,
the choice of Ψ is not unique; the function Ψ should varies from zero to
a certain negative value(𝛽), as the error varies from large value to zero
to change the system damping ratio. 𝜁 = (𝑓 − 𝑓0 )∕(𝑓𝑑𝑒𝑠 − 𝑓0 ), where f0
is the initial value of the force. The main advantage of using the CNF
controller is to achieve fast response with small overshoot. The fast re-
sponse increases the production rate and reduces the production time.
In addition, small overshoot prevents workpiece damage. This perfor-
mance is achieved by adjusting the feedback signal such that the closed
loop poles of the system are changed in a nonlinear manner. We have
verified the effectiveness of this controller via polishing experiments by

62
A.E.K. Mohammad et al. Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65

Fig. 13. The experimental results of Exp. No. IV for low and high force tracking with polishing for the proposed and conventional systems: (a) 8-12-8 N force tracking and (c) 40-50-40 N
force tracking.

Fig. 14. The control signal of the linear HVCM in Exp. No. IV: (a) 8-12-8 N force tracking and (c) 40-50-40 N force tracking.

using the proposed design. The experimental results shows that the CNF
controller improve the transient response as well as the steady state re-
sponse of the force tracking system for contact and polishing cases as
shown in Fig. 15.
In this research, we used a milled stainless steel surface. As shown
in Fig. 16, we can see that the milled surface contains a directional lines
from the milling tool. After mechanical polishing, these directional lines
are removed and a clear surface is achieved.
5. Conclusion and future works
Various precision industries realized the importance of the auto-
mated polishing technologies including CNC machines and industrial

63
A.E.K. Mohammad et al.
Fig. 15. Force tracking during contact (upper) and polishing (lower) by linear feedback (LF) and composite nonlinear feedback (CNF) controllers.
Fig. 16. Microscopic view of the milled surface and the polished workpiece.
robots. Robotic polishing has gained a lot of attraction due to its ad-
vantages over CNC machines such as lower cost and higher flexibility.
In robotic polishing, force control is an essential issue to improve the
surface roughness. In this paper, we presented a new design of a force-
controlled end-effector for robotic polishing tasks. Unlike the conven-
tional designs of macro-mini robot system, the proposed design reduces
the inertial effect results from moving the polishing motor and spindle
by the mini′s actuator. Thus, the mini′s holding force can be used effi-
ciently in the polishing process.
The proposed design is tested experimentally through a robotic pol-
ishing cell. Several experiments are conducted including force tracking
during contact and real polishing with different levels of desired con-
tact/polishing forces and feed rates. In addition, the ability to track low
force as well as high force is examined. The experimental results shows
that the proposed design gives a remarkable force tracking with smaller
overshoot, smaller settling time and smaller tracking error.
64
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
Since the market of industrial robots is dominated by closed archi-
tecture systems, it is hard to modify the closed control system of these
robots. The advantage of the proposed device is that it does not require
any change of the macro′s control system. It can be integrated as add-
on device to any existing robotic polishing system using few auxiliary
input and output ports which are usually available in most of the com-
mercial robot controllers. An adaptive control system and extension of
the mechanical design of the proposed design to achieve force control
in multi-axis are left for future works.
Acknowledgment
The authors gratefully acknowledge the support of the A ∗ STAR In-
dustrial Robotics Program Science and Engineering Research Council
Grant number 122510004.
A.E.K. Mohammad et al.
References
[1] X. Xie, L. Sun, Force control based robotic grinding system and application, in:
2016 12th World Congress on Intelligent Control and Automation (WCICA), 2016,
pp. 2552–2555, doi:10.1109/WCICA.2016.7578828.
[2] W. Lin, P. Xu, B. Li, X. Yang, Path planning of mechanical polishing process for
freeform surface with a small polishing tool, Rob. Biomimetics 1 (1) (2014) 1–15,
doi:10.1186/s40638- 014- 0024- 1.
[3] P. Xu, C.-F. Cheung, B. Li, L.-T. Ho, J.-F. Zhang, Kinematics analysis of a hybrid ma-
nipulator for computer controlled ultra-precision free form polishing, Rob. Comput.-
Integr. Manuf. 44 (2017) 44–56, doi:10.1016/j.rcim.2016.08.003.
[4] F. Tian, C. Lv, Z. Li, G. Liu, Modeling and control of robotic automatic polishing
for curved surfaces, {CIRP} J. Manuf. Sci. Technol. 14 (2016) 55–64, doi:10.1016/
j.cirpj.2016.05.010.
[5] L. Gven, K. Srinivasan, An overview of robot-assisted die and mold polishing with
emphasis on process modeling, J. Manuf. Syst. 16 (1) (1997) 48–58, doi:10.1016/
S0278- 6125(97)88405- 1.
[6] E. Kalt, R. Monfared, M. Jackson, Towards an automated polishing system - captur-
ing manual polishing operations, Int. J. Res. Eng. Technol. 5 (7) (2016) 182–192.
[7] L. Liao, F. Xi, A linearized model for control of automated polishing process, in:
Proceedings of 2005 IEEE Conference on Control Applications, 2005. CCA 2005.,
2005, pp. 986–991, doi:10.1109/CCA.2005.1507258.
[8] F. Nagata, Y. Kusumoto, Y. Fujimoto, K. Watanabe, Robotic sanding system for new
designed furniture with free-formed surface, Rob. Comput. Integr. Manuf. 23 (4)
(2007) 371–379, doi:10.1016/j.rcim.2006.04.004.
[9] E. Pitschke, M. Schinhaerl, R. Rascher, P. Sperber, L. Smith, R. Stamp, M. Smith, Sim-
ulation of a complex optical polishing process using a neural network, Rob. Comput.
Integr. Manuf. 24 (1) (2008) 32–37, doi:10.1016/j.rcim.2006.07.003.
[10] J. Ahn, Y. Shen, H. Kim, H. Jeong, K. Cho, Development of a sensor information
integrated expert system for optimizing die polishing, Rob. Comput. Integr. Manuf.
17 (4) (2001) 269–276, doi:10.1016/S0736- 5845(00)00057- 0.
[11] Y. Kakinuma, K. Igarashi, S. Katsura, T. Aoyama, Development of 5-axis polishing
machine capable of simultaneous trajectory, posture, and force control, {CIRP} Ann.
Manuf. Technol. 62 (1) (2013) 379–382, doi:10.1016/j.cirp.2013.03.135.
[12] M. Khalick, W. Danwei, Electrochemical mechanical polishing technology: recent
developments and future research and industrial needs, Int. J. Adv. Manuf. Technol.
(2016) 1–16, doi:10.1007/s00170- 015- 8119- 6.
[13] H. Huang, L. Zhou, X. Chen, Z. Gong, Smart robotic system for 3d profile turbine
vane airfoil repair, Int. J. Adv. Manuf. Technol. 21 (4) (2003) 275–283, doi:10.1007/
s001700300032.
[14] A.E.K. Mohammad, D. Wang, A novel mechatronics design of an electrochemical me-
chanical end-effector for robotic-based surface polishing, in: 2015 IEEE/SICE Inter-
national Symposium on System Integration (SII), 2015, pp. 127–133, doi:10.1109/
SII.2015.7404966.
[15] S. Yong-jie, D. Zheng, W. Long-shan, H. Li-yong, Polish force control in precise nc
polishing of aspheric parts, Opt. Precis. Eng. 19 (5) (2011) 1013–1021, doi:10.3788/
OPE.20111905.1013.
[16] C. Liu, C.-C. Chen, J.-S. Huang, The polishing of molds and dies using a compliance
tool holder mechanism, J. Mater. Process. Technol. 166 (2) (2005) 230–236, doi:10.
1016/j.jmatprotec.2004.08.021.
[17] A. Sharon, N. Hogan, D.E. Hardt, High bandwidth force regulation and inertia re-
duction using a macro/micro manipulator system, in: Robotics and Automation,
1988. Proceedings., 1988 IEEE International Conference on, 1, 1988, pp. 126–132,
doi:10.1109/ROBOT.1988.12036.
65
Robotics and Computer–Integrated Manufacturing 49 (2018) 54–65
[18] G. Bone, M. Elbestawi, Active end effector control of a low precision robot in debur-
ring, Rob. Comput. Integr. Manuf. 8 (2) (1991) 87–96, doi:10.1016/0736-5845(91)
90024-M.
[19] U.A. Tol, J.-P. Clerc, G.J. Wiens, Micro/macro approach for dexterity enhancement
of pkms, in: Proceedings of the Workshop on Fundamental Issues and Future Re-
search Directions for Parallel Mechanisms and Manipulators, 2002, pp. 34–39.
[20] A. Arifin, M. HAng, C. Lai, C. Lim, General framework of the force and com-
pliant motion control for macro mini manipulator, in: 2013 IEEE/ASME Inter-
national Conference on Advanced Intelligent Mechatronics, 2013, pp. 949–954,
doi:10.1109/AIM.2013.6584216.
[21] L. Liao, F.J. Xi, K. Liu, Modeling and control of automated polishing/deburring pro-
cess using a dual-purpose compliant toolhead, Int. J. Mach. Tools Manuf. 48 (1213)
(2008) 1454–1463, doi:10.1016/j.ijmachtools.2008.04.009.
[22] C.L. Ting-Ying Wu, S. Chen, An adaptive neural network compensator for decou-
pling of dynamic effects of a macro-mini manipulator, in: 2015 IEEE International
Conference on Advanced Intelligent Mechatronics (AIM), 2015, pp. 1427–1432,
doi:10.1109/AIM.2015.7222741.
[23] K. Ma, G. Yang, Kinematic design of a 3-dof force-controlled end-effector module,
in: 2016 IEEE 11th Conference on Industrial Electronics and Applications (ICIEA),
2016, pp. 1084–1089, doi:10.1109/ICIEA.2016.7603743.
[24] L. Jae, Contact control of flexible micro/macro-manipulators, in: 2014 IEEE In-
ternational Conference on Robotics and Automation, 4, 1997, pp. 2850–2855,
doi:10.1109/ROBOT.1997.606719.
[25] D. Whitney, Historical perspective and state of the art in robot force control, Int. J.
Rob. Res. 6 (I) (1985) 3–14, doi:10.1109/ROBOT.1985.1087266.
[26] L.J. Trudnowski, Vibration control of a microhlacro-manipulator system, Int. J. Rob.
Res. (1996) 26–31.
[27] L. Roveda, N. Pedrocchi, F. Vicentini, L.M. Tosatti, An interaction controller for-
mulation to systematically avoid force overshoots through impedance shaping
method with compliant robot base, Mechatronics 39 (2016) 42–53, doi:10.1016/
j.mechatronics.2016.08.001.
[28] J.Y. Lew, S.M. Moon, A simple active damping control for compliant base manip-
ulators, IEEE/ASME Trans. Mechatron. 6 (3) (2001) 305–310, doi:10.1109/3516.
951368.
[29] C.Y. Lai, Improving the transient performance in robotics force control using nonlin-
ear damping, in: 2014 IEEE/ASME International Conference on Advanced Intelligent
Mechatronics, 2014, pp. 892–897, doi:10.1109/AIM.2014.6878193.
[30] L. Roveda, N. Iannacci, F. Vicentini, Optimal impedance force-tracking control de-
sign with impact formulation for interaction tasks, IEEE Rob. Autom. Lett. 1 (1)
(2016) 130–136, doi:10.1109/LRA.2015.2508061.
[31] L. Gven, K. Srinivasan, Force controller design and evaluation for robot-assisted die
and mould polishing, Mech. Syst. Signal Process. 9 (1) (1995) 31–49, doi:10.1006/
mssp.1995.0003.
[32] E. Erlbacher, Force control basics, Ind. Rob. Int. J. 27 (1) (2000) 20–29, doi:10.
1108/01439910010305130.
[33] E. Erlbacher, Development of a force controlled orbital polishing head for free
form surface finishing, J. Prod. Eng. 4 (23) (2010) 269–277, doi:10.1007/
s11740- 010- 0221- x.