Using Unmanned Aircraft Systems

for Mobile Network Verifications

Marjo Heikkilä, Marjut Koskela, Tero Kippola, Jouni Tervonen
Mahmut Kocak, Juha Erkkilä University of Oulu
Centria University of Applied Sciences Kerttu Saalasti Institute
Ylivieska, Finland Nivala, Finland
Email: [email protected] Email: [email protected]

Abstract—This study presents the usage of unmanned in both horizontal and vertical dimensions. Using an
aircraft for site survey and antenna pattern measurements Unmanned Aircraft System (UAS) in testing provides
in-situ in mobile networks. The site survey is one part new possibilities to perform network measurements, it
of the verification and optimization process of the network
operator. Mobile network operators’ responsibility is to en- enables actual on-site 3D beamforming measurements
sure the broadly available, reliable network series that are and allows to perform measurements automatically. An
essential for customer experience. Technically that means Unmanned Aerial Vehicle (UAV) enables measurements
mobile network coverage of the desired area without holes to be taken from different elevation angles. UAV also
and with good capacity. The operator performs network makes it possible to measure in rough terrain or areas
planning in order to optimize coverage and capacity. It is
essential to know the antenna pattern to plan and design an without roads; compared to traditional field testing, one
effective and efficient mobile network. Antenna manufac- can operate a UAV where humans or cars cannot go.
turers provide antenna radiation patterns measured in the UASs are becoming popular for professional use be-
standard anechoic chamber. Understanding how real site cause of the easily operated and cost-efficient aerial
implementation differs from the laboratory measurements technology. A battery-powered UAV equipped with a
is also crucial. This study focuses on developing in-situ
antenna measurements for the antenna patterns recorded Global Positioning System (GPS) and suitable measure-
in the laboratory. The evaluation of antenna radiation ment equipment is capable of gathering data from the
patterns in a real environment will provide additional air. They are commonly used for aerial photography and
information about what kind of effect the antenna site surveys. This study focuses on how UASs can be used
mast, installation, and antenna performance parameters in an efficient way for mobile network site surveys and
will have on the theoretical radiation pattern.
network verification.
I. I NTRODUCTION The Section I introduces a developed measurement
method, which can increase the effectivity of verifying
This paper illustrates the development of the practi- future mobile networks. The related work is described in
cal 3D measurement method for mobile networks. The SectionII. The measurement environment and equipment
complexity of future networks will increase because are presented in Section III. The measurement setup is
they consist of many different technologies and frequen- shown in Section IV. Section V describes the measure-
cies. When the complexity of networks increases, the ments and methods. The results are shown in Section VI.
critical verification of the network performance in-situ The conclusions are drawn, and future work is discussed
also becomes increasingly challenging. The increasing in Section VII.
demand for high reliability and availability of networks
means that a network must be verified with testing. The II. R ELATED W ORK
field testing with smart-phone and measurement software Traditionally, the antenna characterization measure-
is one method of producing practical knowledge of ments, i.e., the evaluations of the antenna radiation pat-
mobile technologies. This method usually provides only terns, are performed either indoors in anechoic chambers
a 2D view of the network performance. Field testing is or in specific outdoor antenna measurement ranges. In
typically conducted with a car. Future mobile technolo- recent years, the in-situ antenna measurements utilizing
gies will utilize 3D beamforming. In 3D beamforming, UASs have aroused scientific interest, e.g., [1], [2], [3],
the beam can be steered horizontally and vertically. and [4]. The obvious advance of in-situ measuring is
In addition to steering the beam, the tilt and azimuth that it enables analyses of antenna radiation patterns in
can also be changed dynamically by using adaptive realistic conditions, and takes into account the influence
antennas. With all these adjustments, network throughput of the surrounding environment when measuring the
can be improved. When these antenna features are stud- radiation pattern via multipath contributions from the
ied, it is also necessary to study network performance ground and surrounding structures. A perfect example of
the latter is the experiment wherein an UAS was used for was used for 3D antenna measurement with the mobile
measuring a circular azimuthal path with a 20◦ elevation network besides a Finnish language Bachelor level thesis
angle of a log-periodic antenna at 433 MHz mounted on work in [11].
a ship [1].
An interesting practical example demanding UAS- III. M EASUREMENTS OF THE E NVIRONMENT AND
based antenna measurement was introduced in [2] for S YSTEM T ESTS
the radio-astronomical usage of the Square Kilometer When the radiation pattern of the antenna is studied
Array (SKA) project. In the SKA experiment, even the in 3D, several issues must be taken into account. The
small subarray was found to be too large to be measured system parameters must be known beforehand when
in an anechoic chamber, requiring the development of measurements are planned, as the test environment as
an in-situ measurement. To prepare the SKA measure- well as weather conditions can effect the measurements.
ments, several experiments from single antennas (e.g.,
[5] and [6]) to a 3x3-aperture array of Vivaldi antenna A. Test Environment
elements were conducted [7]. Another example of radio-
The study was performed in the field trial environ-
astronomical use was introduced in [8] where they
ment of Centria University of Applied Sciences. This
presented the results of the UAS-based measurements
environment covers approximately 4 km2 area in a ru-
of a calibration dipole verifying the UAS method for
ral/suburban environment of Ylivieska, Finland. Cho-
calibrating the radio arrays used in astronomy.
sen as eNodeB (eNB) for these measurements was the
Antenna radiation pattern measurements were success- Puuhkala site. That site has the most versatile antenna
fully performed in Australia using UAS in the far field techniques available in the trial environment. The site
of “live” broadcast antennas using ultra high frequency has a self-supporting mast, which means that there are
(UHF) and very high frequency (VHF) digital television no supporting wires around the mast. The antenna mast
systems [9]. Beyond the examples conducted with either is located at small hill surrounded by pine trees. Reported
VHF or UHF bands, Duthoid et al. [3] reported in- antenna installation height in the mast is 42 m and
situ antenna characterization measurements performed the total height of the mast is 60 m. Figure 1 shows
in super high frequency (SHF) band at 9.8 GHz. This the Puuhkala site and the environment around it. The
system is aimed at radar systems where the in-situ measurements took place in clear weather where the
antenna includes undesirable reflections, diffraction, and temperature was around 15 ◦C and wind speed was 1–4
other external sources of contamination. The antenna is m/s. These conditions were beneficial also for the UAS
surrounded by other elements such as radome, tower, and operation; both wind and rain create a risky environment
lightning protection. Additionally, ground irregularities for the flight.
from geography and weather conditions that are different
for each site impact the performance of radars. B. Antenna
Although it has no direct impact on the measurement In the measurement configuration, an antenna with an
case presented in this paper, the literature review of operating frequency band of 1710 MHz to 2180 MHz
UAV antenna measurements revealed a limiting factor, was used in 2.1 GHz Long Term Evolution (LTE) band
one which is worth mentioning. In cases of very large 1 with 5 MHz bandwidth operated from Puuhkala eNB.
antenna arrays, the far field range is too large for Antenna is commercial quad port Remote Electrical
practical measurements with small UAV with limited Downtilt (RET) compatible antenna with a 65◦ hori-
flight time and clear visibility requirements. To overcome zontal beamwidth [12]. The antenna length is 1300 mm
such limitations, a method including the near field to far and the width is 300 mm. Figure 2 and 3 present
field transformation allowing shorter UAS flights in the manufacturer’s antenna patterns with 8◦ downtilt.
near field was introduced in [4]. The methodology was
validated with measurements in UHF and SHF bands, at
2950 MHz and 4650 MHz.
UASs have been used in the context of mobile network
characterization of 3D RF fields in [10] although the
scope was not primarily the antenna radiation pattern.
The measurements were performed mostly 500 m away
from the base station antenna and the goal was to suggest
potential improvements in the speed and accuracy of
multi-node radio network planning and the deployment
of hastily-formed (e.g., emergency) networks, and gen-
eral network design and optimization. To the best of our
knowledge, this is the first scientific article where UAS Fig. 1. Aerial view of the Puuhkala site.
 the transmitter and receiver, is used [13]. This technology
has an impact on the RSRP pattern because the RSRP
and Signal to Noise Ratio are proportional to each other
on average. In our case, the Spatial Multiplexing Open
Loop MIMO (SM OL MIMO) was active in the eNB
during the measurement. Polarization diversity was used.
As a Key Performance Indicator (KPI) of an antenna,
RSRP is one of the basic physical layer measurements.
RSRP is critical because it represents the linear average
of the downlink reference signals of a given channel
width enabling coverage area evaluation of the LTE cell.
Additionally, there is certainly a correlation between
RSRP and RS-SNR the user plane Quality of Service
Fig. 2. Manufacturer’s horizontal pattern. (QoS).

IV. M EASUREMENT S ETUP

The measurement setup includes all the components
used in operation, both on the ground and in the air (Fig.
4). A UAV with the test device was operated from the
ground level. A suitable ground control station [14] was
used to operate the UAV with automatic mode and a
UAV controller was used to fly manually when needed.

A. Test Device
When using the UAV for measurement purposes, the
payload in the UAV must be light to minimize the take-
off weight. In this case, the used payload i.e., UE, was
a commercial mobile phone with suitable measurement
Fig. 3. Manufacturer’s vertical pattern. software [15]. More detailed information about the test
devices can be found in Table I. During the measure-
ment, the UE was fastened to the UAV with an adjustable
Different mechanical properties, such as antenna in- holder. While the UAV was flying around the eNB’s
stallation and mast structure, can affect the radiation antenna, the UE collected data.
pattern. In our case, the antenna was installed so that All mobile phones have a unique radiation pattern.
the mechanical tilt was approximately 4◦ and the antenna At the beginning of the study, the antenna radiation
RET was set to 4◦ to get the desired 8◦ downtilt. pattern of the UE was measured in the anechoic chamber.
The measurement did not examine the influence of
C. Key Performance Indicators in LTE the UAV on the radiation pattern. Based on the UE’s
antenna radiation pattern results, we determined the most
In an LTE network, in addition to the obstacles such
suitable position for the UE, where both the vertical and
as trees and buildings between the antenna and User
horizontal patterns had a wide enough angle area with a
Equipment (UE), the distance between eNB and the
flat gain, and the gain was sufficient. Radiation pattern
equipment also affects the throughput. The effective use
of dynamic modulation and a coding scheme results in
high data rates and capacity when the UE is located
near the eNB. Using UAV as a testing platform gives us
new possibilities and opportunities to measure the perfor-
mance parameters of an LTE cell, including Reference
Signal Received Power (RSRP), Reference Signal Re-
ceived Quality, downlink throughput, Reference Signal
- Signal to Noise Ratio (RS-SNR), and Transmit Power
values in a real-time environment.
To enhance communication performance via the use
of spatial diversity and spatial multiplexing in data trans-
mission, the Multiple-Input Multiple-Output (MIMO) ra-
dio technology, which utilizes multiple antennas at both Fig. 4. Measurement setup in use.
 TABLE I. Test devices We defined the most suitable measurement points that
ROLE EQUIPMENT SPECS
can give accurate data about the radiation pattern of
the antenna. One must consider issues concerning the
UAS UAV Drone DJI S900 (incl. GPS)
Ground Station SW UgCS Desktop application environment, the measurement setup, and the system
Ground Station HW Laptop Computer under the test. One of the measurement setup issues
Remote Controller Futaba T14SG that has a substantial impact on the measurements is the
UE Test SW Nemo Handy flight time, and that is limited because of the battery-
Test Device Sony Xperia X powered UAV. The flight time of the selected UAV was
18 minutes, which restricts measurement time. National
limitations to UASs operations have to be taken into
results revealed that the most suitable position of the UE account, such as a maximum flight altitude of 150 meters
was the vertical tilt at 45◦ . The UE was facing the eNB above ground level. Antenna mast structure, surrounding
antenna in vertical position rotated 45◦ to the left while buildings, terrain height variations, or vegetation can also
the UAV flew around the eNB’s antenna. limit the measurement. In our case, the mast was self-
B. Unmanned Aircraft System supporting so there were no supporting wires on the
flying route.
The UAS used in these measurements was DJI S900 Antenna installation height was 42 m, so the national
Spreading Wings [16], which is a commercial aircraft for maximum flight limitation did not prevent measure-
professional use. The diagonal wheelbase of the UAV is ments; however, the forest surrounding the site limited
900 mm and the total height is 730 mm. In this trial, measurements below the antenna. In order to find the
flight operations lasted approximately 15 minutes when most suitable measurement points and optimize the
the take-off weight with UE was 6.5 kg. Among other spherical route, while also minimizing the length of the
factors, the take-off weight is one of the issues that affect circumference of the circular paths, we defined our route
the flight time of a UAV. in the far field area. Hence, in the far field the radiation
In order to use the UAS for radio frequency measure- pattern does not change shape as the distance increases.
ment purposes, a series of UAS regulations were studied Equation 1 is to approximate the far field region:
beforehand. The Finnish Transport Safety Agency (Trafi)
has defined these regulations [17] concerning UAS usage 2D2
R= (1)
in Finland. These regulations may vary between different λ
countries (like traffic regulations) [18] concerning UASs. Where D is the length of the antenna, λ is the
In addition to the Finnish traffic regulations, specific wavelength, and R is the radius of the route. With the
permission for the use of the mobile phone (UE) with used antenna at 2100 MHz this means that the far field
the UAS was needed from the Finnish Communications starts approximately from 23.5 m. We chose the radius
Regulatory Authority. Each flight operation with a UAS of 25 m for our measurement routes. We called these
was conducted under these regulations and precautions. circular paths, forming a sphere, a target route. Each
circle had its own height from the ground, defining the
V. M EASUREMENT METHOD tilt of the UE. We have defined 26 different tilt angles for
The target for this study was the mobile network site 26 different circles, starting from a height of 22 m from
performance and the antenna radiation pattern verifica- the ground. The distance between the eNB antenna and
tion using the UAS. The UAS technology has elevated the UE measurement was the same at each circle, and
the possibility of developing field measurements closer this formed a spherical shape as shown in the Fig. 5. A
to the antenna pattern measurements performed in the few circle layers were left out of planned measurements
laboratory. In the laboratory, antenna measurement de- for safety reasons avoiding flying too close to obstacles.
vices have been designed to build a high-resolution 3D In the measurements, the UAV flew around the eNB
scan of the antenna field pattern, using the ring-shaped one circle at a time. In practice, the UAV with UE was
element that surrounds the transmit antenna which col- raised manually near the desired circular level, where
lects the data by rotating a few degrees in each scan. automatic mode was turned on. The UAV flew around
When the UAV is used as a measurement for the UE the antenna at an even speed, which was approximately
platform, it provides the opportunity to collect real-time 1 m/s. Depending on the radius of the circle, samples
data, using an almost perfect spherical shape, spatially. were collected in different amounts in different circular
The spherical route will be beneficial in the analysis layers. The UE gathered samples approximately every
of data in case we compare the results from different 500 milliseconds. The measurement was performed in
directions and for further predictive approaches such as the downlink direction. The UAV flew around the an-
clustering or regression. tenna in a complete circle with automatic mode and was
When the radiation pattern is studied, the signal landed manually. In each circle, the predetermined angle
strength information at a constant radius is needed. of the UE was adjusted, and the UE pointed towards the
Fig. 5. Planned measurement points in spheres around
the antenna. Fig. 6. UAV flying route of one measurement flight.

eNB’s antenna. The flight direction was the same for all points are green in Fig. 6. Filtered measurement points
circles. are red and blue.
The following mathematical operations were used in
VI. R ESULTS order to plot the relationship between the signal strength
and the angle of the radius. First the RSRP results
The target was to visualize the results of the radiation were interpolated. Since the spherical route with the
pattern, depending on the received signal strength from initial point grid was not dense enough, we improved the
the antenna. The signal strength needs to be plotted at grid using linear interpolation. This method utilizes low-
a constant radius to make a diagram with the polar degree polynomials in each of the intervals. It selects the
plot. The idea is to compose two separate graphs of polynomial pieces, so they fit smoothly together. After
the vertical plane and the horizontal plane. The most interpolation, RSRP values were averaged. The values of
important parameter to generate the patterns was signal nine points were averaged, four values before and after
strength, RSRP value in this case. To form the correct the measurement point.
radiation pattern, the distance between the UE and the Finally, from the processed measurement data, the
antenna must be equal while flying. To achieve this the horizontal and vertical planes of radiation pattern were
UAV flew a steady and near-perfect circle at different plotted (Fig. 7 and 8). The measurement data was
heights. This required the information of the geographic compared with the antenna manufacturer’s information
location, given for example by the UE. Geographic loca- (Fig. 2 and 3).
tion information includes latitude, longitude and altitude As the mast structure and the placement of the antenna
parameters. These parameters give information where in the mast may affect antenna’s electrical properties,
the RSRP value has been collected in three different
dimensions. During the study, we realized that UE’s
GPS information is not accurate enough, especially the
altitude information. UAS demands reliable navigation
and accurate GPS receiver for autonomous or semi-
autonomous operation. We realized that we could utilize
the GPS data of the UAV in visualization. In these
results, we have replaced UE GPS data with UAV GPS
data. Figure 6 presents the UAV route and measurement
points with a point cloud.
RSRP values were recorded from the beginning of the
flight until end of the flight. GPS time was needed to
define the starting point of the measured circular layer,
while take-off and landing data was not included. Only
values in a circle around the antenna have been included
when the pattern was generated. These measurement Fig. 7. Measured horizontal pattern.
 ification measurements. Antenna pattern measurement
with the UAS helped to understand how real site imple-
mentation affects the patterns measured in the laboratory
environment. In our study, the measured radiation pattern
corresponded to the manufacturer’s reported pattern quite
well.
The main conclusion of the study is that using UAS for
testing brings new possibilities to performing network
measurements and enables actual 3D beamforming mea-
surements of future mobile networks. UAS also enables
the measurements to be done automatically which have
not been possible in traditional network measurements
while using a car. We were able to measure the beam,
Fig. 8. Measured vertical pattern. even when steered with different tilt and azimuth angles.
Future studies will consider further development of the
measurement and visualization of the 3D radiation pat-
manufacturer’s antenna measurements can vary from tern. Repeated 3D study is important for future networks
measurements carried out after the installations. In our not only because of emerging 3D beamforming but also
case, the measured radiation pattern corresponded to the because it can be seen that UAS data services demand
manufacturer’s reported pattern quite well. a different type of mobile network design.
The major difference in the horizontal plane between
the radiation patterns was the width and direction of ACKNOWLEDGMENT
the main beam. The UAS measured the half power This work has been carried out using the framework
beam width (HPBW) of the horizontal plane, which of the IMAGE 5G project [19]. The authors would like
was approximately 55◦ ; in the antenna specifications, to acknowledge the IMAGE 5G research consortium
the HPBW is 64◦ . The downtilt configuration could be consisting of the Centria University of Applied Sciences,
one issue to cause a difference in radiation patterns. In Nokia, PehuTec Oy, Siipotec Oy, Pohjonen Group Oy,
the measurements, the combination of the mechanical the Finnish Communication Regulatory Authority, the
tilt and the electrical tilt was used. The manufacturer Finnish Meteorological Institute, and Business Finland
gives a radiation pattern with only the electrical tilt in – The Finnish Funding Agency for Innovation.
use. It was noticeable that the measured main beam R EFERENCES
was asymmetrical. Another cause of the differences was
[1] F. Üstüner et al., “Antenna radiation pattern measurement using
most probably the antenna mast structures. The measured an unmanned aerial vehicle (UAV),” in 2014 XXXIth URSI Gen-
Upper Sidelobe Level Suppression (USLS) was 14 dB, eral Assembly and Scientific Symp. ,URSI GASS, Aug 2014, pp.
and in the antenna specifications the USLS was 18 dB. 1–4.
[2] G. Virone et al., “Antenna pattern verification system based on
The measured front-back ratio at 180◦ was 30.7 dB, a micro unmanned aerial vehicle (UAV),” IEEE Antennas Wirel.
while in the antenna specification the front-back ratio Propag. Lett., vol. 13, pp. 169–172, 2014.
was 30 dB. The major difference in the vertical plane [3] S. Duthoit et al., “A new approach for in-situ antenna char-
acterization, radome inspection and radar calibration, using an
was the number of the sidelobes. The measured radiation unmanned aircraft system (UAS),” in IEEE Radar Conf., Radar-
pattern consisted of less sidelobes than the manufac- Conf, May 2017, pp. 0669–0674.
turer’s reported one. The measured sidelobes had wider [4] M. Garcı́a-Fernández et al., “Antenna diagnostics and characteri-
zation using unmanned aerial vehicles,” IEEE Access, vol. 5, pp.
lobe width compared with the antenna specifications. 23 563–23 575, 2017.
[5] G. Virone et al., “Antenna pattern measurements with a flying far-
VII. C ONCLUSION AND F UTURE W ORK field source (Hexacopter),” in IEEE Conf. Antenna Meas. Appl.,
The original target of the study was to find out if CAMA, Nov 2014, pp. 1–2.
[6] F. Paonessa et al., “Effect of the uav orientation in antenna pattern
a UAS could be part of the optimization process used measurements,” in Eur. Conf. Antennas Propag., EuCAP, May
by mobile network operators to ensure a good customer 2015, pp. 1–2.
experience. A new method of measurement needed to be [7] G. Pupillo et al., “Medicina array demonstrator: calibration and
radiation pattern characterization using a uav-mounted radio-
developed for mobile network testing with UAS. When frequency source,” Exp. Astron., vol. 39, pp. 405–421, 04 2015.
the method was developed, many technical, safety, and [8] D. C. Jacobs et al., “An external calibrator for hydrogen obser-
even regulatory issues had to be considered. Because vatories,” in IEEE Conf. Antenna Meas. Appl., CAMA, Oct 2016,
pp. 1–3.
UAV does data gathering measurements from the air, [9] J. Schreiber, “Antenna pattem reconstitution using unmanned
the vertical dimension also needed to be considered in aerial vehicles (UAVs),” in IEEE Conf. Antenna Meas. Appl.,
the measurement result analysis. CAMA, Oct 2016, pp. 1–3.
[10] E. Teng et al., “Aerial sensing and characterization of three-
The study revealed that the UAS suits the mobile net- dimensional RF fields,” in Proc. 2nd Int. Workshop Robot. Sensor
work site performance and antenna radiation pattern ver- Netw., Seattle, WA, USA, 2015, pp. 1–6.
[11] N. Gunell, “Puhelintukiasema-antennin säteilykuvion mittaus
multikopterilla,” p. 30, 2017-01-16. [Online]. Available: http:
//urn.fi/URN:NBN:fi:aalto-201701171089
[12] CommScope Inc., “Hbxx-6516ds-vtm specifications.” [On-
line]. Available: https://www.commscope.com/catalog/wireless/
product details.aspx?id=15653
[13] M. Heikkilä et al., “Development of the measurement method for
challenging nlos conditions in mobile networks,” in Proc. IEEE
Int. Workshop Meas. Netw., M&N, Sept 2017, pp. 1–6.
[14] SPH Engineering Inc., “UgCS universal ground control
software.” [Online]. Available: https://www.ugcs.com/
[15] Keysight Technologies Inc., “Nemo handy measurement
software.” [Online]. Available: https://www.keysight.com/en/
pd-2767485-pn-NTH00000A/nemo-handy?cc=GB&lc=eng
[16] Dà-Jiāng Innovations Science and Technology Co., Ltd., “DJI
S900 spreading wings specifications.” [Online]. Available:
https://www.dji.com/spreading-wings-s900/info#specs
[17] Liikenteen turvallisuusvirasto, “Trafi unmanned aviation
regulation OPS M1-32.” [Online]. Available: https://www.trafi.
fi/filebank/a/1483970125/4a6ac53bf4b1cb434d7f85a15f36dde0/
23661-OPS M1-32 RPAS 2016 eng.pdf
[18] DroneRules.eu, “Drone regulations in EU, Norway and
Switzerland.” [Online]. Available: http://dronerules.eu/
[19] Centria University of Applied Sciences, “The IMAGE 5G
project.” [Online]. Available: http://www.centria.fi/image5g