2580 IEEE TRANSACTIONS ON SIGNAL PROCESSING

Efficient Closed-Form Algorithms for AOA

Based Self-Localization of Sensor Nodes
Using Auxiliary Variables
Hua-Jie Shao, Xiao-Ping Zhang, Senior Member, IEEE, and Zhi Wang, Member, IEEE

Abstract—Node self-localization is a key research topic for I. INTRODUCTION

wireless sensor networks (WSNs). There are two main algorithms,
the triangulation method and the maximum likelihood (ML) esti-
mator, for angle of arrival (AOA) based self-localization. The ML
estimator requires a good initialization close to the true location
to avoid divergence, while the triangulation method cannot obtain
L OCATION information of sensor nodes in wireless sensor
networks (WSNs) is essential since it is the basis of a
number of tasks such as target localization and tracking [1]–[4],
the closed-form solution with high efficiency. Here, we develop a
and animal behavior monitoring [5], etc. In practice, random
set of efficient closed-form AOA based self-localization algorithms
using auxiliary variables based methods. First, we formulate deployment of sensor nodes (e.g., dropped from aircraft) is
the self-localization problem as a linear least squares problem common since regions of interest (ROI) usually reside in harsh
using auxiliary variables. Based on its closed-form solution, a environment. As a result, the location of sensor nodes is often
new auxiliary variables based pseudo-linear estimator (AVPLE) unknown. Obtaining accurate location information for sensor
is developed. By analyzing its estimation error, we present a bias
nodes becomes an important task. A possible method is to
compensated AVPLE (BCAVPLE) to reduce the estimation error.
Then we develop a novel BCAVPLE based weighted instrumental equip each sensor node with the GPS receivers [6], but it is
variable (BCAVPLE-WIV) estimator to achieve asymptotically expensive especially for large-scale networks. With the budget
unbiased estimation of locations and orientations of unknown and energy constraints [7], [8], only limited number of sensor
nodes based on prior knowledge of the AOA noise variance. In nodes can be equipped with GPS receivers. These nodes are
the case that the AOA noise variance is unknown, a new AVPLE
beacons as their accurate locations are known by the GPS
based WIV (AVPLE-WIV) estimator is developed to localize the
unknown nodes. Also, we develop an autonomous coordinate ro- receivers. Other nodes without GPS receivers are unknown
tation method to overcome the tangent instability of the proposed nodes, whose locations have to be estimated.
algorithms when the orientation of the unknown node is near . Many techniques have been proposed for sensor node local-
We also derive the Cramér-Rao lower bound of the ML estimator. ization, such as received signal strength indicator (RSSI), time
Extensive simulations demonstrate the new algorithms achieve
of arrival (TOA), angle of arrival (AOA), and time difference
much higher localization accuracy than the triangulation method
and avoid local minima and divergence in iterative ML estimators. of arrival (TDOA) [6], [9]. In this paper, we focus on the AOA
technique [10], [11] for node self-localization in WSNs. One
Index Terms—Angle of arrival, auxiliary variables, bias com-
pensated auxiliary variable based pseudo-linear estimator, closed- common method to obtain AOA measurements is to use an an-
form pseudo-linear estimator, node self-localization, weighted in- tenna array on each sensor node [12]. Different from target lo-
strumental variables, wireless sensor networks. calization, AOA based node self-localization needs to estimate
both the location and orientation of each unknown node. The
orientation in the paper refers to the reference direction against
Manuscript received June 18, 2013; revised October 23, 2013 and January
23, 2014; accepted March 06, 2014. Date of publication March 26, 2014; date which the AOAs are measured.
of current version April 23, 2014. The associate editor coordinating the re- A number of estimation algorithms have been proposed
view of this manuscript and approving it for publication was Prof. Urbashi
for AOA based self-localization method over the past years,
Mitra. This work was supported in part by the National Natural Science Founda-
tion of China (No. NSFC61273079 No. NSFC11174316), the Strategic Priority such as triangulation [10], [12], maximum likelihood (ML)
Research Program of the Chinese Academy of Sciences (XDA06020300 and estimator [13]–[15] and the semidefinite programming (SDP)
XDA06020201), the Open Research Project of the State Key Lab. of Industrial
method [11]. The triangulation method uses two beacons and
Control Tech., Zhejiang University No. ICT1423, No. ICT1430, the Key Lab.
of Wireless Sensor Network & Communication of Chinese Academy of Sci- their AOAs to find the center and radius of the circumscribed
ences No. WSNC2011001, and the Natural Sciences and Engineering Research circle for the triplet of two beacons and one unknown node.
Council of Canada (NSERC), Grant RGPIN239031. (Corresponding author: Z.
Then the trilateration approach is used to obtain the locations
Wang.)
H.-J. Shao is with the Department of Control Science and Engineering, Zhe- of nodes based on the known radiuses and centers of these
jiang University, Hangzhou, 310027, China (e-mail: [email protected]). circumscribed circles. However, the localization accuracy of
X.-P. Zhang is with the Department of Electrical and Computer Engineering,
the triangulation method is low because of the error accumu-
Ryerson University, Toronto, Ontario, Canada, M5B 2K3 (e-mail: xzhang@ee.
ryerson.ca). lation in the process of obtaining the center and radius of the
Z. Wang is with the Department of Control Science and Engineering, Zhe- circumscribed circle. The ML estimator is an asymptotically
jiang University, Hangzhou, 310027, China (e-mail: [email protected]).
unbiased estimator for AOA based self-localization problem.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. However, it is a nonlinear least squares estimator for Gaussian
Digital Object Identifier 10.1109/TSP.2014.2314064 angle measurement noise. The ML method either requires a

1053-587X © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2581

reasonable initialization close to the true solution or may suffer estimator and an AVPLE-WIV estimator to estimate the un-
from local minima and even divergence problems [16], [17]. known node. Section VI presents the ML estimator to localize
Recently, Biswas et al. [11], [18] proposed a SDP method to the unknown node for AOA based self-localization and derives
estimate the locations of nodes for large scale sensor networks. the CRLB of it. Section VII evaluates the performances of our
However, the SDP relaxation method needs high connectivity proposed algorithms based on extensive simulations. Finally,
among nodes and beacons so that the size of the triplets that Section VIII concludes the paper.
have mutual angle information is large enough for the set of The main notation in this paper is listed as follows:
equations to have a unique solution [11]. It has a complicated
number of beacons
structure and high computational complexity [6]. Also note
that in the SDP for AOA based node localization [11], the bea- position and orientation vector
cons and unknown nodes need to measure each other’s AOAs of the unknown node
mutually, consuming much energy of the nodes and beacons. position vector of th beacon
Moreover, the SDP cannot estimate both the locations and
orientations of unknown nodes simultaneously. Its localization measured AOA with noise for
accuracy depends on the size of the triplets with relative angle the th beacon
information and the locations of beacons. Some beacons are true AOA for the th beacon
required to be placed on the perimeter of the WSNs area for the AOA measurement noise for
SDP method to be effective. the th beacon
In [19], [20], closed-form solutions of the bearings only target
tracking are presented. Those closed-form solutions of bearing- AOA noise variance
only (i.e., AOA) target tracking only estimate the target location vector of AOAs with noise
, and cannot be applied to estimate both the locations and
orientations of unknown nodes simultaneously. vector of AOAs as a function
In this paper, we develop an effective auxiliary variables of position vector of the
based method to obtain the closed-form solution of AOA unknown node
based self-localization problem. It improves the localization accurate auxiliary variables
accuracy compared with the triangulation method and avoids related to position vector of
the problems of iterative ML methods. First, we develop an the unknown node
auxiliary variables based pseudo-linear estimator (AVPLE) estimated auxiliary variables
algorithm, which employs auxiliary variables to formulate for the AVPLE
the self-localization problem as a linear least squares problem
and gives a closed-form solution. By analyzing the estimation estimation bias of auxiliary
error of the AVPLE algorithm, we then present a novel bias variables for the AVPLE
compensated AVPLE (BCAVPLE) scheme to reduce the es- bearing-only likelihood
timation error. Furthermore, we develop a BCAVPLE based function
weighted instrumental variable (BCAVPLE-WIV) estimator to ML cost function
achieve asymptotically unbiased estimation of locations and
orientations of unknown nodes based on the prior knowledge assumed observer location so
of the AOA noise variance. In the case that the AOA noise as to analyze the estimation
variance is unknown, an AVPLE based WIV (AVPLE-WIV) bias
estimator is developed to localize the unknown nodes. We assumed true target location
further investigate the impact of the orientation on estimation
estimated target location for
performance of the new algorithms and present an autonomous
using AVPLE
coordinate rotation (ACR) method to overcome the tangent
instability of our proposed algorithms when the orientation noisy AOA vector for the
of the unknown node is near . To analyze the theoretical observer
performance, we derive the Cramér-Rao lower bound (CRLB) orthogonal unit vector for
of the ML estimator. Extensive simulation results demonstrate
that the new algorithms achieve much higher localization the range from observer to
accuracy than the triangulation method and also avoid local target
minima and divergence problems in iterative ML estimators. Approximate AOA noise
The paper is organized as follows. Section II states the variance when the AOA noise
problem of node self-localization. Section III presents an is small
effective auxiliary variables based method to estimate the estimation errors of position
unknown node with closed-form solutions. Section IV analyzes vector
the estimation error of node self-localization using the AVPLE
algorithm. Section V provides a BCAVPLE estimator to reduce Fisher information matrix for
the estimation error and further presents a BCAVPLE-WIV the ML estimator
2582 IEEE TRANSACTIONS ON SIGNAL PROCESSING

node with respect to its own orientation from the th beacon.

The AOA measurements are given as follows [21]:

(1)

where is the true AOA measurement from the th beacon,

is the measurement noise following a
Gaussian distribution with mean zero and variance of . Note
that in [21], it has been shown that the errors of AOA esti-
mation by the MUSIC algorithm are asymptotically Gaussian
Fig. 1. Illustration of the AOA based node self-localization in WSNs. distributed with zero means.
In this paper, the objective is to estimate the location and ori-
entation of the unknown node based on the known locations of
Euclidean distance from the beacons and their AOAs measured by the unknown node.
unknown node and th beacon
error covariance of the
AVPLE III. AUXILIARY VARIABLE BASED NODE SELF-LOCALIZATION
approximate estimation bias IN CLOSED-FORM
of for large
As shown in Fig. 1, the relationship between the unknown
approximate estimation bias node and its neighbor beacons is:
of
identity matrix
(2)
estimated auxiliary variables
for BCAVPLE
where is the location of the th beacon, and
instrumental variable (IV) denote the location and orientation of the unknown node,
matrix respectively.
approximated AOA obtained After mathematical derivation, we get
from
approximated orthogonal unit
vector for
(3)
estimated auxiliary variables
for the BCAVPLE-WIV
See Appendix A for the detailed mathematical derivation from
estimated auxiliary variables (2) to (3).
for the AVPLE-WIV To solve the unknown node location estimation problem, we
introduce the following auxiliary variables:

II. PROBLEM STATEMENT OF NODE SELF-LOCALIZATION

As shown in Fig. 1, assume there are beacons equipped (4)

with GPS receivers and an unknown node without GPS re-
ceivers in a region. Each beacon knows its accurate location, Then (3) can be written as
denoted by , , but does not know
its orientation. That is to say, we do not use the orientation
information of the beacon in the estimation. The nodes in black
color are beacons. The unknown node, in yellow color, does not (5)
know its location and orientation information. The location and
orientation of the unknown node, denoted by , As the true AOA, , cannot be obtained in the outdoor envi-
are to be estimated. ronment, we replace it with the measured AOA with noise, ,
The beacons are signal emitters and the unknown node is a to formulate the linear least squares problem. Thus, (5) could be
signal receiver. Hence, the AOAs of the beacons can be mea- approximately expressed as
sured by the neighboring unknown node with respect to its own
orientation. Let denote the measured AOA by the unknown (6)
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2583

where
Algorithm 1: ACR algorithm
1: Initialization: , ;
.. .. .. (7a)
. . . 2: Use AVPLE algorithm to estimate the unknown node based
on (2)–(9), ;
3: Adopt the estimated and the locations of beacons
to calculate the estimated AOAs, denoted by
, ;
.. (7b)
. 4: Calculate an error between the estimated
AOAs and the measured AOAs , denoted by
;
(7c) 5: Rotate coordinate system, counter-clockwise 90 degrees.
Then, (2) becomes ;
A least squares criteria can be used to solve (6). We can adopt
6: Estimate the unknown node in Step 5 using AVPLE,
the PLE to get a closed-form solution. The estimated value of
;
the auxiliary variables , denoted by , is
7: Similar to Step 3 and 4, calculate the error between the
(8) estimated AOAs and the measured AOAs based on , denoted
by ;
Based on (8) and (4), the location and orientation of the un-
8: If
known node is given by
Adopt the rotated coordinate system, return;
else
(9)
Adopt original coordinate system, return;
End If
Remark 1: Note that the tangent becomes unstable when the
orientation of the unknown node is near (or ). Thus,
it may result in high estimation error of the auxiliary variables IV. ERROR ANALYSIS FOR UNKNOWN NODE LOCALIZATION
based PLE (AVPLE) method when the orientation is near .
Note that in (6), the noise contaminated AOA measurements
To solve the tangent instability problem, we develop an
are used instead of the true AOA in (5). It leads to the estima-
autonomous coordinate rotation (ACR) algorithm to au- tion error for the unknown node localization. In this section, we
tonomously rotate the coordinate system and therefore avoid analyze the theoretical estimation error of node self-localization
the tangent instability as follows. using the AVPLE algorithm. First we provide the bias analysis
Based on the estimated location and orientation of the un- of the auxiliary variables using the AVPLE, and then derive the
known node, , the locations of beacons, location error of the unknown node based on the estimation error
, and the measured AOAs, , we will test whether to of the auxiliary variables.
choose the original coordinate system or the alternative rotated Different from target localization, the error analysis for
(by ) coordinate system. Specifically, in the original coor- AOA based node self-localization involves three variables
dinate system, we compute an error between the estimated in thetwo-dimension plane. This analysis has not been done
AOAs based on the -th beacon and estimated unknown node, before. We convert the formulation of node self-localization
denoted by , and the measured AOAs , using auxiliary variables in (5) to a similar target localization
geometrical relationship in [20] as depicted in Fig. 2. We
i.e., . Similarly, we can compute assume that represents the
this error in the rotated coordinate system, denoted by . Now observer location, represents the estimated
for beacons, we compare the total square error, and target location, denotes the true target location
, and then we adopt the coordinate system with lower and represents the orthogonal error vector. In addition, we
total square error. define the orthogonal unit vector for noisy AOA vector as
The key of the algorithm is to decide when to use the rotated follows:
(by ) coordinate system. The algorithmic procedures of au-
tonomous coordinate rotation (ACR) algorithm are summarized
(10)
in Algorithm 1.
2584 IEEE TRANSACTIONS ON SIGNAL PROCESSING

Hence, the estimation bias of the auxiliary variables for AVPLE

in (8) can be given by

(20)

Then, based on Slutsky’s theorem [2], [22], for large the

AVPLE bias mentioned above can be approximated by

(21)

Fig. 2. Geometrical relationship of node self-localization using auxiliary vari-

ables for equation (5). As , (21) becomes an equality. From (21), it can be seen
that the AVPLE bias is determined by as well as .
In the following, we further analyze the relationship between
According to the target localization geometry described in with and , and obtain a more accurate theoretical
Fig. 2, we have description of the AVPLE bias. Note that

(11)

where denotes the estimated target location and is the

error vector.
Note that
(12)

Substituting (12) into (11) yields

(13)
(22)
where the orthogonal vector is defined by
Thus, the expectation of can be expressed by
(14)

and is the distance from to .

By pre-multiplying both sides of (13) with , we can get

(15)

After expansion,
(23)

(16)
where
Concatenating (16) for yields

(17) (24a)

where (24b)

If the AOA measurement noise is sufficiently small,

.. (18) , same as the variance of AOA noise in (1).
. Now we examine the autocorrelation matrix .
Based on (7a), we have (25), shown at the bottom of the next
matrices , and are the same as those in (7). Based on (17), page. The elements in (25) are
the accurate solution of AVPLE is

(19) (26a)
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2585

After obtaining the estimation bias of the auxiliary variables

for the AVPLE, denoted by , we can
(26b) get the estimation bias of the unknown node as follows
(26c)

Thus for small AOA measurement noise we can get the expec-
tation (27), shown at the bottom of the page.
Based on (23) and (27), the AVPLE bias in (21) can be
rewritten as
(29a)

(29b)
(28)

(29c)
where is the inverse of formulation (27). In the limit
when , (28) would become an equality. It can be seen
where denotes the estimation error of the
from (28) that for large the estimation bias is determined by
location and orientation for the unknown node,
as well as matrix and , which are related to
are the accurate values of the auxiliary variables for node
the locations of beacons. Moreover, (23) and (27) show that the
self-localization.
estimation error is proportional to , which approximates the
AOA noise variance when the AOA noise is small. Note that
the estimation bias does not vanish with the increase of
even if when , implying that the AVPLE is a biased V. BIAS COMPENSATION FOR AVPLE AND WEIGHTED
estimator. INSTRUMENTAL VARIABLE ESTIMATOR
Remark 2: We employ the chi-squared test to check the error In this section, we provide a bias compensation scheme for
distributions of , and find that their error dis- the AVPLE algorithm based on the prior knowledge of the AOA
tributions are not Gaussian. It is because the cosine and sine noise variance. To further reduce the estimation bias and ap-
functions in (6) and (7) are nonlinear functions that apply to the proximate the CRLB accuracy, a new BCAVPLE based WIV
measured DOA with Gaussian noise . Thus, the algorithm to (BCAVPLE-WIV) estimator is proposed. In addition, when the
estimate the auxiliary variables for node self-localization in our AOA noise variance is unknown, a new AVPLE based WIV
paper is a pseudo-linear estimator (PLE) similar to those defined (AVPLE-WIV) estimator is developed to localize the unknown
in [19], [23]–[25], and it is a biased estimator. node.

(25)

(27)
2586 IEEE TRANSACTIONS ON SIGNAL PROCESSING

A. Bias Compensation for the AVPLE (BCAVPLE) It can be seen in (33) that the approximated estimation bias is
The position estimation bias of the unknown node is deter- biased, thus the BCAVPLE is also biased. However, we can
mined by the estimation bias of auxiliary variables as shown in see from (35) that the BCAVPLE has a reduced estimation bias
(29). Therefore, it is essential to estimate the auxiliary variables compared with the AVPLE.
as accurately as possible. We show that the AVPLE is a biased Based on (35), the estimate bias of BCAVPLE is given by
estimator, motivating us to develop a method to compensate the
bias.
Note that in AVPLE estimation bias (20), the true ranges and (36)
AOA noises in are unavailable to calculate the bias . For
sufficiently large , using the similar methodology as in [19], Remark 3: The BCAVPLE algorithm is effective based on the
we can approximate the estimation bias by accurate prior knowledge of the transformed AOA noise vari-
ance , which approximates the AOA noise
variance when it is small. Without this prior knowledge of
the AOA measurement noise, the BCAVPLE cannot reduce the
(30) estimation bias effectively as shown below.
Assuming that the inaccurate AOA noise variance parameter
For large , the expectation of the approximate estimation bias is , the error between the BCAVPLE without accurate AOA
in (30) is noise variance parameter and the BCAVPLE with accurate prior
knowledge of AOA noise variance is
(31)
(37)
Since the true value of is unavailable, the approximate es-
timation bias cannot be obtained directly. To calculate the and the error between the AVPLE and the BCAVPLE with ac-
approximate estimation bias of the auxiliary variables, the curate prior knowledge of AOA noise variance is
matrix in (30) is replaced by the estimation result of AVPLE,
. Thus the estimation bias can be approximated as
(38)

(32)
Comparing (37) and (38), we can see that BCAVPLE may have
larger estimation bias for the node self-localization compared
Note that . The expectation of the approx-
with the AVPLE when the error between the accurate and
imate estimation bias in (32) can be expressed by
inaccurate is large.

B. Asymptotically Unbiased Weighted Instrumental Variable

Estimator
In general, the BCAVPLE is not an unbiased estimator though
its estimation bias could be reduced compared with the AVPLE.
To construct an asymptotically unbiased estimator, based on the
(33) methodology in [17], [19], a new BCAVPLE-WIV estimator
is proposed to localize the unknown node based on the prior
where is the identity matrix, and knowledge of the AOA noise variance. In addition, in the case
that the AOA noise variance is unknown, we present a new
(34) AVPLE-WIV estimator to estimate the unknown node. We will
elaborate these two new methods respectively below in this
section.
Note that the approximate estimation bias in (31) is unbiased,
The key of the BCAVPLE-WIV is to design and construct
while the estimation error in (33) is asymptotically biased. The
the instrumental variable (IV) matrix . Our new method is de-
estimation bias in (33) could have a small bias only if meets one
scribed as follows.
of the following conditions:
Referring to (6) and (8), the solution of the AVPLE is obtained
• ;
from the following normal equation
• .
As we obtain the approximated estimation bias , the (39)
BCAVPLE estimator can be written as
To reduce the bias caused by the correlation between and
when is nonsingular, the IV matrix is introduced to
replace the matrix in (39), yielding
(35)
(40)
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2587

where is the IV estimator ilar methodology in [19], [26], the closed-form solution of the
WIV estimator, i.e., the BCAVPLE-WIV estimator, is

(41)
(48)
The most challenging task for the IV estimator is to choose
a reasonable such that is nonsingular and If the AOA noise is independent identically distributed (i.i.d.)
, which can make the IV estimator an asymp- with the variance , , the weighted
totically unbiased estimator, i.e., as . matrix in (47) can be rewritten as
Usually, the noise-free can be chosen as the optimal IV
(49)
matrix [19], [22], i.e.,
However, sometimes the AOA noise variance is hard
.. (42) to estimate due to the complicated outdoor environments.
. We present an AVPLE-WIV estimator, which does not need
the prior knowledge of AOA noise variance, to localize the
unknown node. Similar to the BCAVPLE-WIV estimator, the
where
AVPLE-WIV estimator is given by

(43a) (50)

(43b) where is the same as in (49), and

However, the optimal IV matrix is not available since the true ..

. (51)
AOAs cannot be obtained in the open air environment with
noise. A simple method for constructing the IV matrix is to
replace the true AOA measurement in with the approx-
where
imated AOA obtained from the BCAVPLE. Here the approxi-
mated AOA is given by (52a)
(44)
(52b)
where
and
(45a)
(52c)
(45b)
which is the estimation result of the AVPLE.
Remark 4: The BCAVPLE-WIV estimator is effective and
and
feasible when the prior knowledge of AOA noise variance
(45c) is available, while the AVPLE-WIV does not need the prior
knowledge of AOA noise variance. For the BCAVPLE-WIV
which is the estimation result of the BCAVPLE. estimator, the approximated AOA used in IV matrix is
Based on the estimated in (44), the IV matrix can be con- derived from the BCAVPLE, while for the AVPLE-WIV esti-
structed as mator, the approximated AOA used in IV matrix is de-
rived from the AVPLE. When the AOA noise is i.i.d., we can
use the AVPLE-WIV estimator to estimate the unknown node.
.. (46) However, while the AOA noise is not i.i.d., its performance may
. degrade and we should use BCAVPLE-WIV if possible since it
does not assume i.i.d.
The following Remark 5 gives the property of the
To reduce the estimation variance, a WIV estimator is pro-
BCAVPLE-WIV estimator and the AVPLE-WIV estimator:
posed to estimate the unknown node based on the IV estimator.
Remark 5: For sufficiently large , both the AVPLE-WIV
Under the assumption that the AOA noise variance is known,
estimator and the BCAVPLE-WIV estimator are asymptotically
the weighted matrix can be given by
unbiased estimators. Moreover, assume that AOA noise vari-
ance is i.i.d. with variance and the prior knowledge
(47)
of is known, for finite value of , the BCAVPLE-WIV
estimator has lower estimation error than the AVPLE-WIV
where is the estimated range from the observer to esti- estimator.
mated target location based on BCAVPLE. Using the sim- Proof: See Appendix B.
2588 IEEE TRANSACTIONS ON SIGNAL PROCESSING

The algorithmic procedures of both the BCAVPLE-WIV errors, and denotes the determinant of . Using the log-like-
estimator and the AVPLE-WIV estimator are summarized in lihood function to (53), we can obtain the ML cost function for
Algorithm 2. estimating the unknown node as

Algorithm 2: WIV estimator (BCAVPLE-WIV and

AVPLE-WIV)

1: Input: , , , and ,
2: If is known
• Compute ((35)) (54)
•
• , where is the number of beacons within the sensing range of
else the unknown node, is the location of the th beacon,
• Compute ((8)) is the measured AOA for the th beacon, and denote the
• location and orientation of the unknown node, respectively.
• , The location of the unknown node can be obtained by min-
imizing . The minimization of cost function involves a LS
End nonlinear problem, which does not have a closed-form solution.
3: Compute One can adopt search algorithms such as Gauss-Newton algo-
• , where rithm [28], Nelder-Mead simplex algorithm [29] or Newton-
• , Raphson [30] to seek the optimal solutions by setting a reason-
able initialization.
4: Compute
, , B. Cramér-Rao Lower Bound (CRLB) of ML Estimator
The CRLB is the theoretical lower bound of variance of an
5: Compute
unbiased parameter estimate [4], [31], [32]. For the AOA based
.. self-localization, the lower bound can be defined as the inverse
• , . of the following Fisher information matrix,

6: Compute
•
•
(55)
VI. MAXIMUM LIKELIHOOD ESTIMATOR AND CRAMÉR-RAO
LOWER BOUND where is the cost function of likelihood estimation in (54).
We briefly introduce the ML estimator for AOA based node Provided that AOA measurements are subjected to the
self-localization, and then derive its Cramér-Rao Lower Bound Gaussian white noise with variance . Let ,
(CRLB). and represents the Euclidean distance
from the unknown node to the th beacon, denoted by
A. Maximum Likelihood Estimator . Then the CRLB on the variance of the
The ML is an asymptotically unbiased estimator. To compare ML estimator is given by
the localization error with our new methods, the ML estimator is
also used as a benchmark. Under the assumption that the AOA
measurements uncertainty is a Gaussian distribution with zero
mean and known covariance , the likelihood function [27] is

(56)

(53) VII. PERFORMANCE EVALUATION

Matlab simulations are conducted to evaluate the perfor-
where is the vector of AOA mea- mance of our new algorithms. First, without loss of generality,
surements, is the vector the estimation errors of new algorithms, the triangulation
of AOAs as a function of the location and orientation with re- method and the ML estimator are compared in the original
spect to the unknown nodes, is the coordinate system. Next the impact of the prior knowledge
diagonal covariance matrix of the AOA measurement of the AOA noise variance on the BCAVPLE algorithm and
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2589

the BCAVPLE-WIV estimator are evaluated. Then the per-

formance of the new AVPLE-WIV estimator is illustrated
by comparing the estimation errors with other proposed new
algorithms. Finally, we illustrate the effectiveness of the ACR
algorithm to overcome the instability problem (tangent insta-
bility) of our new estimation algorithms when the orientation
of the unknown node is near .
Assume that there are no systematic biases for locations of
the beacons. Also assume that all the AOA measurements are
subject to i.i.d. additive Gaussian white noise with the noise
variance . Root mean square error (RMSE) and estimation
bias are used to evaluate the estimation accuracy, denoted as
and , respec-
tively, where is the estimated location, is the true location of
the unknown node and is the number of the simulation runs.
Simulation results are obtained with Monte Carlo
simulation runs. Fig. 3. Estimation errors of the proposed new algorithms including AVPLE,
BCAVPLE and BCAVPLE-WIV, in comparison with the triangulation and the
ML estimator .
A. Algorithm Comparisons Based on Accurate Prior
Knowledge of AOA Noise Variance
According to Remark 3 and Remark 4, the BCAVPLE and
the BCAVPLE-WIV estimator are feasible and effective with
the accurate prior knowledge of the AOA noise variance .
Thus, to compare our proposed algorithms with the triangula-
tion method and the ML estimator, the AOA noise variance
is supposed to be known in advance. We first consider a scenario
with ten beacons and one unknown node randomly deployed in
the 100 m 100 m region. Each trial is a new random realiza-
tion of the placements of the 10 beacons and one unknown node
based on a Monte Carlo method.
The ML estimator is implemented by the Nelder-Mead sim-
plex algorithm as in [29]. The iterative ML algorithm requires
a good initialization. The estimation result of the BCAVPLE
is selected as the initialization for the ML estimator to have
a fair comparison between the BCAVPLE-WIV estimator and
the ML estimator. Note that BCAVPLE-WIV does not need ini-
tialization, but it uses the estimation result of the BCAVPLE to Fig. 4. Estimation bias of the proposed new algorithms including AVPLE,
BCAVPLE and BCAVPLE-WIV, in comparison with the triangulation and the
obtain the approximated AOA used in IV matrix for the .
BCAVPLE-WIV estimator.
Fig. 3 illustrates the comparisons of estimation error
(i.e, RMSE) for the triangulation method, the AVPLE, the posed new algorithms have much lower estimation bias than the
BCAVPLE, the BCAVPLE-WIV and the ML estimator initial- triangulation method, while the BCAVPLE-WIV and the ML
ized to BCAVPLE. It can be seen that the AVPLE estimator estimator have nearly the same estimation bias.
has much lower estimation error than the triangulation method, To show the effectiveness of the presented new algorithms
which verifies the effectiveness and feasibility of our new aux- and demonstrate that the BCAVPLE-WIV estimator is an
iliary variables based algorithm. Moreover, the BCAVPLE has asymptotically unbiased estimator for large , we randomly
higher localization accuracy than the AVPLE as the AOA noise place 100 beacons and one unknown node in the 100 m
variance is known. In addition, the BCAVPLE-WIV estimator 100 m region. For each Monte Carlo trial, the locations of 100
has lower estimation error than the BCAVPLE and the AVPLE, beacons and the unknown node are randomly deployed in the
but slightly higher than the ML estimator. However, the ML area. We can see from Fig. 5 that the BCAVPLE has lower esti-
estimator is sensitive to initializations and may suffer from mation error than the AVPLE. In addition, the BCAVPLE-WIV
local convergence problem, which will be discussed in detail estimator has much lower estimation error than the AVPLE
in the following Section VII-D. and the BCAVPLE, which approaches the CRLB for large
To compare the actual estimation bias [19] of different algo- . Further, we illustrate the impact of the number of beacons
rithms, we conduct simulations with one unknown node, on the estimation errors of our new algorithms and the ML
, and 10 beacons randomly deployed in the 100 estimator. Fig. 6 shows that the estimation errors of all algo-
m 100 m region, as shown in Fig. 4. We can see that the pro- rithms gradually decrease with the increase of . Note that the
2590 IEEE TRANSACTIONS ON SIGNAL PROCESSING

Fig. 5. Comparisons of estimation errors for the AVPLE, the BCAVPLE, the Fig. 7. The impact of AOA noise variance on the BCAVPLE and the
BCAVPLE-WIV, the ML and the CRLB . BCAVPLE-WIV estimator .

than the AVPLE for node self-localization when the AOA noise
standard deviation is less than 5 degrees. When the AOA noise
standard deviation is beyond 4 degrees, the estimation error of
the BCAVPLE is lower than the AVPLE, but still higher than
that when the prior knowledge of is accurate as in Fig. 3.
Thus, it can be seen that the BCAVPLE is sensitive to the accu-
racy of the prior knowledge of the AOA noise variance . Note
that the BCAVPLE-WIV estimator and AVPLE-WIV estimator
have nearly the same estimation errors that are much lower than
the AVPLE. The BCAVPLE-WIV estimator is more robust than
the BCAVPLE.

C. Performance Evaluation for the AVPLE-WIV Estimator

In practice, the prior knowledge of the AOA noise vari-
ance is difficult to obtain owing to the complicated
Fig. 6. Comparisons of estimation errors for our proposed algorithms and the outdoor environment for node self-localization. Therefore, the
ML as increases . AVPLE-WIV estimator, which does not require prior knowl-
edge of , can be adopted to localize the unknown node.
For the simulations in this subsection, we randomly place
estimation errors of the AVPLE and the BCAVPLE decrease to beacons and one unknown node in the 100 m 100 m area.
a constant larger than the CRLB, while the estimation errors of Fig. 8 shows the comparisons of estimation errors for the
the BCAVPLE-WIV estimator and the ML estimator decrease AVPLE-WIV estimator, the AVPLE, the BCAVPLE and the
to approximate the CRLB as increases. BCAVPLE-WIV estimator. It can be seen that the AVPLE-WIV
estimator has lower estimation error than the AVPLE and the
B. Impact of the Prior Knowledge of the AOA Noise Variance BCAVPLE, and it has almost the same estimation error as the
on the BCAVPLE and the BCAVPLE-WIV BCAVPLE-WIV estimator. However, for a large and finite ,
the AVPLE-WIV estimator has slightly higher estimation error
The effectiveness of the BCAVPLE and the BCAVPLE-WIV than the BCAVPLE-WIV estimator as illustrated in Fig. 9. In
estimator are related to the prior knowledge of the AOA noise addition, we can see from Fig. 8 that the estimation errors of
variance . Assume that the predicted AOA noise standard de- the AVPLE-WIV and the BCAVPLE-WIV gradually decrease
viation for the BCAVPLE, denoted by , is four degrees worse and approach the CRLB as increases, verifying Remark 5.
than the actual , i.e., . In this simulation, with the
same self-localization scene as depicted in Section VII-A, ten
D. Algorithm Comparisons: AVPLE-WIV and BCAVPLE-WIV
beacons and one unknown node are randomly deployed in the
VS. ML Estimator
network area. Fig. 7 shows that when the noise standard devi-
ation is used by the BCAVPLE to compensate the estimate The ML estimator is sensitive to initializations and may suffer
bias of the AVPLE, the BCAVPLE has higher estimation error from local convergence problem. To illustrate the impact of dif-
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2591

Fig. 8. Comparisons of estimation errors for the AVPLE-WIV and other algo-
rithms as the number of beacons increases . Fig. 10. Comparisons of estimation errors of the AVPLE-WIV, the
BCAVPLE-WIV and the ML with different initializations .

Fig. 9. Comparisons of estimation errors for the BCLPLE-WIV and the

AVPLE-WIV estimator as the number of beacons increases .
Fig. 11. Impact of the orientation angle on estimation performance of our pro-
posed algorithms in the original coordinate system ( , ).

ferent initializations on the localization accuracy for the ML es-

timator, we conduct the simulation with ten beacons randomly tangent in (5) becomes unstable. In this paper, we further inves-
placed in the 100 m 100 m area, and the location information tigate the impact of the orientation on estimation performance
of unknown node is . In Fig. 10, ML1 and of the proposed algorithms. We randomly place 10 beacons
ML2 represent the ML estimator initialized to the BCAVPLE and one unknown node in the 100 m 100 m region based on
and other values, respectively. We can see that the ML estimator Monte Carlo method. Fig. 11 shows the localization accuracy
has different estimation biases under different initializations, of the proposed algorithms without coordinate rotation as the
while the BCAVPLE-WIV and AVPLE-WIV estimator is not orientation changes from 0 to 180 degrees. When the orienta-
sensitive to initialization. Sometimes, the ML estimator tends to tion is between 80 degrees and 100 degrees, the estimation error
diverge as the AOA noise increases as shown ML2 in Fig. 10, of AVPLE and BCAVPLE is pretty large, especially when is
while the BCAVPLE-WIV and the AVPLE-WIV remain stable. near 90 degrees, while the estimation error of the AVPLE-WIV
In addition, it should be pointed out that the BCAVPLE-WIV and the BCAVPLE-WIV approximately increases to 5 meters
algorithm can obtain the closed-form solutions that have much and 4 degrees when is near 90 degrees (i.e., ).
lower computational complexity than the multiple iterative ML Based on the ACR algorithm, we evaluate the performance of
estimator. our proposed algorithm again as shown in the following Fig. 12.
Fig. 12 shows that after coordinate rotation, the location error
E. Impact of the Orientation on Estimation Performance
and the orientation error of AVPLE have decreased from about
For our auxiliary variables based closed-form algorithms, 23 meters (in Fig. 11) to 2.2 meters and from 30 degrees to 2.3
when the orientation of the unknown node is near , the degrees, respectively, when the orientation is near 90 degrees,
2592 IEEE TRANSACTIONS ON SIGNAL PROCESSING

and can avoid the local minima and divergence problems in

iterative ML estimators. The AVPLE has lower localization
accuracy than AVPLE-WIV, BCAVPLE, and BCAVPLE-WIV
when the prior knowledge of the AOA noise variance is known.
Note that the effectiveness of BCAVPLE and BCAVPLE-WIV
estimator depends on the accurate prior knowledge of the
AOA noise variance. The BCAVPLE-WIV estimator and the
AVPLE-WIV estimator are both unbiased estimators for large
, but BCAVPLE-WIV estimator has lower estimation error
than AVPLE-WIV estimator for finite value of . However,
the AVPLE-WIV estimator is more robust when we have no
prior knowledge of the AOA noise variance and the AOA noise
is i.i.d. In addition, when the orientation is near 90 degrees,
the proposed autonomous coordinate rotation (ACR) method
could solve the tangent instability problem of the presented
new algorithms effectively.

APPENDIX A
Fig. 12. Impact of the orientation angle on estimation performance of our pro- MATHEMATICAL DERIVATION OF EQUATION (3)
posed algorithms with autonomous coordinate rotation (ACR) ( ,
). Referring to (2), we have

(57)
and those of BCAVPLE have decreased from 13 meters to 2
meters and from 16.2 degrees to 2 degrees, while the estimation According to a trigonometric identity, (57) becomes
errors of the AVPLE-WIV and the BCAVPLE-WIV are approx-
(58)
imately the same as that of ML estimator. Therefore, our pro-
posed ACR algorithm is effective to solve the tangent instability
problem. Rewrite (58) as

(59)
VIII. CONCLUSION
In this paper, we present a set of novel auxiliary variables Rearrange the aforementioned formulation (59),
based close-form algorithms for the efficient AOA based
self-localization. First we obtain the closed-form AVPLE
solution for AOA based self-localization problem based on (60)
auxiliary variables. Based on the analysis of the theoretical es-
timation error of auxiliary variables for the AVPLE, we further Pre-multiply both sides of (60) with , we obtain
develop a bias compensation scheme for AVPLE to reduce the
estimation error of node self-localization, namely BCAVPLE
method. When the prior knowledge of the AOA noise variance (61)
is known, we present a BCAVPLE-WIV estimator to reach the
approximate CRLB accuracy for node self-localization. In the APPENDIX B
case that the AOA noise variance is unknown, a AVPLE-WIV PROOF OF REMARK 5
estimator is presented to localize the unknown node. Compared Similar to the prior work in [33], the correlation between
with the triangulation method, our proposed auxiliary variables the and the noise vector vanishes as the number of
based algorithms can improve the localization accuracy signif- beacons . Thus the approximated AOA obtained from
icantly with lower complexity. The simulation results verify the AVPLE, denoted by , is asymptotically uncorrelated
that the BCAVPLE reduces the estimation error compared with . As a result, the elements of the IV matrix for the
with the AVPLE, and for large number of beacons, , the AVPLE-WIV estimator in (51) are independent of the elements
BCAVPLE-WIV estimator can approximately reach the CRLB. of the matrix for sufficient large . Moreover, the elements
We also show that when the AOA noise variance is unknown, of are uniformly bounded. Based on the Theorem 2 in [22],
the AVPLE-WIV estimator has lower estimation error than the we have , where “ ” denotes the
AVPLE and the BCAVPLE, which approaches the CRLB for probability limit [22].
the large . However, compared with the BCAVPLE-WIV In addition, if the unknown node is observable,
estimator, the AVPLE-WIV estimator has higher estimation is always nonsingular, thus
error for a finite value of . is nonsingular as well. Therefore,
In summary, the new AOA based self-localization algo- according to the two sufficient and necessary conditions
rithms, AVPLE, BCAVPLE, AVPLE-WIV and BCAVPLE- in [22], we can conclude that the AVPLE-WIV estimator
WIV, have higher localization accuracy than the triangulation is asymptotically unbiased. Using the same methodology,
SHAO et al.: EFFICIENT CLOSED-FORM ALGORITHMS 2593

we can also justify that the BCAVPLE-WIV estimator is [12] P. Rong and M. Sichitiu, “Angle of arrival localization for wireless
asymptotically unbiased. sensor networks,” in Proc. 3rd Ann. IEEE Commun. Soc. Sens. Ad Hoc
Commun. Netw. (SECON’06), 2006, vol. 1, pp. 374–382.
Next, we will verify that the BCAVPLE-WIV estimator has [13] V. Cevher and J. H. McClellan, “Acoustic node calibration using a
lower estimation error than the AVPLE-WIV estimator for finite moving source,” IEEE Trans. Aerosp. Electron. Syst., vol. 42, no. 2,
value of when the AOA noise variance is known. Firstly, pp. 585–600, 2006.
[14] R. Moses, D. Krishnamurthy, and R. Patterson, “A self-localization
since the BCAVPLE estimator has higher localization accuracy method for wireless sensor networks,” EURASIP J. Appl. Signal
than the AVPLE when the prior knowledge of is known, Process., pp. 348–358, 2003.
the approximated AOA used in IV matrix obtained from the [15] H. Xue, H. Shao, and Z. Wang et al., “A self-localization mechanism
for acoustic array network based on target assistance under the limited
BCAVPLE, denoted by , is more accurate than obtained beacons,” J. Software, vol. 22, no. 1, pp. 40–50, 2011.
from the AVPLE, i.e., [16] L. Yang and K. Ho, “Alleviating sensor position error in source local-
ization using calibration emitters at inaccurate locations,” IEEE Trans.
Signal Process., vol. 58, no. 1, pp. 67–83, 2010.
(62) [17] K. Doğançay, “On the efficiency of a bearings-only instrumental vari-
able estimator for target motion analysis,” Signal Process., vol. 85, no.
3, pp. 481–490, 2005.
where is the true AOA measurement. [18] P. Biswas, T. Lian, T. Wang, and Y. Ye, “Semidefinite programming
According to (62) and (46), for finite value of , we can based algorithms for sensor network localization,” ACM Trans. Sens.
Netw., vol. 2, no. 2, pp. 188–220, 2006.
see that the IV matrix for BCAVPLE-WIV estimator, denoted [19] K. Doğançay, “Bias compensation for the bearings-only pseudolinear
by , is closer to the optimal IV matrix than that of the target track estimator,” IEEE Trans. Signal Process., vol. 54, no. 1, pp.
AVPLE-WIV. When the AOA noise subjects to i.i.d., the 59–68, 2006.
[20] K. Doğançay, “On the bias of linear least squares algorithms for passive
BCAVPLE-WIV and the AVPLE-WIV have the same formu- target localization,” Signal Process., vol. 84, no. 3, pp. 475–486, 2004.
lation as shown in (48) and (50), except for the different IV [21] P. Stoica and A. Nehorai, “Music, maximum likelihood, Cramer-Rao
matrix. Generally, the more closely the estimated IV matrix Bound,” IEEE Trans. Acoust., Speech Signal Process., vol. 37, no. 5,
pp. 720–741, 1989.
approaches the optimal IV matrix , the higher is the local- [22] K. Wong and E. Polak, “Identification of linear discrete time systems
ization accuracy of the estimator. Therefore, we can conclude using the instrumental variable method,” IEEE Trans. Autom. Control,
that the BCAVPLE-WIV estimator has lower estimation error vol. 12, no. 6, pp. 707–718, 1967.
[23] A. Lingren and K. F. Gong, “Position and velocity estimation via
than the AVPLE-WIV for finite value of . bearing observations,” IEEE Trans. Aerosp. Electron. Syst., no. 4, pp.
564–577, 1978.
[24] S. Nardone, A. Lindgren, and K. Gong, “Fundamental properties and
performance of conventional bearings-only target motion analysis,”
IEEE Trans. Autom. Control, vol. 29, no. 9, pp. 775–787, 1984.
REFERENCES [25] V. J. Aidala and S. C. Nardone, “Biased estimation properties of the
pseudolinear tracking filter,” IEEE Trans. Aerosp. Electron. Syst., no.
[1] Z. Wang, J.-A. Luo, and X.-P. Zhang, “A novel location penalized max- 4, pp. 432–441, 1982.
imum likelihood estimator for bearing only target localization,” IEEE [26] Y. Chan and S. Rudnicki, “Bearings-only and doppler-bearing tracking
Trans. Signal Process., vol. 60, no. 12, pp. 6166–6181, 2012. using instrumental variables,” IEEE Trans. Aerosp. Electron. Syst., vol.
[2] K. Doğançay, “Relationship between geometric translations and tls es- 28, no. 4, pp. 1076–1083, 1992.
timation bias in bearings-only target localization,” IEEE Trans. Signal [27] K. Doğançay, “Bearings-only target localization using total least
Process., vol. 56, no. 3, pp. 1005–1017, 2008. squares,” Signal Process., vol. 85, no. 9, pp. 1695–1710, 2005.
[3] Y. Liu, Y. Hu, and Q. Pan, “Distributed, robust acoustic source local- [28] L. Ljung and T. Söderström, Theory and practice of recursive identifi-
ization in a wireless sensor network,” IEEE Trans. Signal Process., vol. cation. Cambridge, MA: MIT Press, 1983.
60, no. 8, pp. 4350–4359, 2012. [29] J. C. Lagarias, J. A. Reeds, M. H. Wright, and P. E. Wright, “Con-
[4] X. Sheng and Y. Hu, “Maximum likelihood multiple-source local- vergence properties of the nelder-mead simplex method in low dimen-
ization using acoustic energy measurements with wireless sensor sions,” SIAM J. Optimiz., vol. 9, no. 1, pp. 112–147, 1998.
networks,” IEEE Trans. Signal Process., vol. 53, no. 1, pp. 44–53, [30] R. Jennrich and P. Sampson, “Newton-raphson and related algo-
2005. rithms for maximum likelihood variance component estimation,”
[5] A. Ali, S. Asgari, T. Collier, M. Allen, L. Girod, R. Hudson, K. Yao, Technometr., vol. 18, no. 1, pp. 11–17, 1976.
C. Taylor, and D. Blumstein, “An empirical study of collaborative [31] D. Torrieri, “Statistical theory of passive location systems,” IEEE
acoustic source localization,” J. Signal Process. Syst., vol. 57, no. 3, Trans. Aerosp. Electron. Syst., no. 2, pp. 183–198, 1984.
pp. 415–436, 2009. [32] J. Chen, K. Yao, and R. Hudson, “Source localization and beam-
[6] M. Sun and K. Ho, “Successive and asymptotically efficient localiza- forming,” IEEE Signal Process. Mag., vol. 19, no. 2, pp. 30–39, 2002.
tion of sensor nodes in closed-form,” IEEE Trans. Signal Process., vol. [33] K. Doğançay, “Passive emitter localization using weighted instru-
57, no. 11, pp. 4522–4537, 2009. mental variables,” Signal Process., vol. 84, no. 3, pp. 487–497, 2004.
[7] Z. Zhong and T. He, “Wireless sensor node localization by multise-
quence processing,” ACM Trans. Embedded Comput. Syst., vol. 11, no.
1, p. 3, 2012.
[8] R. Stoleru, T. He, S. Mathiharan, S. George, and J. Stankovic, “Asym-
metric event-driven node localization in wireless sensor networks,”
IEEE Trans. Parallel Distrib. Syst., vol. 23, no. 4, pp. 634–642, 2012.
[9] L. Yang and K. Ho, “An approximately efficient tdoa localization algo-
rithm in closed-form for locating multiple disjoint sources with erro- Hua-Jie Shao is currently a graduate student in De-
neous sensor positions,” IEEE Trans. Signal Process., vol. 57, no. 12, partment of Control science and Engineering at Zhe-
pp. 4598–4615, 2009. jiang University. He received his B.E. degree in Au-
[10] D. Niculescu and B. Nath, “Ad hoc positioning system (aps) using tomation from Jiangnan University, China, in 2011.
aoa,” in Proc. INFOCOM 2003, 22nd Ann. Joint Conf. IEEE Comput. His research interests include Internet Data Cen-
Commun., 2003, vol. 3, pp. 1734–1743. ters under Smart Grid and Signal Processing and
[11] P. Biswas, H. Aghajan, and Y. Ye, “Integration of angle of arrival infor- Wireless Sensor Networks.
mation for multimodal sensor network localization using semidefinite
programming,” in Proc. 39th Asilomar Conf. Signals, Syst. Comput.,
2005.
2594 IEEE TRANSACTIONS ON SIGNAL PROCESSING

Xiao-Ping Zhang (M’97–SM’02) received B.S. and and the International Journal of Semantic Computing. He is a tutorial speaker
Ph.D. degrees from Tsinghua University, in 1992 and in ACMMM2011, ISCAS2013, ICIP2013 and ICASSP2014. He is currently
1996, respectively, both in Electronic Engineering. an Associate Editor for IEEE TRANSACTIONS ON SIGNAL PROCESSING, IEEE
He holds an MBA in Finance, Economics and En- TRANSACTIONS ON MULTIMEDIA, IEEE SIGNAL PROCESSING LETTERS and for
trepreneurship with Honors from the University of Journal of Multimedia.
Chicago Booth School of Business, Chicago, IL.
Since fall 2000, he has been with the Department
of Electrical and Computer Engineering, Ryerson
University, where he is now Professor, Director of Zhi Wang (M’08) received the B.E., M.S., and Ph.D.
Communication and Signal Processing Applications degrees from Shenyang Jianzhu University in 1991,
Laboratory (CASPAL). He has served as Program Southeast University in 1997 and Shenyang Institute
Director of Graduate Studies. He is cross appointed to the Finance Department of Automation, CAS in 2000, respectively.
at the Ted Rogers School of Management at Ryerson University. Prior to During 2000–2002, as Post-doc, he has con-
joining Ryerson, he was a Senior DSP Engineer at SAM Technology, Inc., San ducted research in Institut National Polytechique
Francisco, and a consultant at San Francisco Brain Research Institute. He held de Lorraine, France and Zhejiang University, China
research and teaching positions at the Communication Research Laboratory, respectively. His main research focus is on the
McMaster University, and worked as a postdoctoral fellow at the Beckman cyber physical system, including sensor network,
Institute, the University of Illinois at Urbana-Champaign, and the University of collaborative signal and information processing,
Texas, San Antonio. His research interests include statistical signal processing, real-time theory and system, integrated schedule
multimedia retrieval and video content analysis, sensor networks and elec- and control, secure and privacy protection, and related platforms, such as
tronic systems, computational intelligence, and applications in bioinformatics, networked acoustic arrays, networked visual and optical arrays, and industrial
finance, and marketing. He is a frequent consultant for biotech companies and communication networks.
investment firms. He is cofounder and CEO for EidoSearch, an Ontario-based Dr. Wang serves as General Co-Chair and TPC member for a number of
company offering a content-based search and analysis engine for financial data. CPS and sensor networks related international conferences, and the committee
Dr. Zhang is a registered Professional Engineer in Ontario, Canada, a member for China Computer Federation Sensor Network Technical Committee
Senior Member of IEEE and a member of Beta Gamma Sigma Honor Society. and China National Technical Committee of Sensor Network Standardization.
He is the publicity chair for ICME’06 and program chair for ICIC’05 and He is a member of ACM.
ICIC’10. He served as guest editor for Multimedia Tools and Applications,