Wireless Pers Commun (2014) 76:643–656

DOI 10.1007/s11277-014-1731-1

Fast Steering Mirror Control Using Embedded

Self-Learning Fuzzy Controller for Free Space Optical
Communication

Bilal Ahmad Alvi · Muhammad Asif · Fahad Ahmed Siddiqui ·

Muhammad Safwan · Jawad Ali Bhatti

Published online: 29 March 2014

© Springer Science+Business Media New York 2014

Abstract Coarse and Fine tracking in free space laser communication is very crucial. This
paper presents architecture for laser beam acquisition, tracking and pointing mechanism. The
centroid of the received image beam is calculated and then error is computed using reference
position. Embedded self-learning fuzzy controller (SLFC) is used to drive the fast steering
mirror mechanism to point the laser beam on receiver detector. The SLFC is embedded in
hardware using Arduino development board. The performance of the SLFC is compared
with the standard PID control. It is shown that the proposed architecture successfully track
the laser beam and the SLFC shows the robustness against model uncertainties and reject
disturbances.

Keywords Free space optics · Neuro-fuzzy control · Image processing · Fast steering
mirror control

1 Introduction

In recent years, the dependence of our social life has been concentrated upon devices and
applications heavily reliant on access to Internet. These devices range from desktop PC,

B. A. Alvi (B) · M. Asif · F. A. Siddiqui · M. Safwan · J. A. Bhatti

Electronic Engineering Department, Sir Syed University of Engineering and Technology,
Karachi, Pakistan
e-mail: [email protected]
M. Asif
e-mail: [email protected]
F. A. Siddiqui
e-mail: [email protected]
M. Safwan
e-mail: [email protected]
J. A. Bhatti
e-mail: [email protected]

123
644 B. A. Alvi et al.

laptop to hand-held smart phones. Day by day applications are becoming bandwidth hungry.
Especially the smart phone era is proving to be the era of Internet dependence as almost
all the applications these days require some sort of Internet connectivity in order to provide
productivity. Most of the users are addicted to the habit of staying connected and frequently
updating their activities on social networks [8].
In order to satisfy user’s demands of broad band access, various devices of communication
network needs to be linked to develop a ubiquitous network. To achieve universal platform
connectivity between wireless access system and optical fiber core network has been around
and commonly known as radio over fiber (RoF) [4]. This platform is widely used in cellular
phone and other broadband networks. This concept can further be utilized for free space
channels by applying free space optical communication techniques combined with RoF.
These systems make a use of traditional optical fiber network technology and devices such
as Erbium doped fiber amplifiers (EDFAs), single mode fiber (SMF) and DWDM technique.
All free space optic (FSO) systems operate at wavelength 1,550 nm and single mode fiber
from the termination is directly connected [5]. The optical beam is radiated directly from
connected fiber end to the atmosphere using FSO optics. The receiver accepts the transmitted
optical beam and can focus it directly into the core of fiber. Radio over free space optics
(RoFSO) system may transmit light signals and it does not require any conversion from
optical to electrical or vice versa at fiber connection part. This greatly enhances the system
throughput.
This system can replace fiber from RoF system, to transmit optical signals that are confined
various kinds of wireless service signals without modulation through free-space. It also can
be an attractive means for RF signal transmission and may find significant applications in
areas such as Metro network extension, Last mile access, Fiber Back Haul, Fiber backup and
Interconnectivity of distributed antenna systems.
In order to achieve seamless connection with an optical fiber network avoiding any changes
in the transmission signal formats, all-optical connection technology which combines a free-
space optical propagated beam at 1,550 nm and single mode fiber (SMF) is essential [13].
Signal transmission in the free space encounters a lot of noise and distortion due to variety
or reasons including scintillation noise, atmospheric pressure and ambient temperature [15].
Even slight variation in these parameters can significantly distort the signal being sent over
free space.
However, there are many challenges to directly connect the free-space propagated beam
directly to the SMF core. Some of the challenges related to radio over free space optic
include incorporation of high speed and highly precise tracking mechanism [14]. In addition
to effective beam tracking system, an efficient method (active tracking) for focusing the light
into the SMF at the receiver is required to maintain alignment of the received optical signal
to the SMF [6,9]. Further, it is required to develop tracking algorithm to rectify misalignment
issues [7]. The objective of this research is to develop a model of free space optical (FSO)
communication system incorporating a self-learning fuzzy tracking system to improve the
accuracy of the atmospheric fluctuation suppression.
Traditional adaptive controller design technique uses high gain mechanism in feedback
loop to compensate uncertainties and disturbances whereas self-learning fuzzy controller
uses parallel and distributed learning mechanism to approximate the nonlinear function either
using fuzzy logic system or neural network [1]. The main advantage of fuzzy controller is
that it doesn’t require any mathematical model and linguistic information can be directly
incorporated into the controller. So, fuzzy controller has attracted numerous attention and a
number of fuzzy based controllers have been proposed using fuzzy–fuzzy [10], neuro-fuzzy
[11], and wavelet-fuzzy [3]. The main drawback of these techniques is that they are com-

123
Fast Steering Mirror Control 645

plex and requires a lot of computational power. This implies that it is difficult to implement
these techniques on off-the-shelf microprocessor or microcontroller based system. The main
motivation of this work is to implement a computationally effective self-learning fuzzy con-
troller which provide robustness against the model uncertainties and disturbances and can be
implemented using off-the-shelf controller boards.
This paper proposed a fast steering mirror control for free space optical communication
using self-learning fuzzy control. The system use two axis fast steering mirror (FSM) and
image base sensor to localize the optical beam. The error calculated based on image sensor is
then use to actuate the fine pointing mechanism for beam alignment using self-learning fuzzy
logic. The advantage of using the self-learning fuzzy control algorithm is that it provides the
robustness in the system by adapting to the inverse dynamics of the system.
Fuzzy controller is based on rule base created by using the expert opinion and considered
less dependent on actual system model. It can deal with variations in the system model with the
real world system. The self-learning fuzzy controller is based on fuzzy relational model that
is trained online in order to estimate the inverse dynamics of the plant. The main contributions
of this work are listed as: (1) Proposed an architecture for laser beam acquisition, tracking and
pointing mechanism, (2) beam image profile is detected using image processing algorithm
(3) a self-learning fuzzy controller is designed and implemented on Arduino mega 2,560.
The breakup of the further work is structured as follows: Sect. 2 discusses the architecture
of the tracking system. Sect. 3 presents the beam image profile and centroid detection. Sect. 4
discusses the self-learning fuzzy tracking system. The simulation results and discussions are
presented in Sect. 5. Finally Sect. 6 provides the conclusion.

2 Architecture of Tracking System

The block diagram of the complete tracking system is shown in Fig. 1. It includes fast
steering mirror (FSM), CCD sensor, beam splitter, data receiver and fuzzy logic controller.
Data receiver and CCD sensor are connected on a rigid platform and aligned to receive the
laser beam. The FSM is controlled using two DC motor actuators which is use to align a
beam on data receiver. The CCD sensor array detects the position of beam in an image profile

Fig. 1 System model of FSO system

123
646 B. A. Alvi et al.

using centroid detection. The error is calculated with respect to the reference position and
fed into the fuzzy controller. Fuzzy controller calculates the actuator outputs to control the
FSM dc motors.

3 Beam Image Profile and Centroid Detection

The initial and most important task in FSO communication is to track the error between the
current position and the reference point. In order to track this error a laser beam is used
between both nodes which falls on a CCD sensor array. One node transmits the laser beam,
while on the other node, the received laser beam is projected on the CCD sensor array which
transform it into the digital image profile. The beam image profile is represented by using
following function
I (x, y, z) ∈ (0, 1) (1)
where x = 0, 1, . . . , M, y = 0, 1, . . . , N and z = 0, 1, 2 are the dimensions of RGB image
and I (x, y, z)represents the pixel intensity. Figure 2 shows the images captured using CCD
camera to detect the position of laser beam. It can be seen that images are noisy and it is
important to do filtering before detecting the position. Therefore, the images are breakdown
into red, green and blue planes and then converted into the binary images using integral
image thresholding [2]. The resulting binary images of red, green and blue planes are shown
in Figs. 3, 4 and 5 respectively.

Fig. 2 Images captured using CCD camera to detect the position of laser beam

123
Fast Steering Mirror Control 647

Fig. 3 Binary image using red image beam profile

 
I Binar y (x, y) = I B R (x, y) I BG (x, y) I B B (x, y) (2)

where I Binar y (x, y)is the filtered binary image, I B R (x, y) is the binary image obtained using
red plane only, I BG (x, y)is the binary image obtained using  green plane only, I B B (x, y)is
the binary image obtained using blue plane only and is the logical end operator. The
filtered images obtained using Eq. 2 is shown in Fig. 6. The centroid of the final images is
calculated using following function
 
μ1,0 μ0,1
centr oid = (3)
μ0,0 μ0,0
where μ0,0 represents the 0th moments and can be calculated using

M 
N
μ0,0 = I Binar y (x, y) (4)
x=0 y=0

and μi,j is the 1st moments, i, j ∈ [0, 1] and i  = j. The 1st moments can be obtained using
M N
x=0 y=0 x I Binar y (x, y)
μ1,0 = (5)
μ0,0
M N
x=0 y=0 y I Binar y (x, y)
μ0,1 = (6)
μ0,0

123
648 B. A. Alvi et al.

Fig. 4 Binary image using green image beam profile

Figure 7 shows the typical beam profile images with centroid which represents
xandposition of the received laser beam on CCD sensor array.

4 Self-Learning Fuzzy Tracking Control

The centroid of laser beam detected using centroid detection is used to calculate the error
signal which has to minimize using the self-learning fuzzy controller. The output of the
designed controller has to control the motors which derive the FSM in order to adjust the
azimuth and elevation angels of the mirror. This is mandatory for the freespace optical
communication, as without the proper alignment, two nodes cannot communicate properly
and the latency will increase sufficiently. The generalized model of proposed FSM mechanism
is represented as

ẋ (t) = f (x, t) +  f (x, t) + δ (t) + u(t) (7)

where x = [x1 , x2 , . . . xn ]T ∈  is the state vector and u (t) ∈  is the control input of
system, δ f (x, t) represents system uncertainties and δ (t) is the bounded disturbance, i.e.,

δ (t) ≤ 

123
Fast Steering Mirror Control 649

Fig. 5 Binary image using blue image beam profile

where  is the small positive constant. The control objective is the tracking the received beam
profile with respect to the reference such that

lim e(t) = lim xd (t) − x(t) → 0 (8)

t→∞ t→∞

where e(t) represents the tracking error, xd (t) is the target position, x (t) is the current
position and  ·  is the Euclidean norm of a vector.
The control objective is achieved by using self-learning fuzzy logic control (SLFC) track-
ing system [12]. SLFC does not require a system model and accommodate nonlinearities and
input/output disturbances. SLFC uses an adaptive feed-forward fuzzy controller with online
learn mechanism to adapt the inverse plant dynamics and to compensate disturbances. A
PID control is used to provide the close-loop stability during online training of feed-forward
controller. In addition, it compensate for model mismatches as an exact inverse mapping is
difficult. The block diagram of SLFC is shown in Fig. 8 with desired inputs and outputs.
The reference model is used to filter the desired changes in plant output in order to make the
setpoint trajectory achievable. The form of control law for FSM is as follows
de(t)
u (t) = u f (t) + k p e (t) + ki ∫ e (t) dt + kd (9)
dt
A fuzzy relational model (FRM) is used to calculate the feed-forward control output u f (t).
A FRM is a predefined set of linguistic rules, each of which has an associated rule confidence
which is stored in a fuzzy relational array(R). The fuzzy relational array is define between

123
650 B. A. Alvi et al.

Fig. 6 Filtered images of laser beam profile

the range 0–1,which indicates that confidence that associated fuzzy rule correctly describes
the fuzzy relation between inputs and the outputs of system being modeled. Assuming that
the fuzzy set have 50 % overlap, the rule base of FRM can be described as
IFx(t) is Ai ANDũ f (t − 1)is B j THENy(t)is X K (R[i, j, k])
and
ũ f (t) = u f (t − 1) + γ e(t) (10)
where γ is the learning rate. The on-line learning mechanism is implemented using the
modified RSK scheme [12]. RSK is one of efficient and computationally simple schemes
for estimating the rule confidences in the presence of sensor noise and disturbances. The
recursive form of RSK can be defined as

R1 (t) i f f A1,...,An (x (t))  = 0
R̂ A1,...,An (t) = (11)
R2 (t) other wise
where
R1 (t) = f A1,...,An (x (t)) µ B j (ũ D (t)) + λ R̂ A1,...,An (t − 1) F A1,...,An (t − 1) ,

R2 (t) = R̂ A1,...,An (t − 1)

f A1,··· ,An (x (t)) = μ A1 (x1 ) .μ A2 (x2 ) . . . . .μ An (xn )

123
Fast Steering Mirror Control 651

Fig. 7 Filtered binary images with centroid detection

Fig. 8 Block diagram of

self-learning fuzzy control

and λ is the forgetting factor. The recursive F-array function Fis defined as

f A1,...,An (x (t)) + λF A1,...,An (t − 1) i f f A1,··· ,An (x (t))  = 0
F A1,...,An (t) = (12)
F A1,··· ,An (t − 1) other wise
The elements of the F-array are indicators of how strongly and how often each combination
of the inputs has occurred in the training data. Using the height defuzzification, the output
of the feed-forward controller, u f (t) is given by
q1 qn

i 1 =1 · · · i n =1 f A1,...,An R̂ A1,...,An,B1 U1 + R̂ A1,...,An,B2 U2

u f (t) = q1 qn
(13)
i 1 =1 · · · i n =1 f A1,...,An R̂ A1,...,An,B1 + R̂ A1,...,An,B2

where U j , j ∈ [1, 2] is the position of the apex of the output set.

123
652 B. A. Alvi et al.

5 Result and Discussion

The objective of aligning the FSM using the centroid detection method is implemented using
the SLFC which is implemented using an Arduino Mega board. X-vision CCD camera is
used to capture the laser beam images and then position of the laser beam on image plane is
detected using image processing algorithm on MATLAB. The self-learning fuzzy control is
implemented in Simulink and then embedded into Arduino mega 2,560 using target hardware.
Arduino Mega is connected to physical hardware which drives the azimuth and elevation
control dc motor using PWM. The block diagram of the complete system is shown in Fig. 9.
The fuzzy relational model implemented with five by five triangular membership functions
to fuzzify the input with universe of discourse [−20 20] and apexes positioned at [−10 5
0 5 10]. The time constant of the reference model is set to 0.01 to ensure the FSM follow
the desired trajectory with physical constraints. The PID controller, which controls the plant
during the learning cycle of SLFC, is tuned using the Ziegler-Nichols method which suggest
the initial gains as [25 10 5]. Therefore, the on-line learning rate γ is set using γ = K /T1
to ensure that the system remains stable during the initial learning phase, where K is the
proportional gain and Ti is the integral time. The forgetting factor λ is set to 0.7 to optimize
the system memory. Figures 10 and 11 show the response of the self-learning controller for
azimuth and elevation control motors respectively. It can be observed that PID controller
output is due to the fact that the fuzzy relational feed-forward controller is not able to model
the entire dynamics. Overall, the self-learning controller is able to track the laser beam profile
which is essential for FSO communication. It is also observed that both x and y axis controls
have slight variation due to fact that both control motors are supposedly similar in make,

Fig. 9 Block diagram of FSM

physical setup

Fig. 10 Self-learning fuzzy position control in X direction

123
Fast Steering Mirror Control 653

Fig. 11 Self-learning fuzzy position control in Y direction

Fig. 12 Time history of Rule Confidence R1–R9

Fig. 13 Time history of Rule Confidence R10–R18

123
654 B. A. Alvi et al.

Fig. 14 Time history of Rule Confidence R19–R25

but all the physical systems are manufactured with non-linearity. This means the output of
both motors varies slightly with respect to the manufacturing non-uniformity. The plots of
rule confidence is shown in Figs. 12, 13 and 14. It can be observed that the rule confidence
continue to change whenever the associated rule is fire.

6 Conclusion

Coarse and fine tracking in free space laser communication is very crucial. In this paper,
acquisition, tracking and pointing system for free space laser communication is proposed
to rectify the misalignment issues in FSO. The system used CCD sensor array to detect the
laser beam in an image. Using centroid detection the laser beam located in an image. The
error is computed with respect to the reference laser beam image. Self-learning fuzzy logic
controller is used to derive the fast steering mirror mechanism to point the laser beam on
receiver detector for coarse and fine tacking. The fast steering mirror mechanism is derived
using two dc motors. The adaptive fuzzy system is implemented in Arduino mega 2,560.
The results shows that the proposed methodology is stabilize using fuzzy logic controller.
PID controller, which is tuned using Ziegler–Nichols method, controls the plant while fuzzy
controller is learning the plant behavior. The output of fuzzy controller gradually increases
and start replacing the signal from the PID controller, and eventually taking over the control
completely.

References

1. Bououden, S., Filali, S., & Kemih, K. (2010). Adaptive fuzzy tracking control for unknown nonlinear
systems. International Journal of Innovative Computing, Information and Control, 6(2), 541–549.
2. Bradley, D., & Roth, G. (2007). Adaptive thresholding using the integral image. Journal of Graphics,
GPU, and Game Tools, 12(2).
3. Erdinc, O., Vural, B., Uzunoglu, M., & Ates, Y. (2009). Modeling and analysis of an fc/uc hybrid vehicular
power system using a wavelet-fuzzy logic based load sharing and control algorithm. International Journal
of Hydrogen Energy, 34(12), 5223–5233.
4. Hamed Al-Raweshidy, S.K. (2002). Radio over fiber technologies for mobile communications networks,
Artech House.
5. Henniger, H., & Wilfert, O. (2010). An introduction to free-space optical communication, RadioEngi-
neering, 19(2).

123
Fast Steering Mirror Control 655

6. Huang, Y., Ful, C., & Bao, Q. (2007). Control method for high-accuracy fine steering mirror basedon
slow sampling rate of tracking sensor signal. Information, Decision and Control.
7. Kuang, J., Tang, T., Fu, C., Ding, K., & Yu, W. (2009). Simulation of the fast steering mirrorcontrol
system based on gyro velocity feedback. In 2009 international conference on optical instruments and
technology: Advanced sensor technologies and applications.
8. Leung, L. (2004). Net-generation attributes and seductive properties of the internet as predictors of online
activities and internet addiction. Cyber-Psychology and Behavior, 7(3), 333–348.
9. Min, Z., & Yanbing, L. Compound tracking in ATP system for free space optical communication, In
Mechatronic Science, Electric Engineering and Computer (MEC), 2011 International Conference on
19–22 Aug 2011, Jilin, China.
10. Roopaei, M., Zolghadri, M., & Meshksar, S. (2009). Enhanced adaptive fuzzy sliding mode controlfor
uncertain nonlinear systems. Communications in Nonlinear Science and Numerical Simulation, 14(9),
3670–3681.
11. Siddique, N. (2014). Neuro-fuzzy control. In Intelligent Control, Springer, pp. 179–216.
12. Tan, W., & Dexter, A. L. (2000). A self-learning fuzzy controller for embedded applications. Automatica,
36(8), 1189–1198.
13. Wakamori, K., Kazaura, K., & Matsumoto, M. (2009). Research and development of a next generation
free-space optical communication system. SPIE OPTO: Integrated optoelectronic devices (p. 723,404).
Bellingham: International Society for Optics and Photonics.
14. Woo, Y., & Kim, Y. C. (2011). A controller design for precision laser-beam positioning system. Interna-
tional Journal of Optomechatronics, 5(2).
15. Zhu, X., & Kahn, J.M. (2002). Free-space optical communication through atmospheric turbulencechan-
nels. IEEE Transactions on Communications, 50(8).

Bilal Ahmad Alvi is serving Sir Syed University of Engineering &

Technology since May 2007 as Chairman, Department of Electronic
Engineering. He has more than 25 years of experience in the field of
Electronic Engineering and Optical Fiber Communication. He did his
M.S. and Ph.D. from the University of Salford, U.K. in the field of
Optical Fiber Communication and B.E. in Electronic Engineering from
NED University Karachi, Pakistan. He has worked on many important
projects in Pakistan, Saudi Arabia and the U.S.A. He had also served
at the Pakistan Space and Upper Atmosphere Research Commission
(SUPARCO) as Senior Engineer for more than 15 years. Dr. Alvi has
published a number of research papers and chaired many local and
International Conferences.

Muhammad Asif received B.S. degree in Biomedical engineering

from the Sir Syed University of Engineering and Technology, Karachi,
Pakistan in 2003, and M.S. degree in Electrical and Electronic Engi-
neering from the UniversitiSains Malaysia (USM) Malaysia, in 2007.
Currently, he is pursuing his Ph.D. in Control Engineering from NUST,
Pakistan. He is also working as faculty member and researcher in the
Department of Electronic Engineering, Sir Syed University of Engi-
neering and Technology. His research addresses the issues and prob-
lems of industrial automation, navigation, mapping and design and
implementation of statistical control algorithm for autonomous robots.

123
656 B. A. Alvi et al.

Fahad Ahmed Siddiqui is working as Assistant Professor in Elec-

tronic Engineering Department at Sir Syed University of Engineer-
ing and Technology. He is currently pursuing his Ph.D. in the field
of control engineering. He did his M.S. in Electrical Engineering with
Power Electronics from University of Bradford and B.E. in Electronics
Engineering from Usman Institute of Technology. His research interest
includes power systems and control system design.

Muhammad Safwan received B.S. degree in Electronic Engineering

from Sir Syed University of Engineering and Technology, Karachi,
Pakistan in 2012 and doing M.S. degree in Electrical Engineering from
Graduate School of Engineering Sciences and Information Technology,
Hamdard University, Karachi, Pakistan. He is working as junior Lec-
turer in Sir Syed University of Engineering and Technology, Karachi.
He has a keen interest in the field of control engineering and adaptive
signal processing.

Jawad Ali Bhatti graduated from Sir Syed University of Engineer-

ing and Technology in December 2002, Karachi, Pakistan. He com-
pleted his M.S. from Hamdard University, Karachi, Pakistan in Sep-
tember 2007. He is doing Ph.D. from Hamdard University, Karachi,
Pakistan. Presently, he is working as Assistant Professor in Sir Syed
University of Engineering and Technology, Karachi, Pakistan. He has
research interest in image processing and object tracking.

123