TIME OF ARRIVAL BASED LOCALIZATION IN WIRELESS SENSOR NETWORKS: A LINEAR APPROACH

Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.4, August 2013
DOI : 10.5121/sipij.2013.4402 13
TIME OF ARRIVAL BASED LOCALIZATION IN
WIRELESS SENSOR NETWORKS: A LINEAR
APPROACH
Ravindra S1
and Jagadeesha S N2
1, 2
Jawaharlal Nehru National College of Engineering, Shimoga-577204.
Visvesvaraya Technological University,
Belguam, Karnataka,
India.
rsp.cse@gmail.com
jagadeesha_2003@yahoo.co.in
ABSTRACT
In this paper, we aim to determine the location information of a node deployed in Wireless Sensor
Networks (WSN). We estimate the position of an unknown source node using localization based on linear
approach on a single simulation platform. The Cramer Rao Lower Bound (CRLB) for position estimate is
derived first and the four linear approaches namely Linear Least Squares (LLS), Subspace Approach (SA),
Weighted Linear Least Squares (WLLS) and Two-step WLS have been derived and presented. Based on the
simulation study the results have been compared. The simulation results show that the Two- step WLS
approach is having higher localization accuracy.
KEYWORDS
Source Localization, Time of Arrival, Linear Least Squares, Subspace Approach, Weighted Linear Least
Squares (WLLS), Two-step WLS and CRLB
1. INTRODUCTION
Wireless Sensor Network (WSN) consists of a large number of tiny low cost, low-power,
multifunctional sensors which are capable of sensing, computing and communicating between
these wireless devices which are deployed in a large geographic area[1]. WSN can be applied to a
wide variety of diverse areas[2], such as environmental monitoring, military applications, target
tracking, medical care, space exploration, location based services such as Emergency 911 (E-911)
[3], Location sensitive billing, fraud detection, intelligent transport systems, location based Social
Networking and Mobile yellow pages etc, [4]. Due to the developments in wireless
communication WSN have been a new area of research [5-7]. Many applications of WSN require
the sensor nodes to acquire the position information of the sensor nodes deployed. Data gathered
by sensors should be associated with the sensors positions and it is worthless without the
information about the place of its origin.
Despite the huge research effort, still a well accepted approach on how to solve the localization
issue is being realized. Since the sensor nodes are inexpensive and are in huge number it is not
practical to equip these sensors with a Global Positioning System (GPS) receiver. Various
localization approaches have been proposed and can be seen in the literature [8-13] and there is
not a single approach which is simple, distinct and gives decentralized solution for WSNs. The

14
Ultra Wide Band (UWB) techniques [14] give very decent localization accuracy but the systems
are expensive.
The commonly used approaches for measuring position estimate in WSN are Time of Arrival
(TOA) [15], Time Difference of Arrival (TDOA)[16], Received Signal Strength (RSS)[17] and
Angle of Arrival (AOA) a.k.a., Direction of Arrival (DOA)[18]. Where, the TOA, TDOA, and
RSS measurement gives the distance calculation between the source sensor and the receiver
sensors while DOAs provide the information of the angle and the distance measurements from the
source and the receiver. Calculating these distance and angle measurements is not simple because
of the nonlinear relationships with the source.
Given the TOA, TDOA, RSS and DOA information, the main focus of this paper is based on
TOA positioning algorithms. We consider a two dimensional (2D) rectangular area where the
sensors are deployed in Line-of-Sight (LOS) transmission, i.e., there is a direct path between the
source and each receiver [19]. Also, we conclude that the measurements are well inside the
expected range in order to obtain reliable location estimation.
The rest of the paper is organized as follows. In section 2 we present, the measurement model of
TOA, and their positioning principles. In section 3, we provide the linear approach of finding the
position by four methods i.e., Linear Least Squares (LLS), Subspace Approach (SA), Weighted
Linear Least Squares (WLLS) and Two-step WLS. In section 4, the mean square position error
comparison of the above approaches is made. Finally, the conclusions are drawn in section 5.
2. TOA MEASUREMENT MODEL AND PRINCIPLES OF SOURCE
LOCALIZATION
The mathematical measurement model for TOA based Source Localization Algorithm is given as:
r = f(x) + n (1)
Where x is the source position which needs to be estimated, r the measurement vector, ‘n’ is an
additive zero-mean noise vector and f(x) is a known nonlinear function in ‘x’.
2.1. Time of Arrival
TOA is the one-way propagation time of the signal travelling between a source and a receiver.
This means that the source and all the receivers are accurately synchronized to measure the TOA
information, and such an identical system is not needed if two way or round trip TOA is
computed. The computed TOA is then multiplied with a known propagation speed, usually
denoted as c, gives the measured distance between the source and the receivers. The measured
TOA represents a circle with its centre at the receiver and the source must lie on the
circumference in a Two Dimensional (2D) space.
Three or more such circles obtained from the noise free TOAs result in a distinct intersection
point which represents the source position and is as shown in Figure 1(a) and 1(b), specifying that
a minimum of three sensors is necessary for two dimensional position estimate [20]. If the
number of sensors is less than three there is a possibility that there may not be any intersecting
points and hence not a feasible solution. Hence, a minimum of three sensors is required to obtain
the intersection and these can be represented as a set of circular equations, based on the
optimization criterion the source position can be estimated with the knowledge of the known
sensor array geometry [21, 22].

15
(a) Trilateration (b) Multilateration
Figure 1: Geometrically representation of TOA Positioning system.
The TOA measurement model is developed as follows. Let [ ]
T
l l lx y=x be the known
coordinates of the thl sensor, 1, 2,......, .l L= and [ ]=
T
x yx be the unknown position of the
source to be estimated. The number of receivers L must be greater than or equal to 3. The
distance between the source and the sensor, denoted by ld is simply:
2 2
x-x ( ) ( ) , 1, 2,.....l l l ld x x y y l L= = − + − =
(2)
The source radiates a signal at time 0 and the thl sensor receives it at time lt . That is, { lt } are the
TOAs and is represented in a simple relationship between lt and ld :
, 1,2,....,l
l
d
t l L
c
= =
(3)
TOAs are prone to measurement errors. As a result, the range based measurement based on
multiplying lt by c, denoted by ToA, ,lr is modelled as:
2 2
TOA, TOA,( ) ( ) , 1,2,.....TOA,l l l l l lr d n x x y y n l L= + = − + − + =
(4)
where TOA, ln is the range error in TOA, ,lr which is resulted from TOA disturbance.
TOA TOA TOA( )= +r f x n (5)
where,
TOA TOA, 1 TOA, 2 TOA, L.......,
T
r r r =  r (6)
TOA TOA, 1 TOA, 2 TOA, L.......,
T
n n n =  n (7)
and

16
2 2
1 1
2 2
2 2
TOA
2 2
( ) ( )
( ) ( )
( ) = =
( ) ( )L L
x x y y
x x y y
x x y y
 − + −
 
 − + −
 
 
 
− + −  
f x d
M
(8)
Here, fTOA(x) represents the known function which is parameterized by x and in fact, it is the
noise free distance vector. The source position estimation problem based on TOA measurements
is to estimate x given {rTOA, l} or rTOA.
A zero mean uncorrelated Gaussian process with variances {σ2
TOA, l} is assumed for the range
error {nTOA,l}. This helps in to facilitate the algorithm development and analysis as well as
Cramer Rao Lower Bound (CRLB) Computation.
It is noteworthy that the zero – mean property indicates LOS transmission. The Probability
Density Function (PDF) for each scalar random variable rTOA, l, denoted by p (rTOA, l,), has the form
of
( )
2
TOA,l TOA,22
TOA,TOA,
1 1
( ) exp
22
l l
ll
p r r d
 
= − −  πσπσ  
(9)
And is characterized by its mean and variance, dl and {σ2
TOA, l}, respectively. In other words, we
can write
2
TOA, TOA,l( , )l lr N d∼ σ
(10)
While the PDF for rTOA, denoted by p(rTOA), is
( )
( ) ( )1/2/2
1 1
exp
22
T -1
TOA TOA TOAL
TOA
p
π
 
= − − − 
 
TOA(r ) r d C r d
C
(11)
where
2 2 2
,1 ,2 ,( , ,........., )TOA TOA TOA Ldiag= σ σ σTOAC (12)
2.2. Cramer Rao Lower Bound
It is known that the MSPE of a biased estimator cannot be less than the CRLB. The mean square
position error of the various positioning algorithms is computed and compared with CRLB which
gives a lower bound on variance attainable by any unbiased estimators for the same data set [23].
Given the conditional PDF, we may derive the CRLB for TOA based location estimation and the
same is given as follows [24, 25]:
• Calculate the second order derivatives of the logarithm of the measured PDF with respect to
x, that is 2
ln ( ) / ( )T
p∂ ∂ ∂r x x ..
• Take the expected value of 2
ln ( ) / ( )T
p∂ ∂ ∂r x x .
• To yield 2
( ) = { ln ( ) / ( )}T
E p∂ ∂ ∂I x r x x

17
Where I(x) denotes the Fisher Information Matrix (FIM). And the lower bound for x and y are
given by -1
1,1
(x)  I and -1
2,2
(x)  I respectively.
Alternatively, when the measurement errors are zero-mean Gaussian distributed, the FIM, whose
elements are defined as:
( ) 1( ) ( )
X
T
−∂ ∂   
=    ∂ ∂   
f x f x
I C
x x (13)
where C is the covariance matrix. The FIM based on TOA measurements denoted by:
( ) 1TOA TOA
TOA
( ) ( )
T
TOA
−∂ ∂   
=    ∂ ∂   
f x f x
I X C
x x (14)
It is straightforward to show that
1 1
2 2 2 2
1 1 1 1
2 2
2 2 2 2TOA
2 2 2 2
2 2 2 2
1 1
( ) ( ) ( ) ( )
( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
L L
L L
x x y y
x x y y x x y y
x x y y
x x y y x x y y
x x y y
x x y y x x y y
− − 
 
− + − − + − 
 − −
 ∂ 
= − + − − + −  ∂   
 
 − −
 
− + − − + −  
f x
x
M M
(15)
1 1
1 1
2 2
T O A
2 2
( )
L L
L L
x x y y
d d
x x y y
d d
x x y y
d d
− − 
 
 
− − 
∂   =   ∂ 
 
 
− − 
  
f x
x
M M
(16)
Employing eq. (16) and eq. (12), eq. (14) becomes
( )
( )
2
2 2
1 1, ,
TOA 2
2 2
1 1, ,
( )( )
( )
( )( )
L L
l l l
l lTOA l l TOA l l
L L
ll l
l lTOA l l TOA l l
x x x x y y
d d
y yx x y y
d d
2 2
= =
2 2
= =
 − − −
 
σ σ 
=  
−− − 
 σ σ 
∑ ∑
∑ ∑
I x
(17)
Where the lower bound for x and y are denoted by -1
1,1
( )  I x and -1
2,2
( )  I x respectively, and the

18
( )TOACRLB x , is
-1 -1
TOA TOA1,1 2,2
( ) ( ) ( )TOACRLB    = +   x I x I x
(18)
3. LINEAR APPROACHES FOR SOURCE LOCALIZATION
Given the nonlinear expressions, the linear localization methodology tries to convert the nonlinear
expressions of eq. (4), into a set of linear equations with zero mean disturbances. And a global
solution is obtained based on the corresponding optimization cost function. In this paper four
linear positioning approaches namely Linear Least Squares (LLS), Subspace Approaches (SA),
WLLS and two Step WLS are presented.
3.1. Linear Least Squares
The LLS approach utilizes the ordinary Least Squares (LS) technique to estimate the position of x
by reorganizing eq. 4 into linear equations [26]. And an intermediate variable is added which is a
linearization function to estimate the source position.
The linear TOA measurement model in x can be obtained by squaring eq. 4 on both sides,
( ) ( ) ( ) ( )
2 2 2 22 2
, , ,2 ,TOA l l l TOA l TOA l l lr x x y y n n x x y y= − + − + + − + − (19)
1,2,..........,where l L=
Let
( ) ( )
2 22
, , ,2TOA l TOA l TOA l l lm n n x x y y= + − + − (20)
be the noise term in eq. (19) and introduce a dummy variable R of the form:
2 2
R x y= + . (21)
Substituting eq. (20) – eq. (21) into eq. (19) yields:
( ) ( )
2 22
, ,TOA l l l TOA lr x x y y m= − + − +
2 2 2 2 2
, ,2 2TOA l l l l l TOA lr x x x x y y y y m⇒ = − + + − + +
2 2 2
, ,2 2 , 1,2,....,l l TOA l TOA l l lx x y y R m r x y l L⇒ − − + + = − − = (22)
Let
1 1
2 2
2 2 1
2 2 1
2 2 1L L
x y
x y
x y
− − 
 − − =
 
 
− − 
M M M
A (23)

19
[ ]y
T
x R=θ (24)
,1 ,2 ,
T
TOA TOA TOA Lm m m =  Lq (25)
and
2 2 2
,1 1 1
2 2 2
,2 2 2
2 2 2
,
TOA
TOA
TOA L L L
r x y
r x y
r x y
 − −
 
− − =
 
 
− −  
M
b (26)
The matrix form for eq. (22) is then:
Aθ + q = b (27)
where the observed TOAr of eq. (5) is now transformed to b, θ contains the source location to be
determined and A is constructed from the known receiver positions. When { ,TOA lm } are
sufficiently small such that
( ) ( )
( ) ( )
( ) ( )
2 2
,1 1 1
2 2
,1 2 2
2 2
,1
2
2
2
TOA
TOA
TOA L L
n x x y y
n x x y y
n x x y y
 − + −
 
 
− + − ≈
 
 
 
− + − 
M
q (28)
can be considered a zero-mean vector, that is { } 0,E ≈q we can approximate eq. (27) as:
≈Aθ b . (29)
The LS cost function based on eq. (29), denoted by ( ),LLS TOAJ %θ , is:
( ) ( ) ( ),
T
LLS TOAJ =% % %θ Aθ - b Aθ - b
= 2T T T T T
− +% % %θ A Aθ θ A b b b (30)
Which is a quadratic function in θ% , and is a sole minimum in ( ),LLS TOAJ θ% . The LLS estimate
corresponds to:
( ),arg min LLS TOAJ= %
)
%
θ
θ θ (31)
which can be easily computed by differentiating eq. (30) with respect to θ% and equating the
resulting expression to zero:

20
( ),
0
LLS TOAJ
=
=
)
%
%
%
θ θ
θ
θ
2 2 0T T
⇒ − =
)
A Aθ A b
T T
⇒ =
)
A Aθ A b
( )
1T T−
⇒ =
)
θ A A A b (32)
The LLS position estimate can be obtained from the first and second entries of θ
)
, that is,
1 2
T
    =     
) ))
x θ θ (33)
Eq. (33) is also known as the least squares calibration method [27].
3.2 Subspace Approach
The subspace positioning approach using TOA measurement is presented as follows. We first
define a ( x 2L ) matrix X:
1 1
2 2
L L
x x y y
x x y y
x x y y
− − 
 − − =
 
 
− − 
M M
X (34)
which is parameterized by x. With the use of X, the multidimensional similarity matrix [28],
denoted by D, is constructed as:
= T
D XX (35)
whose ( , )m n entry can be shown to be
[ ] ( )( ) ( )( ), m n m nm n
x - x x - x + y - y y - y=D
( ) ( ) ( ) ( )
2 2 2 2 2 2
0.5 ( ) ( )m m n n m n m nx x y y x x y y x x y y = − + − + − + − − − − −
 
( )2 2 2
0.5 m n mnd d d= + + (36)
where ( )
2 2
( )mn nm m n m nd d x x y y= = − + − is of known value because it represents the
distance between the m th and n th receivers. We then represent D using Eigen Value
Decomposition (EVD):
= Λ T
D U U (37)

21
Where [ ]= L1 2 LU u u u is an orthonormal matrix whose columns are corresponding eigen
vectors and ( )1 2, ,......., Ldiag= λ λ λΛ is the diagonal matrix of eigen values of D with
1 2 0Lλ ≥ λ ≥ ≥ λ ≥L . Noting that the rank of D is 2, we have 3 4 0.Lλ = λ = λ =L As a result
eq. (37) can also be written as:
T
S S S=D U Λ U
( )1/2 1/2 T
S S S S= U Λ U Λ
( )1/2 1/2 T
S S S S= U Λ U Λ (38)
Where [ ] ( )1 2, ,S S diag= = λ λ1 2U u u Λ and
1 1 1
2 2 2
1 2,S diag
 
= λ λ 
 
Λ denote the signal subspace
components while is the rotation matrix such that T
= I . Comparing eq. (35) and eq. (38)
yields:
1/2
S S=X U Λ (39)
We then determine the unknown rotation matrix
( )
†1/2
S S= U Λ X
( ) ( ) ( )
1
1/2 1/2 1/2T T
S S S S S S
−
 =   
U Λ U Λ U Λ X
1/2 T
S S
−
= Λ U X (40)
where
† is moore penrose pseudo inverse.
Substituting eq. (40) into eq. (39) results in
T
S S=X U U X (41)
which implies that the position x can be extracted from the eigen vectors of the signal subspace.
As , 1,2,...., ,ld l L= is not available, we construct a practical D according to
[ ] ( )2 2 2
,
0.5 m n mnm n
d d d= + +D . (42)
If the measurement error is present [28], 1/2
S SU Λ is the LS estimate of X up to a rotation. And
hence eq. (41) becomes an approximate relation.
To derive the position estimate, we first rewrite X as
1 T
= −X Y x (43)
Where

22
1 1
2 2
L L
x y
x y
x y
 
 
 =
 
 
 
Y
M M
(44)
Using the subspace relation T T
S S n n= −U U I U U where [ ]3 4n L= LU u u u corresponds to the
noise subspace and Substituting eq. (43) into eq. (41), we get:
1T T T
n n n nx ≈U U U U Y (45)
Following, the LLS procedure in eq. (29) – eq. (32), the subspace estimate of x using TOA
measurements is computed as:
( )( )†
1
T
T T
n n n n=
)
x U U U U Y (46)
1
1 1
=
T T
n n
T T
n n
Y U U
U U
(47)
The classical multidimensional scaling approach [29] is a modified subspace technique.
3.3 Weighted Linear Least Squares Approach
Since LLS is a simple approach and provides an optimum estimation performance only when the
disturbances in the linear equations are independent and identically distributed. From eq. (28), it
is clear that because of the noise vector q the LLS TOA-based positioning approach is
suboptimal. The localization accuracy can be improved if we include a symmetric weighting
matrix, say, W, in the cost function, denoted by ( ),WLLS TOAJ %θ . The final obtained expression is the
WLS cost function which is of the form:
( ) ( ) ( ), =
= 2
T
WLLS TOA
T T T T T
J
− +
% % %
% % %
θ Aθ - b W Aθ - b
θ A WAθ θ A Wb b Wb
(48)
According to eq. (27) – eq. (28), we have { }E =b Aθ which corresponds to the linear unbiased
data model. The optimum W, can be obtained similarly as best linear unbiased estimator (BLUE)
[30, 31], which is equal to the inverse of the covariance of q. That is, the weighting matrix is
similar to that if the maximum likelihood methodology. Employing eq. (28), we obtain:
{ }
( )
1
1
2 2 2 2 2 2
,1 1 ,2 2 ,
2 2 2 2 2 2
,1 1 ,2 2 ,
=
4 ,4 ,........,4
1 1 1 1
= diag , ,........,
4
T
TOA TOA TOA L L
TOA TOA TOA L L
E
diag d d d
d d d
−
−
 
 
 ≈ σ σ σ 
 
  σ σ σ 
W qq
(49)

23
As { }ld are not available, a small error condition can be obtained by replacing ld with
,TOA lr which is valid an optimum W, for sufficiently small error condition:
2 2 2 2 2 2
,1 ,1 ,2 ,2 , ,
1 1 1 1
= , ,........
4 TOA TOA TOA TOA TOA L TOA L
diag
r r r
 
  σ σ σ 
W (50)
Following eq. (31) – eq. (32), the WLLS estimate of θ is:
( )
( )
,
1
= arg min
=
WLLS TOA
T T
J
−
)
)
%
θ
θ θ
A WA A Wb
(51)
The Weighted Linear Least Square position estimate is given by eq. (33).
With only a moderate increase of computational complexity [31], eq. (51) is superior to eq. (32)
in terms of estimation performance. The localization accuracy can be improved by making use of
3
  
)
θ according to the relation eq. (21) as follows. When
)
x of eq. (51) is sufficiently close to x,
we have:
( )( )
( )
2
2
1 1 1
1
=
2
x x x
x x
     − + −     
 ≈ − 
) ) )
)
θ θ θ
θ
(52)
Similarly, for
2
  
)
θ :
( )
2
2
2 2
2y y y   − ≈ −   
) )
θ θ (53)
Based on eq. (21) and with the use of eq. (52) – eq. (53), we construct:
= +h Gz w (54)
where
2 2
1 2 3
T
      =        
) ) )
h θ θ θ (55)
1 0
0 1
1 1
 
 =  
  
G (56)
2 2 T
x y =  z (57)
and
( ) ( )1 2 3
2 2
T
x x y x R      = − − −      
) ) )
w θ θ θ (58)

24
Note that z is the parameter vector to be determined. The result of BLUE is used to compute the
covariance of w in x
)
and is of the form [31]:
{ } ( )
1
1 2 3 1 2 3
T
T
E x y R x y R
−
              − − − − − − =              
) ) ) ) ) )
θ θ θ θ θ θ A WA (59)
Employing eq. (58) – eq. (59), the optimal weighting matrix for eq. (54), denoted byΦ , is then:
( )( ) ( )
11
2 ,2 ,1 2 ,2 ,1T
diag x y diag x y
−−
 =
  
Φ A WA (60)
As a result, the WLLS estimate of z is
( )
1T T−
=
)
z G ΦG G Φh (61)
As there is no sign information for x in z , the final position estimate is determined as:
( ) [ ] ( ) [ ]1 21 2
sgn sgn
T
    =     
) )) ) )
x θ z θ z (62)
Where sgn represents the signum function [32]. This technique is called the two-step WLS
estimator [33] where eq. (21) is used in an implicit manner. Similarly an explicit way is to use
Lagrangian multipliers [34, 35] to minimize eq. (48) subject to the constraint of eq. (21).
4. SIMULATION RESULTS
The performance evaluation of the various linear TOA based localization approaches is simulated
using MATLAB[TM] Version 7.10.0.499 (R2010A) on Microsoft Windows XP®, Professional
Version 2002, Service Pack 3, 32 bit operating system installed on Intel[R], Core[TM] 2 Duo
CPU, E4500 @ 2.20GHz, 2.19GHZ, 2.0Gb of Ram.
The simulation is done in a 2- Dimensional region with a size of (1100m x 1100m), where the
unknown source is assumed to be at position (x, y) = (200, 300), and the receivers are positioned
in known coordinates at (0, 0), (0, 1000), (1000, 1000) and (1000, 0) respectively. And the
sensors are deployed in a rectangular area where the source is surrounded by four receivers and is
shown in Figure 2. It is also assumed that { },TOA ln are zero- mean uncorrelated Gaussian process
with variances { }2
,TOA lσ , and zero- mean property indicates LOS transmission. The range error
variance 2
,TOA lσ is proportional to 2
ld with 2 2
,= /l TOA lSNR d σ . The signal-to- noise-ratio (SNR) =
30dB has been assumed. All the methods estimate the position. The implementation flow of the
two step WLS and the CRLB is shown in Figure 3 and 4 respectively. Figure 5, shows the plot of
Mean Square Position Error (MSPE) defined as ( ) ( ){ }2 2
x x y y− + −
) )
E of different linear
approaches and CRLB for SNR in the range [-10, 25] dB, based on 1000 independent runs, which
is given by ( ) ( )
1000 2 2
1
/1000i ii
x x y y=
 − + −
 ∑
) )
, where( ),i ix y
) )
, denotes the position estimate of
the i th run. From the Figure, it can be seen that the Two-step WLS estimator achieves the
optimal estimation performance, while the LLS, SA and WLLS approaches are suboptimal.

25
Figure 2: Position of Source and Receivers
Table 1: Estimated TOA positions obtained using different linear approaches.
Method x
)
in meters y
)
in meters
Linear Least Squares 201 302
Subspace Approach 197 297
Weighted Linear Least Squares 199 300
Two Step Weighted Least Squares 200 300
Table 1 gives the results of the position estimate, and results of the two step WLS approach is
better when compared with the other linear approaches. The two step WLS accurately estimates
the positions. The accuracy is higher in case of Two- step WLS while the other approaches have
lower localization accuracy.

26
Figure 3: Flowchart showing the computation of Two- Step WLS approach

27
Figure 4: Flowchart showing the computation of CRLB
Figure 5: Mean Square Position Error Computation of the different linear based approaches

28
5. CONCLUSIONS
The work presented addresses the problem of position estimation of a sensor node in a Wireless
Sensor Network, using TOA measurements in LOS environments. The CRLB for the position
estimation problem has been derived first and later four methods namely LLS, SA, WLLS and
Two Step WLS methods of linear approach have been derived and presented. Extensive
simulations have been carried out and the results of different methods have been compared. The
comparison reveals that the two step WLS method is superior to the rest of the linear approaches
in LOS environments.
We have restricted our studies to the linear approaches. The work can be extended to the
nonlinear approaches also and shall be reported in a future communication. Tracking mobile
nodes is an interesting problem which may require a combination of two or more approaches to
improve the accuracy of the position estimate.
ACKNOWLEDGEMENTS
The authors would like to thank Dr. Yerriswamy T, (PDIT Hospet), and their Colleagues at Dept.
of CSE, JNNCE Shimoga, for the valuable inputs and the reviewers for their useful comments
and suggestions.
REFERENCES
[1] Ravindra. S and Jagadeesha S N, (2013), “A Nonlinear Approach for Time of Arrival Localization in
Wireless Sensor Networks”, 6th
IETE National Conference on RF and Wireless IConRFW-13,
Jawaharlal Nehru National College of Engineering, Shimoga, Karnataka, India 09-11-May, pp 35-40.
[2] Yerriswamy. T and Jagadeesha S. N, (2011), “Fault Tolerant Matrix Pencil method for Direction of
Arrival Estimation”, Signal & Image Processing: An International Journal, Vol. 2, No. 3, pp. 42-46.
[3] The Federal Communications commission about “Emergency 911 wireless services”,
http://www.fcc.gov/pshs/services/911-services/enhanced911/Welcome.html
[4] Pawel Kulakowski, Javier Vales-Alonso, Esteban Egea-Lopez, Wieslaw Ludwin and Joan Garcia-
Haro, (2010), “Angle of Arrival localization based on antenna arrays for wireless sensor networks”,
article in press, Computers and Electrical Engineering Journal Elsevier,
doi:10.1016/j.compeleceng.2010.03.007
[5] Karl H, Willing A, (2005), Protocols and architectures for wireless sensor networks. John Wiley and
Sons.
[6] Stojmenovic I, (2005), Handbook of sensor networks algorithms and architectures. John Wiley and
Sons.
[7] Wang C, Xiao L. (2007), Sensor localization under limited measurement capabilities. IEEE Network,
21:16–23.
[8] Patwari N, Ash J N, Kyperountas S, Hero III AO, Moses RL, Correal NS, (2005), “Locating the
nodes: Cooperative Localization in Wireless Sensor Networks”. IEEE Signal Process Mag., 22:54–
69.
[9] H. Wymeersch, J. Lien and M. Z. Win, (2009), “Cooperative Localization in Wireless Networks,”
IEEE Signal Processing Mag., vol.97, no.2, pp.427-450.
[10] R Zekavat and R. M Buehrer (2011), “Handbook of Position Location: Theory, Practice and
Advances,” Hoboken, NJ: John Wiley & Sons, INC.
[11] A. H. Sayed, A. Tarighat, and N. Khajehnouri, (2005), “Network based Wireless Location:
Challenges faced in developing techniques for accurate wireless location information,” IEEE Signal
Processing. Mag., Vol. 22, no. 4, pp, 24-40.
[12] Neal Patwari, Alfred O Hero III, Matt Perkins, Neiyer S Correal and Robert J O’dea. (2003),
“Relative location estimation in Wireless Sensor Networks,” IEEE Transactions on Signal
Processing, 51(8):2137-2148.
[13] Yerriswamy. T and Jagadeesha S. N, (2012), “IFKSA- ESPRIT – Estimating the Direction of Arrival
under the Element Failures in a Uniform Linear Antenna Array,” ACEEE International Journal on
Signal & Image Processing, Vol. 3, No. 1, pp. 42 – 46, 2012.

29
[14] Gezici S, Tian Z, Giannakis GB, Kobayashi H, Molisch AF, Poor HV, et al. (2005), “Localization via
ultra-wideband radios,” IEEE Signal Process Mag.; 22:70–84.
[15] A. Jagoe (2003), Mobile Location Service: The Definitive Guide, Upper Saddle River: Prentice- Hall.
[16] M. Ilyas and I. Mahgoub, (2005), Handbook of Sensor Networks: Algorithms and Architectures, New
York: Wiley.
[17] J. C. Liberti and Theodore S Rappaport, (1990), Smart Antennas for Wireless Communications: IS-95
and Third Generation CDMA Applications, Upper Saddle River: Prentice- Hall.
[18] Yerriswamy. T and S. N. Jagadeesha, (2012) “Joint Azimuth and Elevation angle estimation using
Incomplete data generated by Faulty Antenna Array,” Signal & Image Processing: An International
Journal, Vol. 3, No. 6, pp. 99-114.
[19] Theodore S Rappaport et al (1996), Wireless Coommunication: principles and practice, Vol. 2,
Prentice Hall PTR New Jersy.
[20] S. M. Kay, (1988), Modern Spectral estimation: Theory and Application. Englewood Cliffs, NJ:
Prentice Hall.
[21] S. M. Kay, (1993), Fundamentals of Statistical Signal Processing: Estimation Theory, Prentice-
[22] Y. Huang and J. Benesty, Eds., (2004), Audio Signal Processing for Next Generation Multimedia
Communication systems, Kluwer Academic Publishers.
[23] E. G. Larsson, ( 2004), “Cramer Rao Bound analysis of distributed positioning in sensor networks,”
IEEE Signal processing Lett., Vol. 11, no. 3, pp. 334-337.
[24] Feng Yin, Carsten Fritsche, Fredrik Gustafsson and Abdelhak M Zoubir, (2013) “ TOA- Based
Robust Wireless Geolocation and Cramer Rao Lower Bound Analysis in Harsh LOS/NLOS
Environments, IEEE Transactions on Signal Processing, (61), 9, 2243-2255.
http://dx.doi.org/10.1109/TSP.2013.2251341
[25] D. J. Torrieri, (1984) “Statistical theory of passive location systems,” IEEE Transactions on
Aerospace and Electronic Systems, vol 20, no.2, pp. 183-198.
[26] K. W. Cheung, H.C. So, W.-K. Ma and Y.T.Chan, (2004) “Least squares algorithms for time-of-
arrival based mobile location,” IEEE Transactions on Signal Processing, vol.52, no.4, pp.1121-1128.
[27] C. Mensing and S. Plass (2006), “Positioning algorithms for cellular networks using TDOA,” Proc.
IEEE International Conference on Acoustics, Speech and Signal Processing, vol. 4, pp. 513-516,.
[28] H. C. So and K W Chan, (2005) “A generalized subspace approach for mobile positioning with time-
of-arrival measurements,” IEEE transactions on Signal Processing, Toulouse, France Vol.53, no.2,pp
460-470.
[29] H.-W. Wei, Q. Wan, Z.-X. Chen and S. -F. Ye, (2008) “Multidimensional Scaling – based passive
emitter localization from range-difference measurements,” IET Signal Processing, vol.2, no.4,
pp.415-423.
[30] F.K.W. Chan, H.C.So, J. Zheng and K.W.K. Lui, (2008), “Best Linear Unbiased Estimator approach
for time-of –arrival based localization,” IET Signal Processing, vol.2, no.2,pp.156-162.
[31] J. C. Chen, R.E. Hudson and K. Yao, (2002), “Maximum-likelihood source localization and unknown
sensor location estimation for wideband signals in the near field,” IEEE Transactions on Signal
Processing, vol.50, no.8, pp.1843-1854.
[32] G. C. Carter, Ed. (1993), Coherence and Time Delay Estimation: An Applied Tutorial for Research,
Development, test, and Evaluation Engineers. New York: IEEE.
[33] M. Marks and E. Niewiadomska - Szynkiewicz, (2007), “Two-Phase Stochastic optimization to
sensor network localization,” Proc. IEEE International Conference on Sensor Technologies and
Applications, pp. 134-139, Valencia, Spain.
[34] Y. T. Chan and K.C. Ho, (1994) “A Simple and Efficient estimator for hyperbolic location,” IEEE
Transactions on Signal Processing, vol. 42, no. 8, pp. 1905 – 1915.
[35] H. C. So, Y. T. Chan and F.K.W. Chan, (2008), “Closed-form formulae for optimum time difference
of arrival based localization,” IEEE Transactions on Signal Processing, vol.56, no.6, pp. 2614-2620.

30
Authors
Ravindra. S. received his B.E., in Electrical and Electronics Engineering., and
M.Tech., in Networking and Internet Engineering, from Visvesvaraya
Technological University, Belgaum, Karnataka, India in 2006 and 2008
respectively. He is currently working towards a Doctoral Degree from
Visvesvaraya Technological University, Belgaum, Karnataka, India. At present
he is working as Assistant Professor, in Computer Science and Engineering
department of Jawaharlal Nehru National College of Engineering (affiliated to
Visvesvaraya Technological University), Shimoga, Karnataka, India.
Dr. S.N. Jagadeesha received his B.E., in Electronics and Communication
Engineering, from University B. D. T College of Engineering., Davangere
affiliated to Mysore University, Karnataka, India in 1979, M.E. from Indian
Institute of Science (IISC), Bangalore, India specializing in Electrical
Communication Engineering., in 1987 and Ph.D. in Electronics and Computer
Engineering., from University of Roorkee, Roorkee, India in 1996. He is an
IEEE member. His research interest includes Array Signal Processing, Wireless
Sensor Networks and Mobile Communications. He has published and presented
many papers on Adaptive Array Signal Processing and Direction-of-Arrival
estimation. Currently he is professor in the department of Computer Science and
Engineering, Jawaharlal Nehru National College of Engg. (Affiliated to
Visvesvaraya Technological University), Shimoga, Karnataka, India.

TIME OF ARRIVAL BASED LOCALIZATION IN WIRELESS SENSOR NETWORKS: A LINEAR APPROACH

More Related Content

What's hot (16)

Viewers also liked (19)

Similar to TIME OF ARRIVAL BASED LOCALIZATION IN WIRELESS SENSOR NETWORKS: A LINEAR APPROACH (20)

Recently uploaded (20)

TIME OF ARRIVAL BASED LOCALIZATION IN WIRELESS SENSOR NETWORKS: A LINEAR APPROACH