Content uploaded by Sabyasachi Chatterjee
Author content
All content in this area was uploaded by Sabyasachi Chatterjee on Jan 15, 2020
Content may be subject to copyright.
Enhancing Accuracy of Localization for Primary Users in
Cognitive Radio Networks
S CHATTERJEE P BANERJEE M NASIPURI
ECE Department ECE Department CSE Department
Heritage Institute of Technology Heritage Institute of Technology Jadavpur University
KOLKATA, INDIA KOLKATA, INDIA KOLKATA, INDIA
sabyasachi.chatteree@heritageit.edu prabir.banerjee@heritageit.edu mnasipuri@cse.jdvu.ac.in
Abstract—Cognitive Radio (CR) technology constitutes a new
paradigm where wireless devices can access the spectrum left
unused by licensed or primary users in an opportunistic way.
At the same time, these unlicensed secondary users (SUs) must
reliably detect the presence of licensed primary transmission to
avoid interfering with them. Information about primary users
(PUs) locations are very important in Cognitive Radio
Networks (CRNs) to improve spectrum hole sensing, intelligent
location-aware routing as well as efficient and accurate
detection. In this paper, we investigate localization techniques
based on Received Signal Strength Indicator (RSSI)
measurements to estimate the position of PUs. Compared to
other proposed non-cooperative localization algorithms, the
weighted centroid localization (WCL) scheme uses only the
received signal strength information and this approach has
been proposed here. In this paper, we present the theoretical
framework of WCL algorithm. Then to improve localization
accuracy, classical propagation model has been introduced
with correction factor.
Keywords-Cognitive Radio Network, Classical Propagation
Model, Localization Accuracy, Non-Cooperative Localization,
Primary Users, Received Signal Strength Indicator, Secondary
Users, Weighted Centroid Localization.
I. INTRODUCTION
The Radio Frequency (RF) spectrum popularly known as
electromagnetic spectrum is a natural resource for wireless
communication and is divided into multiple bands. The
radio frequency band has been regulated by Federal
Communication Commission (FCC) through the process of
spectrum allocation in which the particular frequency band
is restricted to the licensed user. As the entire spectrum
bands are already allocated to different services, a
fundamental problem of spectrum scarcity arises during
high demand for spectrum resource. However, the real
situation with these allocated bands is that often these are
underutilized with large spectrum holes at different
geographic location over a certain period of time [1].
Cognitive Radio (CR) is a new way to mitigate the spectrum
shortage problem by allowing smart and dynamic spectrum
management in future wireless communication network [2].
It is a radio or system that senses its operational
electromagnetic environment and can dynamically or
automatically adjust its radio operating parameters to
improve system performance.
In wireless network scenario, fixed frequency allocated
users are defined as the Primary Users (PUs) or licensed
users whereas users with dynamic or variable frequency
allocation is defined as the Secondary Users (SUs) or CR
users. In Cognitive Radio Networks (CRNs), Secondary
Users can opportunistically access the spectrum holes or
licensed spectrum by proper spectrum sensing mechanism,
as long as they does not cause excessive interference with
actual owner of that spectrum or PU. However PUs can
access the spectrum at any time, even during the SU
transmission. In that case, PUs may suffer from severe
interference until the end of transmission. So an essential
issue in successful operation of CRN is the determination of
the precise location of primary users so that SUs can locate
PUs and switch to free frequency upon detection.
To detect primary user in cognitive radio network, position
information or location awareness is an essential service
which increases context knowledge. Positioning refers to the
determination of the physical position of an object w.r.t
some coordinate system whereas localization may refer to
the determination of a logical or symbolic location [3].
Localization occurs in a few steps. First selected
measurements are performed between nodes and second,
these measurements are correlated to some earlier reference
measurement to determine position. The knowledge of
primary users' position w.r.t the secondary users is
important for the cognitive radio network to identify &
choose proper spectrum access opportunities at the same
time avoiding destructive interference. Localization
techniques can be classified based on measurements
between nodes, such as range based, angle based and
proximity based localization. Among them, range based
systems (i.e. based on distance estimates) are more suitable
for localization accuracy [3]. However, in the cognitive
radio context, PU localization poses a few unique
challenges. Here a primary user does not communicate
directly with secondary cognitive radio user during the
localization process. This scenario is referred to as non-
cooperative localization and does not allow the use of
2015 International Conference on Computational Intelligence & Networks
2375-5822/15 $31.00 © 2015 IEEE
DOI 10.1109/CINE.2015.24
72
2015 International Conference on Computational Intelligence & Networks
2375-5822/15 $31.00 © 2015 IEEE
DOI 10.1109/CINE.2015.24
74
conventional techniques such as time-of-arrival (ToA) and
Time-Delay-of-Arrival (TDoA) ranging [4].
In this paper, we have focused on analyzing the performance
of a non-cooperative localization algorithm, referred to as
Weighted Centroid Localization (WCL). Compared to other
proposed non-cooperative localization algorithms, WCL
scheme uses only the received signal strength information
(RSSI), which makes it simple to implement and robust to
variations in the propagation environment. When primary
users waveform is not known a prior, RSS measurements can
be an efficient tool where the received power from the
primary users’ is the only information to the secondary users
in CRN [3]. For accurate prediction of RSSI signal, a
correction factor, caused by presence of shadowing and
fading effects, have been considered. In our work, we have
observed various environmental conditions and have
provided design guide lines such as node placement and
spacing.
II. RELATED SCHEMES
Several techniques have been proposed in the literature for
estimating the accurate position of active primary users in a
cognitive radio network (CRN). Localization algorithms
assume the presence in the network of a limited number of
reference nodes, referred to as beacon or anchor nodes,
which know their own spatial coordinates and are used as
reference points for localizing the other nodes, hereafter
referred to as strayed nodes. Broadly speaking, the
localization algorithms can be divided in two wide
categories: range–based [5] and range–free [6]. Range free
positioning schemes or non cooperative schemes have
attracted a lot of interest. At present, there are three different
detection techniques which have been proposed. In [7], S.
Shobana et al. have worked on the matched filter detection
method that can achieve high processing gain by employing
coherent reception. In [8], energy detection scheme has been
applied when secondary users can not gather sufficient
information about the primary user signal such as the
modulation type, the pulse shape & so on. But here problem
arises due to variation in signal types which cannot be
differentiated by secondary users, thus making it vulnerable
to false detection. The cyclostationary feature presented in
[9], can differentiate among various modulated signal.
However, this scheme is computationally complex and
requires long observation time. On the other hand, range-
based techniques such as multi-lateration [5], offer better
detection capability but require proper cooperation between
the two terminals or a very precise knowledge of the path
loss model.
To detect PUs in CRNs, limit on the SUs burst
transmission time to increase the sensing time for achieving
higher detection probability have been proposed to be
increased. This approach causes spectrum utilization to
decrease significantly. Another approach is to restrict the
transmission power of SUs. However, this method might
limit the coverage and throughput of the CRNs. In RSSI-
based range free algorithms, distance between nodes can be
estimated by RSSI-mapping techniques [10]. This requires a
preliminary knowledge about a map of the radio signal
strength received in the area of interest. Comparing the
RSSI values received from the different nodes of cognitive
radio networks with the pre-built RSSI map, a node can
estimate its own position in the area. However there can be
ranging errors due to noisy RSSI value as because this
scheme does not consider different channel conditions or
path loss variations.
.
III. RANGE FREE LOCALIZATION OF PUS
In the absence of distance information, the easiest way to
localize the Primary Users (PUs) is by Centroid
Localization (CL). In CL, position information and activities
of all neighboring nodes (SUs) which are located within the
transmission range of the particular Pus, is maintained by all
the Secondary Users by certain routing protocol. Therefore,
each SU node knows its location coordinates and its
neighbors’ SUs locations coordinates and uses this
information to find out location coordinates of unknown
PUs. Among all CL based non cooperative localization
algorithms, the Weighted Centroid Localization (WCL)
scheme uses only the Received Signal Strength Indicator
(RSSI) which makes it simple, cost effective and robust to
variations in the propagation environment. In WCL, the key
idea is to perform a weighted sum of SU’s position where
the received powers from unknown PUs define weights.
However, nodes or secondary users that are closer to the PU
have a stronger impact on location estimation. In particular
WCL collects all known coordinates of SUs that detect the
presence of PU [3].
More Precisely, assuming that there is a PU at unknown
coordinates (X,Y) and that N number of SUs are placed at
known coordinates ( ) with n= 1,....N. Here has
been defined as subset of those SUs that detect the presence
of the PU within their transmission range [3].
Reference Node
Detector SU
SU
PU
RSSI (Power Received
From All the Nodes within Subset Range)
Figure 1: Location of the Units
7375
Figure 1 shows that there are five SUs within a transmission
range of a PU. The four SUs, as can be seen, are at different
distances from the detector SU. Hence the required power or
RSSI at the detector SU corresponding to the other five will
be different. All the five SUs including the detector from a
subset in which K, a variable representing the identity of
SU, will be within one to five. Since in WCL method, the
main idea is to find out a weighted sum of SUs position
where the weights are related to the received power or RSSI
from the PU. Therefore, SUs that are closer to PU have
larger impact on location estimation.
…….. [1]
Where
SU (RSSI)
(Difference of Maximum &
Minimum Power received among the nodes in ).
The equation show that closer is the PU to the detector SU,
the weighted RSSI will be higher.
A. RSSI Relation with Distance
In communication system, Received Signal Strength
Indicator (RSSI) is a measurement of the ratio of signal
power & noise power present in a received radio signal. It is
generic radio receiver technology metric which is usually
invisible to the users and measurement often is done in the
Intermediate Frequency (IF) stage before the IF amplifier.
The RSSI-based localization techniques arise from the fact
that there is a strong correlation between the distances of
wireless links and RSSI. Typically, RSSI is measured in
dbm, which is ten times the logarithm of the ratio of power
of received signal and a reference power [11].
………… [2]
This would mean that- [Assume
It is known that received power decreases with increment in
distance-whatever RF propagation model is applied. For
example, as per Friis equation, the received RF power is
inversely proportional to the square of the distance between
the transmitter and the receiver. In other words-
Power
Simplifying this relationship in terms of RSSI can be
expressed as
RSSI-based location algorithm uses the average of all
calculated location as position estimation of the unknown
nodes, also affected by environmental factors. To improve
accuracy of localization, RSSI uses theoretical or empirical
propagation models with combined effect of path loss and
shadowing to calculate distance.
B. Power Law Model
Most radio propagation models are derived using a
combination of analytical and empirical methods. Both
theoretical and measurement-based propagation models
indicate that average received signal power decreases
logarithmically with distance. Radio Transmission system
often takes place over irregular terrain. The terrain profile
may vary from a simple curved earth profile to a highly
mountains profile. Over time, a classical model, Log
Normal Distribution, have emerged which use noise
analysis techniques and make it possible to predict the
signal strength. By considering the presence of trees,
buildings and other obstacles, model describes the random
shadowing effects which occur over a large number of
measurement locations which have the same transmitter-
receiver separation, but have different of terrain conditions
in the propagation path. This phenomenon is referred to as
log-normal shadowing [12]. Most models do not consider
these environmental variations between two different
locations of transistor (Tx) and receiver (Rx). This leads to
variation in signal strength at particular receiving point with
same separation distance from transmitter for different
terrain conditions. This can be expressed mathematically,
In a cognitive Radio Network (CRN) Transmitter (PU) and
Receiver (SU) generally operate in Non Line-of-Sight
(NLOS) environment. Based on empirical data, a fairly
classical model has been developed for NLOS propagation.
This model predicts the mean path loss when
transmitter receiver separation r (in meters)
.
………. [4]
Furthermore Power received from Tx to Rx can be
expressed in terms of Received Signal Strength Indicator
(RSSI). Without considering the condition of the signal
gain, the received signal strength is equal to transmitted
power minus the signal propagation loss. RSSI ranging uses
the received signal strength theoretical or empirical model
of the propagation path loss.
r ………..[5]
Where
7476
n = Propagating environment Characteristics
C. Modified Power Law for Localization
The RSSI is measured by a cognitive receiver corresponding
to a PU at a distance ‘r’. The signal strength also becomes
known from that particular PU. There is no standardized
relationship of any physical parameter to the RSSI. But in
this paper, we have shown a linear relationship among
power level, RSSI and RF signal strength. Mathematically,
for RF Propagation models obeying Friis equation,
Power
Converting Power equation to a RSSI equation-
..[6]
Equation [6] therefore represents a direct relationship
between the received power and the measured distance. The
accuracy of the measured value of RF signal or RSSI is
affected by shadowing and fading effects. This correction
factor can be expressed by the product as shown in
above equations.
The relevance of factor can also be explained thus. Let
the value of factor be y dBm and the uncorrected RF
signal strength is x dBm. Therefore the corrected RF signal
strength corresponding to PU under localization will be
given by-
= (x
= (x
Various values of for different scenarios will help to
predict PU location precisely.
Here we have assumed value known for any particular
scenario and Path Loss Exponent (PLE) or propagation
environment characteristics ‘n’ varies among 1 to 5. The
standard deviation (SD) represented by the Greek letter
sigma, measures the amount of variation or dispersion
from the average. Here a zero-mean Gaussian random
variable, with standard deviation : G has been used, were
is often estimated by empirical measurements. Commonly
accepted values for are between 6 dB and 12 dB.
Measured values of itself seem to display Gaussian
distribution as well, in their variations from one area to
another, and depend on the radio frequency, the type of
environment (rural, suburban, or urban), base station and
subscriber station height. Some reference values are listed in
the table 1.
Frequency
(GHz)
Path loss
Exponent
(n)
Environmental
Terrain
0.9 2.8 9.6 Suburban
3.7 3.2 9.5 Some Denser
Urban
5.8 2.93 7.85 Residential
5.8 2.0 6.9 LOS
5.8 3.5 9.5 NLOS
Table 1: Variation of
Furthermore, ‘ ’ distance between detector SU and
known PU can be calculated accurately by-
… [7]
Assume n=1
= inv log
= m dBm
Meter
IV. SIMULATED RESULT
The radio channel places fundamental limitations on the
performance of wireless links. The paths between
transmitters and receivers are extremely random due to
various obstructions. The standard deviation with Gaussian
random variable occurs when wave interference occurs at
the receiver as because transmitted wave travelled via
different paths and variable losses such as-rainfall,
atmospheric refraction, ionosphere attenuation, vehicles or
person blocking, building losses. A simple method for
accounting standard deviation ( ) factor is set to a marginal
loss figure which represents maximum bound for the fading
losses. This factor is typically selected on a statistical basis,
with various types of distribution function applicable
dependent on the environment. The statistics depend
strongly on multiple reflections from various objects, since
the transmitted waves travel along different paths of varying
lengths. The interaction between these waves causes
multipath fading whether there is a strong Line-of-Sight
(LOS) or Non-Line-of-Sight (NLOS) communication
between Transmitter-Receiver.
To generate the performance curves, the simulator software
“RF Propagation Calculator, version 1.03, Serial Data Code
980619” has been taken help of in this paper. We have
assumed that the transmit power level is uniform for all
measurements. The antenna heights will remain unchanged
and reference distance is also kept constant for our exercise
to localize PUs. Based on different environmental, terrain
conditions and line-of-sight properties the deviation of
received power from log-normal distribution is seen to be
7577
very high. This behavior is reflected in our modified ‘power
law model’. When the correction factor represented by as
shown in ‘table 1’ is taken into account, the received power
patterns obtained from the simulator, changes appreciably
from the curve, which does not consider this .
0
5
10
15
20
25
dBm
Power Received from Unknown User (PU)
With
Correction
Factor
Without
Correction
Factor
Figure 2: Simulated RSSI Reception from PU
Figure 2 shows the power reception from a particular node
(PU) to the same detector or receiver node (SU) for different
terrain conditions. It has been observed that the correction
factor corrects the power reception level significantly. This
will help to localize that particular PU more accurately.
Figure 3: Simulated RSSI Reception from Reference Node
The reception of received power from reference user (SU)
has been plotted in figure 3. By keeping constant reference
distance for all the terrain conditions, simulation has been
performed here for both the cases. Though distance is
constant, the received power levels or RSSI are changing
randomly as because standard deviation factor was not
applied in WCL localization algorithm. By considering
variable factor for different terrain model, almost constant
reception of RSSI can be observed.
Figure 4: Simulated PU Position
In figure 4, it can be observed that localization of PU or
predicted position of PU changes with different propagation
scenarios for WCL positioning scheme. Information of PU
is not accurate when localization algorithm is applied
without correction factor. On the other hand, position of
primary user is almost constant for different environmental
conditions when localization algorithm has been applied
with correction factor.
V. CONCLUSION
The Cognitive Radio system requires a signal detection
technique that detects reliably the signal received from
the Primary Users (PUs). In this paper, we have proved
that RSSI measurement would be a good technique for
detection of PU signals. This paper presents a new RSSI
based localization algorithm which considers significant
changes on the received signals due to reflection,
multipath propagation and non-line-of-sight reception
for more accurate measurement. Simulation results show
the performance of WCL for localization of PU is less
efficient compared with fading factor correction scheme.
Here with correction factor, the RSSI value changes at
the same reception node for any particular scenario, at
any instant, while comparing with WCL scheme.
Therefore the localization of PUs will be definitely more
accurate than when the correction factor is not applied.
This will ensure more reliable network performance of
Cognitive Radio.
44
46
48
50
52
54
56
dBm
Power Received from Reference User
(SU)
With
Correction
Factor
Without
Correction
Factor
58
60
62
64
66
68
70
72
74
76
78
meter
Unknown User (PU) Localization
With
Correction
Factor
Without
Correction
Factor
7678
REFERENCES
[1] J. Unnikrishnan, V. V. Veeravalli “Cooperative Sensing for primary
detection in Cognitive Radio” IEEE, DoI
10.1109/JJTSp.2007.914880 , FEBRUARY 2008.
[2] Yonghong Zeng, Senior Member, IEEE & Y. C. Wang “Spectrum-
Sensing Algorithms for cognitive radio based on statistical
covariance” September-2008.
[3] Sithamparanathan, Kandeepan. “Cognitive Radio Techniques :
Spectrum Sensing, Interfernce Mitigation, and Localization” Artech
House Id-10818323, 2012.
[4] J. Wang, P. Urriza, Y. Han and D. Cabric “Weighted Centroid
Localization Algorithm: Theretical Analysis and Distributed
Implementation” arxiv:1011.2313v3, 9 May,2011.
[5] K. Langendoen and N. Reijers, “Distributed localization in wireless
sensor networks: A quantitative comparison,” Computer Networks,
vol. 43, no. 4, pp. 499–518, Nov. 2003
[6] T. He, C. Huang, B. M. Blum, J. A. Stankovic, and T. Abdelzaher,
“Range-free localization schemes for large scale sensor networks,” in
Proc. ACM/IEEE MobiCom, San Diego, CA, USA, Sep. 14–19,
2003.
[7] S.Shobana, R.Saravanan, R.Muthaiah, "Matched Filter Based
Spectrum Sensing on Cognitive Radio for OFDM WLANs" ISSN :
0975-4024, International Journal of Engineering and
Technology,Feb-Mar 2013
[8] V.I. Kostylev, “Energy detection of a signal with randomamplitude,”
in Proceedings of the IEEE International Conference on
Communications (ICC 2002), vol. 3, pp. 1606–1610, 28 April–2 May
2002.
[9] A. Fehske, J. Gaeddert, and J. H. Reed, “A new approach to signal
classification using spectral correlation and neural Networks,” in
Proceedings of the 1st IEEE International Symposium on New
Frontiers in Dynamic Spectrum Access Networks (DySPAN ’05), pp.
144–150, November 2005.
[10] T. He, C. Huang, B. Blum, J. Stankovic, and T. Abdelzaher,“Range-
free localization schemes in large scale sensor networks,” in
MobiCom, 2003. [Online].
Available:citeseer.ist.psu.edu/he03rangefree.html
[11] A.T Parameswaran, M.I Husain, S Upadhyaya "Is RSSI a Reliable
Parameter in Sensor Localization Algorithms – An Experimental
Study"U.S.Department of Defense Grant No. H98230-08-1-0333.
[12] T. s. Rappaport "Wireless Communication" Prentice Hall
Coomications Engineering and Emerging Technologies Series ISBN
0-13-042232-0, 2002.
7779
