See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/274704302

Angle of Arrival Measurement Using Multiple Static Monopole Antennas

Article  in  IEEE Sensors Journal · December 2014

DOI: 10.1109/JSEN.2014.2386537

CITATIONS READS
12 196

3 authors, including:

Marko Malajner
University of Maribor
23 PUBLICATIONS   303 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Marko Malajner on 09 April 2015.

The user has requested enhancement of the downloaded file.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
1

Angle of arrival measurement using multiple static

monopole antennas
Marko Malajner, Dušan Gleich, Peter Planinšič,

Abstract—The Angle of Arrival (AoA) is an important factor modules support the measuring of RSSI, which enables the
in the localization of a wireless sensor network. This paper deals calculations of received power for each received packet. The
with AoA measurement using omnidirectional antennas. In our power or energy distribution of a RF signal traveling between
case microstrip monopole antennas are used which have radiation
patterns with two sharp minimums. Therefore, an algorithm two nodes is a signal parameter which can be used for distance
based on an approach where an AoA is obtained along a direction estimation, depending on the path-loss and shadowing effects.
where the measured received strength signal indicator (RSSI) is RSSI measurements are very unpredictable because there are
minimal. Multiple stationary microstrip antennas are placed on several disturbing sources like the many delayed multipath
a printed circuit–board and no moving parts are needed for signals arriving at the receiver [4], [5]. AoA or directions to
measurement. In comparison with rotating antennas this simple
method has lower resolution. We show that the resolution can be neighboring sensors can be estimated from ToF and/or RSSI.
improved using interpolations and approximations. When using There are two common ways that sensors measure the AoA,
the proposed algorithm, the experimental results for the outdoor i) multiple static smart antennas or arrays and ii) rotatable
measurements reached a root-mean-square error (RMSE) of less antennas.
than 10◦ , and an indoor environment of less than 25◦ . Many localization systems use multiple static antennas or
Index Terms—Angle of Arrival (AoA), Omnidirectional An- arrays for AoA measurements. Usually they are called smart
tenna, Received Signal Strength Indicator (RSSI), Wireless Sen- antennas because they use signal processing units for AoA
sor Network (WSN).
measurements. The most common method is to use antenna ar-
ray with known array geometry, and measuring the differences
I. I NTRODUCTION
of signal arrival times at different receivers [6], [7]. Authors

L OCALIZATION is a crucial mechanism for obtaining the

locations of data source within wireless sensor networks
(WSN). WSN consists of small so-called sensors equipped
in [8] estimated AoA at each receiving node via frequency
measurement of the local RSSI. Another method uses antenna
arrays and RSSI measurements on each antenna and from the
with environmental sensing devices, power sources, radio and RSSI ratio the AoA can estimated [9], [10], [11]. On the
processor units. Sensors can communicate with each other or other hand, AoA systems can use beamforming [12], [13].
with the base–station. In many cases the sensors are randomly Beamforming on static antennas’ systems can shape radiation
deployed on fields where locations’ information does not exist pattern without the physical shaping of antennas. This can be
in advance. Some localization systems must be performed on done by switching each individual antenna ”on” or ”off” with
sensors’ fields for locating each sensor. various algorithms. Adaptive smart antennas are the second
Two physical values are primarily measured, separately or category of antennas in which adaptive beam is formed based
in combination, when positioning RF equipped sensor nodes: on the direction of the desired and interfering sources [13]. In
(i) time of RF signal flight (ToF) and (ii) the power of the [14] the authors proposed AoA estimation based on subarray
received RF signal. The distances or angles between the nodes beamforming. The target AoA is estimated from the phase
are then calculated from these primary values for determining shift introduced in the target signal by subarray beamforming,
the locations of each node. Three main localization techniques which is a function of the target AoA. In literature many
are known in WSN: (i)ToF – Time of Flight, (ii) RSSI – algorithm can be found for estimating AoA from antenna
Received Signal Strength Indicator and (iii) AoA – Angle arrays or smart antenna, like MUSIC [15], root MUSIC [16],
of Arrival [1]. Measuring the ToF is costly because the RF maximum liklehood estimator [17], ESPIRIT [18], etc.
signal propagation has the speed of light and therefore the Many AoA techniques use rotatable antennas with shaped
nodes must have a common accurate clock, or exchange timing radiation patterns in one direction and then seek the maximal
information using certain protocols as, for example, a two- power of the received signals around the axis [19], [20].
way ranging protocol [2], [3]. Most 802.11 and 802.15.4 radio This paper presents our proposed method of AoA estimation
M. Malajner is with the Faculty of Electrical Engineering and Computer using a system of multiple omnidirectional monopole static
Science, University of Maribor, Smetanova ul. 17, 2000 Maribor, Slovenia antennas placed on the circumference of a circular plate. The
e-mail: [email protected] radiation shape of the used monopole antenna is not ideally
D. Gleich is with the Faculty of Electrical Engineering and Computer
Science, University of Maribor, Smetanova ul. 17, 2000 Maribor, Slovenia isotropic and has a tiny minimum in the direction of the
P. Planinsic is with the Faculty of Electrical Engineering and Computer antenna axis. The idea presented in this paper is to find from
Science, University of Maribor, Smetanova ul. 17, 2000 Maribor, Slovenia which direction the minimal RSSI is measured. Searching for
Copyright (c) 2013 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be the minimum was chosen because better selectivity is obtained
obtained from the IEEE by sending a request to [email protected]. due to the tiny sharp minimum within the radiation pattern

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
2

of the antenna. Instead of rotatable monopoles [21], here

we use twelve static monopoles placed on a circular plate.
The advantage of a static plate is that all movable parts are
eliminated. We will also show that the static system has a
slightly lower resolution, but can be improved by interpolation.
The rest of the paper is organized as follow: Section II
provides the theoretical background, Section III presents the
used hardware, Section IV presents the simple estimation
algorithm, Section V shows the experimental results using a
simple algorithm, Section VI introduces the improved algo-
rithm, with the experimental results, Section VII compares the
used algorithms, and Section VIII concludes the paper.

II. T HEORETICAL BACKGROUND OF APPROACH

The dipole antenna is considered as an omnidirectional
antenna. The ideal dipole has a spherical radiation pattern, Fig. 1. E-plane of used antenna.
which radiates isotropically. A real dipole has some degree of
directivity and radiates weaker in the direction of the antenna
therefore
axis. The performance of a linearly-polarized antenna, which is 3 cos θ sin2 θ
almost an approximation of an ideal antenna, is described with √ =0 (5)
E– and H–plane patterns [22]. The H [A/m]– and E [V/m]– 2 sin3 θ
planes are perpendicular to each other. The cross–product of and thus
E and H* (* denotes complex conjugation) divided by 2, gives θ = nπ , n ∈ Z (6)
the time-average Poynting vector of the radiated field. We can see that the minimums of the simplified radiation
The quarter-wavelength monopole, as was used in our pattern of dipole appear every 180◦ . Real dipole has little
experiment, can be represented as a half-wavelength dipole, difference between two minimums, as is shown in Fig. 1. This
as explained in [22, p. 2-17]. For the half-wavelength dipole, difference within used the dipoles is about a few dBs and can
the E– and H–planes can be written as [23, p. 182]: be detected. Therefore, the minimums of real dipole appear
every θ = 2π.
"
#
I0 e−jkr cos π2 cos θ
Eθ ' jη (1) In our previous experiment presented in [4] we physically
2πr sin θ
rotated one or multiple omnidirectional antennas placed on the
circular plate within an E-plane, in order to find the minimal
"
#
Eθ I0 e−jkr cos π2 cos θ
Hφ ' 'j (2) measured value of RSSI for estimating AoA. This paper
η 2πr sin θ
presents a static system for AoA estimation which eliminates
where η is the intrinsic impedance of the medium, r is the all mechanical moving parts. In other words, this AoA system
radius distance, θ is the azimuth angle, and φ is the latitude is static. In order to achieve similar effects to those of the
angle. I0 is a traveling wave current. The 2–D omnidirectional rotatable antenna, the multiple antennas were attached to a
radiation patterns are presented in Fig.1 obtained by Ansoft static circular plate with a diameter of 12 cm.
HFSS [24]. It is the simulation of used antenna in our Fig. 3 shows radiation pattern of single antenna, obtained
experiments. by equation (1). Fig. 4 shows radiation patterns of 4 shifted
It can be observed from the radiation patterns in Fig.1 that antennas, as are placed on our plate. We shows on Fig. 4 only
the patterns have all the characteristics of omnidirectional four antennas instead of 12 because of better representation.
monopoles or dipoles, respectively. The right side in Fig. 1
shows the orientation of the used antenna in respect to the III. H ARDWARE SET- UP
radiation pattern.
The device for the proposed AoA measurement in our
The radiation pattern in Fig.1 consists of two zones of
case consists of twelve transceivers placed on a circular
decreased intensities at 90 and -90 degrees. The RF signal,
plate (Fig.5). The plate with the transceivers was attached
which arrives from the transmitter in the direction of a
to our SPaRCMosquito WSN module with a Cortex M3
radiation pattern with less sensibility, causes lower RSSI. Let
NXP microcontroller, integrated MRF24J40MA transceiver,
us take equation (1) there it can be shown that:
and a battery-powered source, as shown in Fig.6b. Fig.6c
"
#2
cos π2 cos θ also shows another SPaRCMosquito module which serves as a
3
sin θ ' (3) transmitter. The transceivers are the Microchip MRF24J40MA
sin θ
[25] with microstrip monopole antenna operating within the
In order to extract angle θ, lets differentiate the equation: ISM 2.4 GHz band. The monopole antenna uses ground plane
dp 3 of MRF24J40MA (see Fig. 1) module and ground plane of
sin θ = 0 (4) circular plate as a counterpoise ground plane. An additional
dθ

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
3

y
Radiation patterns
E-planes

Rx 1
Rx 12 Rx 2

Rx 11 Rx 3

Rx 10 Rx 4
x
Rx 9
Rx 5
Rx 8
Rx 7 Rx 6

AoA

Tx

Fig. 5. The AoA measuring device with 12 MRF24J40MA modules.

Fig. 2. Principle scheme of static AoA measuring system with multiple

antennas.

10
Amplitude of electric field
0

−10

−20
c) a)
b)
E (dBm)

−30

−40

−50

−60

−70
Fig. 6. Static AoA measuring system. a) Plate with transceivers. b) Main
−80 measurement control device – SPaRCMosquito. c) SPaRCMosquito as a
0 50 100 150 200 250 300 350
Angle (°)
transmitter.

Fig. 3. Radiation pattern of single antenna ground plane on our circular plate enhances the performance
Antennas number
of the module [25] and does not reshape the radiation pattern.
1 2 3 4 5 6 7 8 9 10 11 12
10

0
The transmitter continually transmits RF signals during the
−10
experiments. The SPaRCMosquito with receivers on a plate
collects the RSSI values from each receiver. Approximately
−20
40 packets of RSSI data are collected and averaged. Each
−30
transmitted packet was received on all receivers simultane-
RSSI (dBm)

−40 ously. The averaged RSSI data were then sent to the laptop
−50 for further processing.
min. at antenna 3 (AoA=270°−180°)
−60
min. at antenna 6 (AoA=360°−180°)
min. at antenna 8 (AoA=240°−180°)
−70 min. at antenna 11 (AoA=330°−180°) A. Preliminary measurements
−80 Reference measurements of RSSI from different directions
50 100 150 200
Angle [°]
250 300 350 were done in order to prove the simulation from Fig. 4. Fig. 7
shows the RSSI values from four directions of preliminary
measurements with the transmitter at a distance of 1 m.
Fig. 4. Simulation of radiation pattern for static measuring device.
Transmitter were placed in front of antennas 3, 6, 8 and 11.
We can observe typical simulation–predicted minimas in the
directions of antennas 9, 12, 2 and 5. Opposite minimas are at

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
4

Antennas number
1 2 3 4 5 6 7 8 9 10 11 12
−40

−45

Start of algorithm
−50
End of online and
start of offline
−55
algorithm
Transmitter (Fig. 4c)
RSSI (dBm)

−60
sends data packet
Simple algorithm
−65

−70 Receiving device Improved algorithm

(Fig. 4a) receives
−75
packet in all
min. at antenna 9 (true AoA=270°−180°) Searching which Interpolation of
antennas
min. at antenna 12 (true AoA=360°−180°) antenna received measured RSSI
−80 min. at antenna 2 (true AoA=240°−180°)
minimum RSSI points
min. at antenna 5 (true AoA=330°−180°)

−85
50 100 150 200 250 300 350 40 Received
Angle [°] AoA 180° is in
packets? Searching minimal
direction of antenna
interpolated RSSI
with minimum RSSI

Fig. 7. Preliminary measurements of static AoA estimation system with

Board (Fig. 4b)
multiple antennas. average RSSI and AoA 180° is in
send them to PC End direction of antenna
together with with minimum RSSI
receiver‘s number
antennas 3, 6, 8 and 11 and they are smaller at about 20 dBm.
End
It is because of thus property of the measurement device that
it is possible to obtain the right AoA. Right AoA’s are on
the opposite side of observed minimums. ±180◦ is added to
angle where is measured minimum in order to obtain true
Fig. 8. Algorithm for AoA estimation.
AoA. There are some inequalities between minimas because
the ground plane of measurement system is not equally placed
behind each antenna and the transmitters are not 100% equal writes the averaged RSSI of each receiver in the file, including
each other. the receiver’s number. This file serves as input for the Matlab
algorithm for AoA estimation which analyzes the RSSI data.
IV. S IMPLE AOA ESTIMATION ALGORITHM WITH Basically, the simple algorithm searches for which receiver
MULTIPLE ANTENNAS minimum of RSSI data is measured. The true AoA is in the
The radiation pattern monopole is well-known and can be opposite direction (±180◦ ) of minima. By using the proposed
calculated from equations (1) and (2). Twelve static dipole method, the resolution of AoA is 360◦ divided by the number
patterns arranged around the circle can be imagined using the of receivers, in our case 30◦ , because the antennas are equally
proposed antenna arrangement. Fig. 2 shows only 4 of the 12 arranged around the circle. Fig. 8 shows the flow–chart of the
antennas patterns (E–planes) because of transparency, where algorithm.
the E–plane patterns are calculated by equation 1. The E–plane
for a single antenna is shown in Fig. 3.
V. E XPERIMENTAL ESTIMATIONS OF AOA USING A SIMPLE
Fig. 4 presents radiation patterns for four antennas within
ALGORITHM
the cartesian coordinate system, instead of the polar. Each
minima presents the decreased zone of the radiation pattern All experiments were carried-out within a real environ-
caused by a single antenna. We used this specific feature for ments, outdoor and indoor. Measurements were limited to
AoA estimation. line of sight (LOS) between transmitter and receivers. Moving
The simple algorithm for gathering the RSSI from each and static obstacles between the transmitters caused significant
receiver is as follows: firstly, the transmitter sends one packet fluctuations in the measurements of the received power. Each
with dummy data. All the transceivers on the circular plate experimental set-up was repeated three times and the results
are in receiving mode and waiting for the incoming packet. averaged.
When the receivers receive a packet, they calculate the RSSI
from the voltage of the incoming RF signal. One of receivers
is connected to a microcontroller interrupt line. When the A. Outdoor measurements
receiver sends an interrupt to the microcontroller, then the The outdoor experiments were carried-out on an asphalt
reading phases starts. The microcontroller via SPI reads the floor with line-of-sight between transceivers. The transceivers
RSSI data from each receiver in sequence. When the reading is were placed on a rack at 1 m heights from the ground. The
done, the transmitter sends a new data packet. RSSI measuring transmitter was placed at distances from 1 m to 30 m, at steps
is repeated till 40 packets are processed and averaged. The on– of 1 m. Fig. 9 shows the measured RSSI points versus angle at
line phase of the algorithm is finished with the last operation. a distance between the transmitter and the receiver of 1 m. The
The off–line part of the algorithm is performed on a laptop. true azimuth AoA was set at 90◦ . The algorithm just searched
Special software was developed but only for gathering data the minimal RSSI and added +180◦ or −180◦ (depends from
from the measuring device via USB to the laptop. The software angle of minima), which was in this case found at 90◦ i.e. at

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
5

Antennas number
1 2 3 4 5 6 7 8 9 10 11 12
−50
RSSI
estimated AoA−180°
true AoA
−55

−60
Measured RSSI [dBm]

−65

−70

−75

−80
50 100 150 200 250 300 350
Angle [°]

Fig. 9. Outdoor measurements of RSSI versus angle at a distance of 1 m. Fig. 11. Indoor measurements test environment.

Antennas number
Antennas number
1 2 3 4 5 6 7 8 9 10 11 12
−50 1 2 3 4 5 6 7 8 9 10 11 12
−50
RSSI
estimated AoA−180°
−55 true AoA

RSSI
−60 estimated AoA−180°
Measured RSSI [dBm]

−55

Measured RSSI [dBm]

true AoA

−65

−70
−60

−75

−80
50 100 150 200 250 300 350 −65
50 100 150 200 250 300 350
Angle [°]
Angle [°]

Fig. 10. Outdoor measurements of RSSI versus angle at a distance of 3 m Fig. 12. Indoor measurements of RSSI versus angle at 1 m
and true AoA at 105◦ .

the 3rd antenna. The estimated AoA completely matched the However, despite small differences between the minimas, the
true AoA, set in advance. AoA estimation was very close to the true value. In the case
In the second case the true AoA was set between antennas depicted in Fig. 12, the algorithm found smaller RSSI at the
3 and 4. The RSSI measurements are shown in Fig. 10. The 9th antenna (270◦ ), the true AoA was set at 270◦ − 180◦ .
algorithm estimated AoA was again at 270◦ − 180◦ , and the Fig. 13 shows the case where the true AoA was set at
true AoA was set at 105◦ . The Estimation error in this case 105◦ and the distance between the transceivers was 3 m. Two
was about 15◦ and was a consequence of the fixed resolution major minimas were observed at 90◦ and 300◦ . The algorithm
of this method, in our case 30◦ , as mentioned in the previous estimated AoA at 300◦ where the second smallest RSSI was
section. measured. The greater errors during indoor estimation were
the consequences of unconsidered reflective signals from the
obstacles. Due to the disturbing reflections from the obstacles,
B. Indoor measurementes the estimation errors within the indoor environment were, in
The indoor experiments were carried-out in a 6 m-by-8 m general, greater than a resolution of 30◦ .
room and in the underground garage (Fig. 11). The transceivers
were placed at 1 m heights. The experiments were carried-out
in the same way as the outdoor experiments. Fig. 12 shows VI. I MPROVED AOA ESTIMATION ALGORITHM
the measured RSSI versus azimuth angle. The measurements
were done at a distance of 1 m and an angle of 90◦ . The indoor The interpolations and approximations between RSSI points
measurements were influenced by the reflected signals from were realized in order to improve the resolution and accuracy
the obstacles in the room. Due to reflected signals, the RSSI of the simple algorithm of the AoA method, and then the
measurements contained more local minimums and maximums angles with minimum interpolated RSSI-values were searched
and it was more difficult to accurately estimate the AoA than for. Different interpolations and approximations were used
during outdoor measurements. The differences between local and compared. In this section all the experiments were taken
minimas were smaller than for the outdoor measurements. outdoors at distances of 2 meters.

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
6

Antennas number Antennas number

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
−45 −55
RSSI
estimated AoA−180°
true AoA
−50
−60

−55

Measured RSSI [dBm]

Measured RSSI [dBm]

−65
−60

−65 Spline interpolation

−70
Estimated AoA−180°
Linear interpolation
−70 Measured RSSI points

−75
−75

−80 −80
50 100 150 200 250 300 350 50 100 150 200 250 300 350
Angle [°] Angle [°]

Fig. 13. Indoor measurements of RSSI versus angle at 3 m. True AoA was Fig. 15. Outdoor measurement using linear and spline interpolation at true
at 105◦ AoA at 105◦ (distance 2 m).

Antennas number
Antennas number
1 2 3 4 5 6 7 8 9 10 11 12
−55 1 2 3 4 5 6 7 8 9 10 11 12
−55

−60
−60
Measured RSSI [dBm]

Measured RSSI [dBm]

−65
−65

−70 Spline interpolation

−70
Estimated AoA−180° Estimated AoA−180
Linear interpolation Measured RSSI points
Measured RSSI points Polynomial interpolation
−75
−75

−80
50 100 150 200 250 300 350 −80
Angle [°] 50 100 150 200 250 300 350
Angle [°]

Fig. 14. Outdoor measurement using linear and spline interpolation (distance
Fig. 16. Outdoor measurement using polynomial approximation using 9th
2 m).
degree polynomial (distance 2 m).

A. Linear and spline interpolations

degree of the polynomial, then interpolation is obtained and
The simplest interpolation is linear interpolation which the polynomial goes through all the data points. The Gaussian
connects points with lines. However, this interpolation does process is a powerful non-linear interpolation tool. In addition
not improve the accuracy of AoA estimation by searching the the Gaussian function, can not only be used for fitting an
minimum. The next interpolation is cubic spline. Spline inter- interpolant that passes exactly through the given data points
polation uses low-degree polynomials at each of the intervals, but also for regression, i.e., for fitting a curve through noisy
and chooses the polynomial pieces so that they fit smoothly data [26].
together. Fig. 14 shows the linear and spline interpolations A curve–fitting tool by Matlab [27] was used for calculating
where the true AoA was at 90◦ and in Fig. 15 where the true the polynomial coefficients. Fig. 16 shows the approximated
AoA was at 105◦ . Where the true AoA was set in the direction data with polynomial of the 9th degree:
of antenna 3, both the linear and spline interpolations gave the
same minima of close to 90◦ . Fig. 15 depicts a case where the y(x) = p1 · x9 + p2 · x8 + ... + p8 · x2 + p9 · x + p10 (7)
true AoA was between antenna 3 and 4 (105◦ ). Using linear
The algorithm with polynomial approximation returned 91◦ ,
polarization, the algorithm returned AoA at 90◦ , which meant
the estimated AoA was a little closer in comparison to linear
the error was about 15◦ . The algorithm with spline polarization
polarization. The error was about 14◦ .
estimated AoA at 99◦ and the error was much lower, at about
6◦ . Gaussian approximation used the general model Gauss2
with equation:
2 2
B. Polynomial and Gaussian approximations y(x) = a1 · e(−((x−b1 )/c1 ) ) + a2 · e(−((x−b2 )/c2 ) )
(8)
In the next case the Polynomial and Gaussian approxima- The result of Gaussian approximation is in Fig. 17. The
tions were included within the algorithm. When the number estimated AoA was almost exactly at 105◦ , with an error of 0◦ .
of data points (in our case twelve) is equal to an order of In this case the Gaussian approximation was the best choice.

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
7

Antennas number Antennas number

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
5

−60

−80 MUSIC
Estimated AoA−180° Estimated AoA−180°
Measured RSSI points Measured and mirrored RSSI
−5
Measured RSSI [dBm]

Gaussian interpolation
−100

RSSI
−10

−120

−15

−140

−20

−160

−25
50 100 150 200 250 300 350 50 100 150 200 250 300 350
Angle [°] Angle[°]

Fig. 17. Outdoor measurement using Gaussian approximation at true AoA at Fig. 18. MUSIC estimator used on mirrored RSSI measurements.
105◦ (distance 2 m).

the maximums in the signals appear. In our case it is necessary

C. MUSIC estimator to search for the minimums of the received signal and because
In regard to comparisons between interpolations techniques of that we arranged an input signal for the MUSIC estimator.
we implemented the MUSIC (Multiple Signal Classification) The raw measured signal depicted with dots in Fig. 15 is
estimator [13] for obtaining AoA. This MUSIC algorithm is mirrored (depicted in Fig. 18) before going to the MUSIC esti-
well–known in AoA estimation. The inputs of the MUSIC mator. The mirrored signal exhibits peaks instead of minimums
algorithm are usually the amplitude and phase of the received which are crucial for correct MUSIC estimation. In regard to
signal. MRF24J40 radio modules return just the amplitude of a this modification, the parameters of the MUSIC estimator must
signal (RSSI). The authors in [13] (Chapter 10, p.343) reported be set appropriately. Instead of the beam width of the antenna,
a MUSIC estimator which neglected information about the we used the width of the so–called ‘cone of silence’ which is
signal phase. This RSSI estimator is defined as follow: the opposite of the beam width, in our case 20◦ . Fig. 18 shows
The RSSI value on each antenna (i) can be written as: the measured mirrored signal and the pseudospectrum of a
signal as an output of the MUSIC estimator. The amplitudes
Si [k] = Gi (θ)xi [k] + ni [k], (9) of both signals are adjusted for better representation in Fig.
where Gi (θ) are the gains of each antenna, xi [k] are those 18.
signals assumed as being non–correlated from snapshot to
snapshot [k]. An estimation of the correlation matrix Rss of VII. C OMPARISON BETWEEN IMPROVED ALGORITHM
the received signal is: ACCURACIES WITH DIFFERENT
K INTERPOLATIONS / APPROXIMATIONS
1 X
R̂ss = S[k]S[k]T (10) The interpolations method improved the accuracy of AoA
K
k=i estimation using, in our case, multiple omnidirectional an-
By applying the single-value decomposition, a set of three tennas. The resolution of the estimation was 30◦ using the
following matrices is obtained: simple algorithm without interpolations and approximations.
This section provides a comparison between the differently
R̂ss = U SU ∗ (11) used interpolations/approximations.
The space spanned by the signal is partitioned U = [Us , Un ], Fig. 19 shows AoA estimation using five interpolations at
the matrix Us contains the singular vectors corresponding to a few different pre-set AoA’s. Measurements were carried-
the largest singular value, and the matrix Un containing the out outside in LOS between transceivers. The linear interpo-
singular vector corresponding to the smallest singular value. lations provided an average error of about 15◦ . Polynomial
Us is the signal subspace, and Un is the signal null space approximation was slightly better and returned about 10◦
(complementary space of the signal subspace). Because U is of error. The MUSIC estimator returned similar results on
a unitary matrix, the signal and noise subspaces are orthogonal average as polynomial approximation. Much more accurate
(Us Un = 0). This can be defined as a pseudospectrum of the were the Spline and Gaussian interpolations/approximations
MUSIC algorithm: where the errors reached about 6◦ and 4◦ , respectively. RMSE
1 was calculated for measurements between 1 to 30 m for all
PM U SIC (θ) = , (12) interpolation/approximation methods at different angles, as is
G(θ)Un
shown in Fig. 20.
which exhibits a peak of angle (θ) close to the actual angle The indoor measurements were more problematic because
ˆ
(θ). of the influences of reflected signals from the obstacles.
Basically the RSSI MUSIC estimator searches areas where Meanwhile, because LOS was between the transceivers, large

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
8

16
15
Linear 14
Spline
Polynomial 12
10
Gaussian
10
AoA estimation error (°)

AoA estimation error (°)

MUSIC
8
5
6

4
0
2

0
−5
−2 Linear
Spline
AoA = 90° AoA = 105° AoA = 165° AoA = 360° Polynomial
−4
−10 Gaussian
1 2 3 4 1 2 3 4 5 6
MUSIC
Set No. Distance (m)

Fig. 19. Outdoor AoA estimation error at different pre-set AoA’s (distance Fig. 21. Outdoor AoA estimation errors vs. distance at 105◦ .
2 m).

20

70
10 AoA = 195°
Linear
60 Spline
Polynomial
0

AoA estimation error (°)

Gaussian
50
MUSIC
AoA = 90° AoA = 105° AoA = 300°
−10
AoA RMSE (°)

40

30 −20

20 Linear
−30
Spline
10 Polynomial
−40 Gaussian
0 MUSIC
−50
−10
AoA = 135° AoA = 225° AoA = 300° 1 2 3 4
1 2 3 Set No.
Set No.

Fig. 22. Indoor AoA estimation errors at different preset AoA’s (distance of
Fig. 20. Outdoor RMSE at different pre-set AoA’s and distances from 1 m 1 m).
to 30 m.

fluctuations were caused by reflected signals from the floor, 120

Linear
ceiling, walls, furniture, etc. Reflections of the signal causes Spline
”fake” minimas, which can not be very successfully eliminated 100 Polynomial
Gaussian
by our improved algorithm. Therefore the errors of AoA MUSIC
80
estimates were grates within the indoor environment. Again,
AoA RMSE (°)

Fig. 22 shows that the best approximation was the Gaussian 60

at all AoA measurement points. Fig. 23 shows RMSE for

40
measurements taken in an underground garage at distances
from 1 to 30 m and at different pre-set angles. 20

Fig. 21 and Fig. 24 show the averaged errors for the six
0
carried-out outdoor and indoor estimations respectively. The AoA = 135° AoA = 225° AoA = 300°
errors are given for each interpolation/approximation and at 1 2 3
Set No.
distances from 1 m to 6 m between transmitter and receiver.
In Fig. 21 the outdoor estimations are shown. It can be
Fig. 23. Indoor RMSE at different pre-set AoA’s and distances from 1 m to
observed that the distance had negligible influence on the 30 m.
AoA estimations. It was a completely different story inside
the building where the measurements were also taken. In
Fig. 24 extremely large errors can be observed at distances different interpolation/approximation methods within different
of 3 m. These errors reached up to 70◦ . These significant environments. It was expected that the bigger errors would
deviations were caused because of environmental specifics, be within the indoor environment. For each method RMSE is
and interpolations/approximations were not able to reduce calculated and is presented in Fig. 25. RMSE was calculated
them. from all the taken measurements (about 100 for each method).
The last Fig. 25 shows overall comparisons between all the It is evident that the Gaussian approximation provided the best

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
9

iterpolations/aproximations, we improved the resolution and

70
accuracies by the proposed simple measurement method in
60 free–space outdoor measurements and indoor environment
Linear
Spline measurements with reflected signals from obstacles.
50 Polynomial
AoA estimation error (°)

Gaussian
40
MUSIC
R EFERENCES
30
[1] L. Bras, N.B. Carvalho, and P. Pinho. Pentagonal patch-excited sector-
ized antenna for localization systems. Antennas and Propagation, IEEE
20
Transactions on, 60(3):1634–1638, 2012.
[2] N. Patwari, A.O. Hero, M. Perkins, N.S. Correal, and R.J. O’Dea. Rela-
tive location estimation in wireless sensor networks. Signal Processing,
10
IEEE Transactions on, 51(8):2137–2148, 2003.
[3] S. M. Lanzisera. RF Ranging for Location Awareness. PhD thesis, UC
0
1 2 3 4 5 Berkeley, 2009.
Distance (m) [4] M. Malajner, K. Benkic, P. Planinsic, and Z. Cucej. The accuracy
of propagation models for distance measurement between wsn nodes.
Fig. 24. Indoor AoA estimation errors vs. distance at 105◦ . In Systems, Signals and Image Processing, 2009. IWSSIP 2009. 16th
International Conference on, pages 1–4, 2009.
[5] N. Patwari, J.N. Ash, S. Kyperountas, A.O. Hero, R.L. Moses, and N.S.
Correal. Locating the nodes: cooperative localization in wireless sensor
120
networks. Signal Processing Magazine, IEEE, 22(4):54–69, 2005.
Linear [6] S. Maddio, A. Cidronali, and G. Manes. An azimuth of arrival detector
100 Spline based on a compact complementary antenna system. In Microwave
Polynomial
Gaussian
Conference (EuMC), 2010 European, pages 1726–1729, 2010.
80 MUSIC [7] Zizheng Cao, H.P.A. van den Boom, Rongguo Lu, Qing Wang, E. Tang-
diongga, and A.M.J. Koonen. Angle-of-arrival measurement of a
AoA RMSE (°)

microwave signal using parallel optical delay detector. Photonics

60
Technology Letters, IEEE, 25(19):1932–1935, Oct 2013.
[8] Weile Zhang, Qinye Yin, Hongyang Chen, Feifei Gao, and N. Ansari.
40 Distributed angle estimation for localization in wireless sensor networks.
Wireless Communications, IEEE Transactions on, 12(2):527–537, 2013.
20 [9] M. Abusultan, S. Harkness, B.J. LaMeres, and Yikun Huang. Fpga
implementation of a bartlett direction of arrival algorithm for a 5.8ghz
0 circular antenna array. In Aerospace Conference, 2010 IEEE, pages
Outdoor Indoor 1–10, 2010.
1 2 [10] M.R. Kamarudin, Y.I. Nechayev, and P.S. Hall. Onbody diversity
Outdoor/Indoor
and angle-of-arrival measurement using a pattern switching antenna.
Antennas and Propagation, IEEE Transactions on, 57(4):964–971, 2009.
Fig. 25. Comparison of all interpolation/approximation methods. [11] Yuan Shen and M.Z. Win. On the accuracy of localization systems
using wideband antenna arrays. Communications, IEEE Transactions
on, 58(1):270–280, 2010.
[12] K.A. Gotsis, K. Siakavara, and J.N. Sahalos. On the direction of
results, with outdoor errors of about 10◦ and indoor about arrival (doa) estimation for a switched-beam antenna system using
20◦ , respectively. The spline and Linear interpolations and neural networks. Antennas and Propagation, IEEE Transactions on,
MUSIC estimator returned approximately similar results. The 57(5):1399–1411, May 2009.
[13] R. M. Buehrer) Seyed A. (Reza, editor. Handbook of Position Local-
maximum errors were obtained by Polynomial approximation ization Theory, Practice, and Advances. Wiley, 1 edition, 2012.
because of oscillations. The results show that the Gaussian [14] Nanyan Wang, P. Agathoklis, and A. Antoniou. A new doa estimation
approximation improved the accuracy of AoA estimation for technique based on subarray beamforming. Signal Processing, IEEE
Transactions on, 54(9):3279–3290, Sept 2006.
class, better than others. [15] Qiaowei Yuan, Qiang Chen, and K. Sawaya. Accurate doa estimation
using array antenna with arbitrary geometry. Antennas and Propagation,
IEEE Transactions on, 53(4):1352–1357, April 2005.
VIII. C ONCLUSION [16] Y. Takahashi, H. Yamada, and Y. Yamaguchi. Array calibration
These results from our research show that it is possible to techniques for doa estimation with arbitrary array using root-music
algorithm. In Microwave Workshop Series on Innovative Wireless Power
obtain quite accurate estimations of AoA from RSSI measure- Transmission: Technologies, Systems, and Applications (IMWS), 2011
ments by using multiple, in our case twelve, omnidirectional IEEE MTT-S International, pages 235–238, May 2011.
antennas placed circularly on a plate and the plate is statically [17] Xin Chen, Yu. Morton, and F. Dovis. A computationally efficient
iterative mle for gps aoa estimation. Aerospace and Electronic Systems,
placed. It was assumed that there were no moving obstacles in IEEE Transactions on, 49(4):2707–2716, OCTOBER 2013.
the line-of-sight between the transceivers and that the trans- [18] R.L. Johnson and G.E. Miner. An operational system implementation
mitter and receiver lay within the same azimuth plane. The of the espirit df algorithm. Aerospace and Electronic Systems, IEEE
Transactions on, 27(1):159–166, Jan 1991.
expected resolution of the system with the simple algorithm [19] B.N. Hood and P. Barooah. Estimating doa from radio-frequency
is 30◦ . For the purpose of increasing the resolution/accuracy rssi measurements using an actuated reflector. Sensors Journal, IEEE,
of AoA estimation, a few interpolations/approximations were 11(2):413–417, 2011.
[20] J. Graefenstein, A. Albert, P. Biber, and A. Schilling. Wireless node
used and compared. The greater improvements were archived localization based on rssi using a rotating antenna on a mobile robot.
by Gaussian interpolation between the RSSI measurement In Positioning, Navigation and Communication, 2009. WPNC 2009. 6th
points. The estimated RMSE decreased on average, from Workshop on, pages 253–259, 2009.
[21] M. Malajner, P. Planinsic, and D. Gleich. Angle of arrival estimation
30◦ to 10◦ for outdoor and from 60◦ to 25◦ for indoor, using rssi and omnidirectional rotatable antennas. Sensors Journal,
in comparison with the simple algorithm. Using different IEEE, 12(6):1950–1957, 2012.

1530-437X (c) 2013 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JSEN.2014.2386537, IEEE Sensors Journal
10

[22] R. Dean Straw. The ARRL Antenna Book. The national association for
ARRL amateur radio, 21 edition, 2007.
[23] C. A. Balanis. Antenna Theory, Analysis and Design. Wiley, 3 edition,
2005.
[24] Ansys hfss.
[25] Mrf24j40ma - rf.
[26] Interpolation - wikipedia, the free encyclopedia.
[27] Curve fitting toolbox - matlab.

1530-437X (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
View publication stats http://www.ieee.org/publications_standards/publications/rights/index.html for more information.