IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO.

9, SEPTEMBER 2012

3501

Development of a Spherical Air Bearing

Positioning System
Kok Kiong Tan, Member, IEEE, Sunan Huang, Wenyu Liang, Abdullah Al Mamun, Senior Member, IEEE,
Eng Kiat Koh, and Huixing Zhou

AbstractIn line with the general trend of the automation,

spherical motion systems are increasingly required in applications
such as robot joints, manipulators, orientation control systems,
and angular alignment systems. In this paper, a spherical air bearing positioning system (SABS) aiming at providing highly precise
rotational motions in two degrees of freedom is designed. The
SABS mainly consists of direct-drive voice coil actuators and pneumatic bearing. This paper presents the mechanical structure and
control system of the SABS. The model of the SABS is identified
based on adaptive control concepts. To eliminate the measurement
noise, a noise filter is designed on the basis of the model. Then,
an observer-based PID controller is designed and implemented on
a dSPACE development card. The experiment results show that
the designed controller can achieve higher precision and better
tracking performance of about ten times compared to that from
a traditional PID controller.
Index TermsAir bearing, filter design, PID control, position
control, voice coil actuators (VCAs).

N OMENCLATURE
B
L
a, b

x
xd

Magnetic flux density.

Effective length of each turn of the coils.
Parameters of the system.
Dirac delta function.
State variables of the system.
Desired performance of the system.
I. I NTRODUCTION

S THE LEVEL of automation rises in the modern world,

more and more equipment (particularly in industrial
applications) is required to achieve multi-degree-of-freedom
(DOF) (multi-DOF) motion. Moreover, those motions are also
required to be ever smoother, more accurate, and flexible.

Manuscript received May 24, 2011; revised August 20, 2011; accepted
September 16, 2011. Date of publication October 13, 2011; date of current
version April 13, 2012.
K. K. Tan, S. Huang, W. Liang, and A. A. Mamun are with the Department
of Electrical and Computer Engineering, National University of Singapore,
Singapore 117576 (e-mail: eletankk@nus.edu.sg; elehsn@nus.edu.sg; liangwenyu@nus.edu.sg; eleaam@nus.edu.sg).
E. K. Koh is with the Data Storage Institute, Agency for Science, Technology
and Research (A STAR), Singapore 117608 (e-mail: koh_eng_kiat@dsi.astar.edu.sg).
H. Zhou is with the College of Engineering, China Agricultural University,
Beijing 100083, China (e-mail: huixingzhou@yahoo.cn).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIE.2011.2171179

One example of a multi-DOF system is the spherical motor

which offers two- or three-DOF motion (roll, yaw, and pitch).
These spherical motion systems can be applied to various
applications, such as robot joints (include wrist, elbow, etc.),
manipulators, steering systems, vehicle wheels, machines
which require orientation control, medical applications (where
there is a need to adjust angular or swing movements), angular
alignment or positioning systems, and so on. Specifically, in
[1], in order to resolve the alignment challenges for doublesided nanoimprint lithography, a three-DOF spherical motion
system was designed. The relative orientation to the disk
and imprinter head can be efficiently manipulated with the
spherical motion system.
In conventional designs, at least two single-DOF motors
and mechanisms (like gears and link mechanisms) are used
to implement multi-DOF movements. Such a method is generally simple to design and control, but achieving a spherical configuration with it has several drawbacks as follows:
1) It is difficult to achieve high precision due to low stiffness,
mechanism backlash, and elasticity; 2) because of nonlinear
friction and thermal effects, the velocity is limited so that the
dynamic response is slow; 3) the output torque and efficiency
are also limited; 4) most of them are usually complicated in
mechanical structure, bulky in size, and heavy in weight; and
5) there is lack of flexibility. Thus, exploring new and better
methods for spherical motion system is necessary. Direct-drive
configuration is one of the important approaches to improve the
conventional multi-DOF spherical motion system. In the past
two decades, various direct-driven spherical motion systems
have been developed, which can be found in [1][8]. Different
from the conventional configuration, a direct-drive system does
not need any mechanism (e.g., gearbox) to convert movements;
therefore, it offers plenty of advantages. Precision and stiffness
can be increased, while friction is decreased, leading to a more
efficient motion. Furthermore, the direct-drive system allows a
faster positioning time due to its high torque output and low
moment of inertia. The device noise due to coupling effect is
reduced, and the working life is extended.
Generally, because of the installation of rolling element
bearings, direct-drive systems will incur friction, which degrades the control precision. In order to reduce friction, an air
bearing can be considered. Air bearing uses pressurized air to
provide an air film between two objects so as to support the
payloads. Comparing air bearing with other types of bearings
(include rolling bearings and fluid bearing), air bearing has low
viscosity, offers lower friction, and has smaller sensitivity to
temperature variation [9]. Meanwhile, it offers a number of

0278-0046/$26.00 2011 IEEE

3502

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012

Fig. 2.

Cross-sectional view of SABS.

Fig. 3.

Stator and rotor of SABS.

Fig. 1. Mechanical structure of SABS.

advantages to conventional precision positioning systems, such

as no contact, lack of stiction, and a clean operation. Hence,
adopting air bearing can greatly help a multi-DOF system to
improve its performance. Two multi-DOF spherical air bearing
systems are presented in [1] and [4], and another two different three-DOF precision positioning systems using direct-drive
motors/actuators and air bearing are illustrated in [10] and [11].
In this paper, to achieve a high-performance multi-DOF system for angular positioning, the design and implementation of
a novel spherical air bearing positioning system (SABS), which
combines a direct-drive electromagnetic motor with pneumatic
bearing, will be presented. Significantly, the SABS is designed
to provide precise rotational motions in two DOFs. This paper
consists of two parts: In the first part, the hardware components
of SABS, including the mechanical structure and the control
system, will be introduced in detail; in the other part, the
mathematic model of SABS will be derived, and a noise filter
and a controller for angular positioning will be developed.
II. D ESIGN OF S PHERICAL A IR B EARING S YSTEM
In this section, the design of SABS will be provided in
detail. The mechanical structure will be elaborated alongside
the working principle of SABS, followed by a presentation of
the control system.
A. Mechanical Structure
The mechanical structure of SABS is shown in Fig. 1. It
is cylindrical with an outer diameter of 182 mm and mainly
consists of two components: a hemispherical solid rotor with
a radius of 112 mm and a bowl-shaped stator for holding the
rotor with three adjustable legs. The material of the stator and
the rotor is aluminum alloy which is not ferromagnetic. Fig. 2
shows the cross-sectional view of SABS. When a thin air film
is introduced between the two components with a pneumatic
system, the rotor floats on the air film so that the friction
between the two surfaces is extremely low, due to the low
viscosity of air. Thus, the rotor can rotate in all directions freely
without any resistance.
However, achieving the angular position control by only
using air bearings is not possible. Therefore, a spherical motor

[which is made up of four voice coil actuators (VCAs)] is

designed, as shown in Fig. 2. In particular, the motor is used
to enable positioning in two directions. In what follows, the
two key components (VCAs and pneumatic system) will be
elaborated.
1) VCAs: A VCA is usually classified as a brushless dc
actuator which provides linear force and motion. One of the
common applications of VCA is in the positioning actuator of
the read-and-write head in a hard disk drive. Due to its simple
structure, it has many advantages such as miniaturization, rapid
response, high accuracy, high stiffness (i.e., high resistance to
deformation), and easy operation. Therefore, VCA is a popular
choice for mechatronic equipment, particularly for precision
positioning systems.
In the SABS, four VCAs are uniformly and symmetrically
distributed inside the stator and the rotor, as shown in Fig. 3.
Each VCA has the same structure and specification, and it is
composed of two parts: the coils mounted in the stator and
the permanent magnets (with magnetic yokes) in the rotor. In
addition, every two motors on the same axis are set as a pair;
each pair is used for one-DOF angular positioning around each
corresponding axis (X- and Y -axes), respectively. In this paper,
the pair of VCAs placed on the X-axis is named X-axis VCA
while the other one is named Y -axis VCA.

TAN et al.: DEVELOPMENT OF A SPHERICAL AIR BEARING POSITIONING SYSTEM

3503

Thus, the equivalent force along the z-axis also equals zero
(because Fd = Fd ), but a turning torque Md around the yaxis is inevitably generated
Md Fd d

Fig. 4.

Working principle of VCA. (a) Magnetic flux path. (b) Force analysis.

The working principle of each VCA is shown in Fig. 4.

More specifically, Fig. 4(a) displays the magnetic flux path
of the VCA, and Fig. 4(b) shows the distribution of VCA.
As can be seen in Fig. 4(a), two permanent magnets made of
NdFeB (neodymium iron boron) are placed side by side on one
soft iron yoke while the coils are immovable due to the fixed
stator.
The permanent magnets generate a stable magnetic field. Its
magnetic flux density B is a vector, which, in the Cartesian
coordinate system, can be resolved into three orthogonal components: Bx , By , and Bz . In particular, the displacement of
this VCA is quite small, and the coils rarely reach the end of
the magnets. Moreover, the value of Bz is minuscule which is
neglected in this case.
When the current is applied through the coils, several forces
are exerted on the coils as a result of the interaction between
magnetic field and electric current, as shown in Fig. 4(b). As
can be seen, the forces Fa . Fb , Fb , Fc , and Fc are generated
in the effect By , where, due to the symmetrical structure, Fb =
Fc and Fb = Fc , i.e., Fb + Fb + Fc + Fc = 0. Therefore,
all forces along the y-axis direction are canceled so that the
equivalent force in this direction equals zero. In other words,
there is no motion generated in the y-axis direction when the
current is applied. In the x-axis direction, according to the
Lorentz force law, the force Fa is given by

Fa = k By L N i

(1)

where k is a constant relating to the structure of the magnetic

flux path, By is the effective flux density, denotes the vector

cross product, L is the effective length of each turn of the

coils, N is the number of turns of the coils, and i is the applied
current. Moreover, (1) can be rewritten as follows:
Fa = kBy LN i

(3)

where d is the distance between Fd and Fd .

The torque Md is too small to affect the performance and
stability of SABS since the value of d is relatively small.
Therefore, this torque can be ignored in this system.
In conclusion, the VCA mainly generates a force in the
tangential direction of the rotor when the coils are energized.
Thus, by controlling the current of different pair of VCAs, the
spherical motor is able to implement angular positioning in two
DOFs, where the X-axis VCA is used for adjusting the angle
around the Y -axis (yaw) while the Y -axis VCA is for the angle
around the X-axis (roll).
2) Pneumatic System: The pneumatic system of SABS includes the spherical air bearing and other components which
are used to provide pressurized air. For air bearing, in terms
of the pressure generation principles, it can be categorized
as follows: 1) Hydrodynamic (or self-acting) typeair film
is generated internally by relative motion at high speeds;
2) Hydrostatic typeair film is provided from external pressure supply; and 3) Squeeze-film typeair film is created
by imposing oscillations on the nonmoving part of bearings.
Generally, the hydrodynamic type is common for fluid bearings
in which the viscosity of fluid (such as oil) is relatively high,
so it is hard for air bearings to utilize this type due to the low
viscosity of air. The squeeze-film type is frequently used for
damping applications. Moreover, it is easy to obtain and control
a designated air pressure by an external air compressor. As a
result, the hydrostatic type is applied in the SABS.
In order to maximize the stiffness of the air bearing and help
it maintain a constant air gap, it is better to preload the air
bearing, although air bearing can work without any preload.
In the SABS, a vacuum preload is mainly used, which avoids
unnecessary moving mass for preloading in the system.
The structure and the working principle schematic diagram
of hydrostatic air bearing in the SABS are shown in Fig. 5.
As can be seen, eight microholes (not all shown in the figure)
distributed in two circles uniformly and symmetrically are
used to transmit compressed air into the floating area, in turn
generating forces for sustaining the rotor. One intake is used
to evacuate gas so as to form the vacuum room, where vacuum
preloading is provided. The forces generated by the air pressure
can be calculated by the (4) shown as follows:

(2)
FP = P A

where By and L are scalars.

Meanwhile, arising from the force Fa , an equal but opposite
force Fa is produced on the magnets. Because the coils are
fixed as mentioned previously, the magnet part will be driven
under the action of the force Fa in the x-axis direction when
the coils are subjected to current.
Aside from the effect of Bx , a couple of forces Fd and Fd
in the z-axis direction are generated, which are approximately
equal in magnitude and opposite in direction but not collinear.

(4)

where P is the average pressure in the floating area or vacuum

room and A is the equivalent area of those areas.
Since SABS is a hydrostatic type, external clean pressurized
air is required for the bearing (i.e., an air compressor and filters
are required). Additionally, due to the vacuum preload method,
the air bearing needs a vacuum generator. Those components
and the air bearing compose the whole pneumatic system, as
shown in Fig. 6.

3504

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012

Fig. 7.

System diagram of SABS.

Fig. 8.

Angle measurement principle.

Fig. 5. Structure and working principle of air bearing in the SABS.

(a) Arrangement of air vents. (b) Design scheme and working principle.

Fig. 6. Pneumatic system of SABS.

B. Control System
Controlling an air bearing stage can be defined as influencing
it in such a way as to force it to operate according to certain
requirements. Moreover, the control objective of the SABS is
to find a control mechanism for every bounded smooth desired
output so that the controlled output converges to the desired output as closely as possible. This involves system modeling, parameter identification, eliminating noise, and controller design.
1) Hardware of Control System: A control system for the
SABS is developed to control each VCA in order to implement
angular positioning control. The control system and mechanical
components together compose the SABS. The system diagram
is shown in Fig. 7 which includes a sensor, motor drivers, and a
controller.
The sensor is a dual-axis position sensing diode (PSD),
whose outputs are bipolar voltage analogs of the X and the Y
position of the light spot centroid. The light spot is provided
by a compact laser generator which is mounted on the rotor.
Moreover, the PSD exports the measured values of the X and

the Y position separately. Therefore, the angle of roll and

yaw can be calculated by the sensor outputs, respectively. The
measurement principle for angles is shown in Fig. 8.
As can be seen, when the rotor is stationary, the light spot and
the center of the PSD are coincident; therefore, the outputs are
zero. When the rotor rotates around its pivot, an angle between
the axis of the rotor (which can be represented by the laser
beam) and the direction of gravity is formed. Meanwhile, the
distances between the light spot and the PSD central point are
measured by the PSD. Thus, according to the trigonometric
function, the angles can be calculated by (5) as follows:

= arctan( hy )
(5)
= arctan( hx )
where and are the angles of roll and yaw, x and y are the
outputs of the PSD, and h is the height between the PSD and
the rotors pivot.
Since the angle outputs of the SABS are designed to be less
than 10 , (5) can be linearized and written as

hy
.
(6)
hx
Two motor drivers are required to drive two pairs of VCAs,
respectively. Each drive consists of one high-voltage highcurrent operational amplifier (OPA 548), which controls the
current output in terms of the analog voltage control signal. The
continuous output current of OPA 548 is 3 A while the peak output current is 5 A which meets the requirements of the VCAs.
The drive circuit constructed by OPA 548 is shown in Fig. 9.

TAN et al.: DEVELOPMENT OF A SPHERICAL AIR BEARING POSITIONING SYSTEM

Fig. 9.

3505

Drive circuit of SABS.

Fig. 11. Spectrum of the SABS. (a) FFT in normal scale. (b) FFT in logarithmic scale.

Consider the control problem of the system in (7). The

definition of the filtered error S is given by
Fig. 10. Force caused by actuator.

The controller is the center part of the control system. In

the SABS, a dSPACE 1104 control development card and a
personal computer (PC) are used to implement the positioning
control.
2) Modeling of Air Bearing Stage: The use of the air bearing
stage for an accurate control depends greatly on the model
obtained for the controller. The purpose of the model presented
in this section is to map the relationship between input and
output at the electrical port of the air bearing system in a special
form that can be represented by a set of equations.
In the spherical system, there are two pairs of voice-coil actuators which can control the XY directions. These actuators
are placed inside the stator and rotor as mentioned before (see
Fig. 3). According to the theory of VCA, the force generated by
the actuator is given by
F = kBy LN i.
Applying the physical law, the SABS can be described by the
following mathematical equations:

m1 x
1 = F1 P 1
(7)
m2 x
2 = F2 P 2
where Pr , r = 1, 2, is a force caused by the weight of the air
bearing stage (see Fig. 10), x1 = R, and x2 = R, R is the
rotors radius of sphere. We assume that the actuators on XY
are decoupled.
3) Parameter Identification: Although the model structure
is built in the above section, the model parameters Pr and mr ,
r = 1, 2, are difficult to obtain. In this section, we design an
estimator to obtain the parameters of the system. The idea of
the algorithm is based on adaptive control concepts.

S = e + e

(8)

where is a constant, the error e = xd x, xd represents the

desired performance, and x = [x1 x2 ]T represents the state
variables of the system. Thus, the system can be rewritten as
kBy LN i
P
S = e + x
d +

= e + x
d + a bi
m
m

(9)

where a = P/m and b = kBy LN/m.

Since the parameters a and b are not known exactly, the
following adaptive control law is used:
i=

d + a

Kv S + e + x
b

(10)

with
a
= 1
b = 2 S

(11)
(12)

where a
and b are estimated values of a and b, respectively.
Substituting the control law into the system yields
S = a
bi Kv S

(13)

where a
=aa
and b = b b are the errors between the
estimated and actual values of a and b, respectively.
a2 + (1/2 )b2 . The time derivative
Define V = S + (1/1 )
of V is given by
2
2
V = 2Kv S 2 + 2S
a 2Sbi a
a
bb.
1
2

(14)

Substituting the adaptive learning laws produces

V = 2Kv S 2 .

(15)

3506

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012

This shows that S and b, a

are bounded. This also implies that
S is bounded. From (15), we have
t
2Kv S 2 d = V (0) V (t) V (0).

(16)

Applying Barbalats lemma, we obtain

lim S2 = 0.

Fig. 12.

Overall structure of PID control.

Fig. 13.

Block diagram of control system.

(17)

Although the online parameter estimation scheme guarantees

that the tracking error S converges to zero as t , this result,
however, is not sufficient to establish parameter convergence to
the true parameter values unless the input signal i is sufficiently
rich. The system noise (white noise) contributes toward the
cause.
4) Noise Filter: In the SABS, noise is introduced due to the
sensor, circuit, and amplifier. To analyze the signal noise, we
adopt a fast Fourier transform (FFT) algorithm. It is a popular
offline approach widely used to obtain information of the frequency distribution required for the filter design. Fig. 11 shows
the spectrum of the air bearing stage, where the frequency of
10 Hz is a working signal. It is observed that the noise exists
over a wide range with small amplitude. Moreover, the expected
value of the noise approaches to zero. Since it exhibits the
characteristics of white noise, an observer-type filter is used for
handling such noise.
In the presented state space model, the position of the air
bearing stage is available. Rewriting (7) into a compact form
yields

X = AX + B(bi a) + w
(18)
y = CX + v
where X = [x x]
T , y is the output (position), w and v are the
system noises, and


 
0 1
0
A=
, B=
, C = [1 0]
(19)
0 0
1
KBy LN
P
.
(20)
a= , b=
m
m

=X X
and A = A KC. Basically, the gain
where X
vector K is required to be designed to satisfy the following
condition: All the eigenvalues of A have negative real parts,
< 0. Moreover, the criterion for determining K
i.e., Re({A})
is the error between the observer and actual outputs being as
small as possible.
5) Observer-based Controller Design: In this section, we
will design a PID controller. The main reason is that it has a
simple structure, which is easily understood by engineers, and
works well under practical conditions.
The PID controller uses the following law:
t
i = KP (yd y) + KI

(yd y)d + KD

d(yd y)
. (26)
dt

It should be noted that y is the observed output. Fig. 12 shows

the schematic diagram of the overall model of a PID controller.
Unlike traditional PID control, this observer-based PID control
is based on the observer output which is designed based on the
system model. The PID parameters are tuned based on trial and
error method.

Assumption 1
(i) E(w) = 0, E(wwT ) = Qc
(ii)
E(v) = 0, E(vv T ) = Rc
(iii)
E(wv T ) = 0

(21)
(22)
(23)

where E(.) denotes the expectation and is the Dirac delta

function.
Because both output and state have noise contents, the
observed outputs are used to replace the actual outputs. The
following observer-type filter is designed

= AX
+ B(bi a) + K(y y)
X
.
(24)

y = CX
In (24), K is the steady-state gain of the filter. The estimated
error is given by
= AX
+ w Kv
X

(25)

III. P ERFORMANCE A NALYSIS OF S PHERICAL

A IR B EARING S YSTEM
The evaluation of the state of the SABS, in terms of whether
it fulfills the assumed requirements, is done on the basis of
the system control performance. Moreover, it is done on the
basis of observations of the position variable and characterizing
the process behavior. This is a real-time experimental study.
The block diagram of the whole closed-loop control system
is shown in Fig. 13. The setup of SABS is shown in Fig. 14.
This involves air bearing, a sensor, drivers, and a controller. The
controller is designed based on a dSPACE control development
card, which utilizes a Texas Instruments TMS320C31 32-b
floating point processor. The control algorithm is implemented
via MATLAB/Simulink block diagrams with real-time interface, compiled on a PC and which can be downloaded into the
dSPACE board.

TAN et al.: DEVELOPMENT OF A SPHERICAL AIR BEARING POSITIONING SYSTEM

3507

Fig. 16. Filter performance analysis (I).

Fig. 14. Setup of SABS.

Fig. 17. Filter performance analysis (II).

the estimated parameter b converges the fastest. This is due

to the lack of knowledge about the plant. Through adaptive
learning, the parameters converge to their true values after
230 s. Finally, in this experiment, the values of a and b are
identified approximately, but they are approaching the actual
values, since the system noise is almost white noise which adds
to the excitation of signal partially.
Noise Filter
Based on the model (18), the filter (24) is designed, i.e.,
 


+ 0 (0.0755i 0.0001) + K(y y)
= 0 1 X
X
1
0 0

y = [ 1 0 ] X.
(27)

Fig. 15. Model identification.

Model Identification
We first identify the model parameter of the air bearing motor
from the entire system. The range of the desired signal used
is chosen as 1 Hz. The proportional-derivative control gain
is chosen as KP = 0.1, KD = 0.001. In this experiment, the
initial values of the parameters in the adaptive laws (11) and
(12) are chosen as [b(0), a
(0)] = [0.08, 0.0001].
The adaptation rates in the laws are selected as 1 = 0.1
and 2 = 0.1. The time evolution of the parameter estimates
using the proposed adaptive algorithm is shown in Fig. 15.
The parameters b and a
converge to 0.0755 and 0.0001, respectively. It is observed that, during the initial learning phase,

The choice of the gain K depends on the condition.

 5  Initially,
we choose K = 24 . Then, we change to K = 25
to compare
the filter performance. Fig. 16 shows the filter performance
compared with the actual
  sine wave signal. It is found that,
for the case of K = 24 (0 22 s), the estimated value is not
5
satisfactory, while for the case of K = 25
(after 22 s), the
estimated value converges to the actual sine wave signal. To
show the convergence clearly, we plot the figure under a small
time window (3040 s), and the result is shown in Fig. 17.
It is seen from the figure that the noise is removed by the
designed filter and the estimated output approaches the actual
one closely. Now, this filter
 5 can represent the system output

well.
For
the
case
of
K
=
25 , we can obtain A = A KC =


5 1
, and its eigenvalues are 2.5 4.33i. The real
25 0
parts of all the eigenvalues of A are negative, which satisfies
< 0.
the stability condition: Re({A})
Control Results
The observer-based PID controller in (26) is applied to
the SABS. The parameters of the PID controller are adjusted
according to the general tuning rules for PID parameters. It

3508

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012

IV. C ONCLUSION
In this paper, a spherical air bearing positioning system
(SABS) has been developed. This system combines VCAs
with pneumatic bearing in order to achieve higher accuracy
and better performance than the traditional multi-DOF system.
The mechanical structure is carefully designed. The design
and the construction of the control system for the SABS is
presented, which involves system modeling, parameter identification, eliminating noise, and controller design. The model of
the SABS is identified based on the adaptive control concepts
first. Then, a noise filter is designed to remove the measurement
white noise on the basis of the model. Finally, an observerbased PID controller is applied to the SABS and implemented
on a dSPACE development card.
The experimental results indicate that the observer-based
PID controller is effective in controlling the SABS. Comparing
it with the traditional PID controller, it is clear that the observerbased PID controller achieves higher precision and better tracking performance of about ten times than the traditional one.
R EFERENCES
Fig. 18. Observer-based PID control.

Fig. 19. Traditional PID control.

is obtained that KP = 0.01, KI = 0.12, and KD = 0.008.

Fig. 18 shows the control result, and the maximum position
error is about 0.0125 . If using the traditional PID controller
without an observer, the maximum error is about 0.25 , and the
result is shown in Fig. 19. It is clearly observed from the figures
that the tracking error is significantly improved by the proposed
control.

[1] K. K. Tan and S. Huang, Problem and solution of designing an air

bearing system, in Proc. 2nd Int. Conf. Ind. Inf. Syst., Dalian, China,
2010, pp. 212215.
[2] K. M. Lee, G. Vachtsevanos, and C. K. Kwan, Development of a spherical stepper wrist motor, in Proc. IEEE Int. Conf. Robot. Autom., 1988,
vol. 1, no. 3, pp. 225242.
[3] K. M. Lee and C. K. Kwan, Design concept development of a spherical
stepper for robotics applications, IEEE Trans. Robot. Autom., vol. 7,
no. 1, pp. 175181, Feb. 1991.
[4] D. E. Ezenekwe and K. M. Lee, Design of air bearing system for fine motion application of multi-DOF spherical actuators, in Proc. IEEE/ASME
Int. Conf. Adv. Intell. Mechatronics, Atlanta, GA, 1999, pp. 812818.
[5] M. Hu, H. Du, S. F. Ling, and J. K. Teo, A piezoelectric spherical motor
with two degree-of-feedom, Sens. Actuators A, Phys., vol. 94, no. 1/2,
pp. 113116, Oct. 2001.
[6] K. M. Lee, H. Son, and J. Joni, Concept development and design of a
spherical wheel motor (SWM), in Proc. IEEE ICRA, Barcelona, Spain,
Apr. 1822, 2005, pp. 36523657.
[7] C. K. Lim, I. M. Chen, L. Yan, and G. Yang, Motion generation methodology of a permanent magnet spherical actuator, in Proc. IEEE/ASME
Int. Conf. Adv. Intell. Mechatronics, Singapore, 2009, pp. 13771382.
[8] C. Xia, C. Guo, and T. Shi, A neural-network-identifier and fuzzycontroller-based algorithm for dynamic decoupling control of permanentmagnet spherical motor, IEEE Trans. Ind. Electron., vol. 57, no. 8,
pp. 28682878, Aug. 2010.
[9] K. C. Fan, C. C. Ho, and J. I. Mou, Development of a multiple-microhole
aerostatic air bearing system, J. Micromech. Microeng., vol. 12, no. 5,
pp. 636643, Sep. 2002.
[10] S. Q. Lee and D. G. Gweon, A new 3-DOF Z-tilts micropositioning
system using electromagnetic actuators and air bearings, Precis. Eng.,
vol. 24, no. 1, pp. 2431, Jan. 2000.
[11] D. J. Lee, K. Kim, K. N. Lee, H. G. Choi, N. C. Park, Y. P. Park, and
M. G. Lee, Robust design of a novel three-axis fine stage for precision
positioning in lithography, Proc. IMechE, Part C: J. Mech. Eng. Sci.,
vol. 224, no. 4, pp. 877888, 2010.
[12] Q. Zhang, X. Shan, G. Guo, and S. Wong, Performance analysis of air
bearing in a micro system, Mater. Sci. Eng.: A, vol. 423, no. 1/2, pp. 225
229, May 2006.
[13] J. L. Schwartz and C. D. Hall, System identification of a spherical airbearing spacecraft simulator, in Proc. AAS/AIAA Space Flight Mech.
Meeting, Maui, HI, 2004.
[14] K. K. Tan, Development of a new and precise spherical air bearing
alignment system, NRF Proof-of-Concept Grant Scheme, Singapore,
2009.
[15] H. Son and K. M. Lee, Open-loop controller design and dynamic characteristics of a spherical wheel motor, IEEE Trans. Ind. Electron., vol. 57,
no. 10, pp. 34753482, Oct. 2010.

TAN et al.: DEVELOPMENT OF A SPHERICAL AIR BEARING POSITIONING SYSTEM

[16] L. Yan, I. M. Chen, C. K. Lim, G. Yang, W. Lin, and K. M. Lee, Design

and analysis of a permanent magnet spherical actuator, IEEE/ASME
Trans. Mechatronics, vol. 13, no. 2, pp. 239248, Apr. 2008.
[17] S. Ikeshita, A. Gofuku, T. Kamegawa, and T. Nagai, Development of
a spherical motor driven by electro-magnets, J. Mech. Sci. Technol.,
vol. 24, no. 1, pp. 4346, Jan. 2010.
[18] Z. Qian, Q. Wang, L. Ju, A. Wang, and J. Liu, Torque modeling and
control algorithm of a permanent magnetic spherical motor, in Proc. 12th
Int. Conf. Elect. Mach. Syst., Tokyo, Japan, 2009, pp. 16.
[19] C. Schmidt, J. Heinzl, and G. Brandenburg, Control approaches for highprecision machine tools with air bearings, IEEE Trans. Ind. Electron.,
vol. 46, no. 5, pp. 979989, Oct. 1999.
[20] L. Yan, I. M. Chen, C. K. Lim, G. Yang, and K. M. Lee, Empirical formulation of torque output for spherical actuators with low-cost
rotor poles, in Proc. IEEE/ASME Int. Conf. Adv. Intell. Mechatronics,
Singapore, 2009, pp. 16251630.
[21] C. K. Lim, I. M. Chen, L. Yan, G. Yang, and K. M. Lee, Electromechanical modeling of a permanent-magnet spherical actuator based on
magnetic-dipole-moment principle, IEEE Trans. Ind. Electron., vol. 56,
no. 5, pp. 16401647, May 2009.
[22] Z. Zhou and K. M. Lee, Real-time motion control of a multi-degree-offreedom variable reluctance spherical motor, in Proc. IEEE Int. Conf.
Robot. Autom., Minneapolis, MN, 1996, pp. 28592864.
[23] K. S. Chen, D. L. Trumper, and S. T. Smith, Design and control for
an electromagnetically driven X Y stage, Precis. Eng., vol. 26,
no. 4, pp. 355369, Oct. 2002.
[24] M. Hu, S. F. Ling, H. Du, and J. K. Teo, Design of a novel ultrasonic
spherical motor, in Proc. IEEE Ultrason. Symp., San Juan, PR, 2000,
pp. 667670.
[25] D. H. Yeom, N. J. Park, and S. Y. Jung, Digital controller of novel
voice coil motor actuator for optical image stabilizer, in Proc. Int. Conf.
Control, Autom. Syst., Seoul, Korea, 2007, pp. 22012206.

Kok Kiong Tan (M04) received the B.Eng. degree

in electrical engineering with honors and the Ph.D.
degree from the National University of Singapore,
Singapore, in 1992 and 1995, respectively.
He is currently an Associate Professor with the
Department of Electrical and Computer Engineering, National University of Singapore. His current
research interests are in the areas of advanced control
and autotuning, precision instrumentation and control, and general industrial automation.

Sunan Huang received the Ph.D. degree from

Shanghai Jiao Tong University, Shanghai, China,
1994.
Since 1997, he has been a Research Fellow
with the Department of Electrical and Computer
Engineering, National University of Singapore,
Singapore. His research interests include error compensation of high-precision machine, adaptive control, neural network control, and automated vehicle
control.

3509

Wenyu Liang received the B.Eng. and M.Eng. degrees in mechanical engineering from China Agricultural University, Beijing, China, in 2008 and
2010, respectively. Since 2010, he has been working
toward the Ph.D. degree in the Department of Electrical and Computer Engineering, National University
of Singapore, Singapore.
His research interests include mechatronics and
precision motion control.

Abdullah Al Mamun (M93SM04) received the

B.Tech. (Hons.) degree in electronics and electrical
communication engineering from the Indian Institute
of Technology, Kharagpur, India, in 1985, and the
Ph.D. degree in electrical engineering from the National University of Singapore, Singapore, in 1997.
He is currently an Associate Professor with the
National University of Singapore. His research interests include servomechanism in data storage systems, precision mechatronics, and autonomous mobile robots.

Eng Kiat Koh received the M.S. degree in electrical

engineering and the Ph.D. degree for his work on
multivariable control from the National University
of Singapore, Singapore, in 1990 and 1997, respectively.
He is currently a Research Scientist with the
Data Storage Institute, a research institute under the
umbrella of the Agency for Science, Technology
and Research (A*STAR), Singapore. His current research interests include proprietary file systems design and systems architecture for solid-state storage.

Huixing Zhou received the B.Eng. degree from

Dalian University of Technology, Dalian, China, and
the Ph.D. degree in mechanical engineering from
the Tsinghua University, Beijing, China, in 1983 and
1998, respectively.
From 1998 to 2000, he was a Postdoctoral Research Fellow with the Department of Electrical
and Computer Engineering, National University of
Singapore, Singapore. Since 2005, he has been a
Professor with the College of Engineering, China
Agricultural University, Beijing, China. His research
interests include linear motors, direct drive technology, DSP embedded system,
and precision motion control.
Dr. Zhou is a member of Institution of Mechanical Engineers, a Chartered
Engineer, and Corresponding Member of IMechE in China.