You are on page 1of 4

2017 International Conference on Military Technologies (ICMT)

May 31 – June 2, 2017, Brno, Czech Republic

Drone Detection by Ku-Band Battlefield Radar

Ján Ochodnický*, Zdeněk Matoušek*, Marián Babjak* and Ján Kurty*


*Armed Forces Academy of General M. R. Štefánik, Liptovský Mikuláš
e-mail: jan.ochodnicky@aos.sk, zdenek.matousek@aos.sk, marian.babjak@aos.sk, jan.kurty@aos.sk

Abstract—The intensive research and development efforts in the II. THEORETICAL ANALYSIS OF SMALL UAV DETECTION
area of unmanned aerial vehicles (UAV) detection for security
applications is observed in the last decade. The published A. General Radar Target Detection Theoretical Assumptions
research results are focused on applications in particular The radar range RMAX for a given radar target detected by a
domains (e.g. acoustic, ultrasonic, optical, infrared, particular radar system, from an energetic parameters point of
radiofrequency…) whereas the probability characteristics for view is described by the radar equation [6], [10]
each assumed application evaluation are needed. This paper
includes the theoretical analysis of battlefield pulse radar
detection probability and range for small UAV. The PT G 2 λ 2 RCS
interdependence between Signal-to-Noise Ratio (SNR) and small RMAX = 4
3
(1)
UAV Radar Cross Section (RCS) is derived. The results of (4π ) FkT0 BL SNR
dynamic flight test series conducted to evaluate the detection
performance of three drone size categories by pulse Ku-band where PT is the radar transmitter output power, G is the radar
short range battlefield radar are presented. Some dependences of antenna gain, λ is the signal wavelength, F is the noise figure of
the SNR and probability of detection on the drone sizes are radar receiver, k is the Boltzmann constant 1.38.10-23, T0 is the
determined. temperature, B is the bandwidth of radar receiver, L represents
the transmission loses, RCS is the radar target cross section and
Keywords-battlefield radar; unmanned aerial vehicle; radar cross-
SNR represents the Signal-to-Noise Ratio at the radar receiver
section; radar measurements
output. Equation (1) shows, that radar detection range for a
I. INTRODUCTION given target depends on the fourth root of ratio between the
target RCS and the SNR at the receiver output.
Unmanned Aerial Vehicles (UAV)/drones are becoming
attractive solutions for non-military and military applications. As long as RCS for given target fluctuates slowly, the
Some special systems for drones detection and defense reflected signal phase has an uniform distribution and the
solutions are already designed and developed. Primary amplitude has a Rayleigh distribution. With respect to that, one
surveillance radars were not designed specifically for detection may describe SNR as relation between the radar target
and tracking of small, low flying, non-cooperative UAV/drones detection probability PD and the false alarm probability PFa by
at ranges up to 1 mile [1]. In addition, the conventional defense formula [6], [7]
radars and short rage battlefield radars have difficulty with
detection of low Radar Cross Section (RCS) targets. Besides ln ( PFa )
the radar characteristics which usually are provided by SNR = −1 (2)
producers, the probability characteristics of small UAVs should ln ( PD )
be determined. The analysis of small RCS target detection in
primary radar system is presented in [2]. An early warning The graphical interpretation of relation between SNR and
radar to detect the existence of UAV is proposed in [3]. The the radar target detection probability PD for the chosen false
multisensor system based on the light camera laser pointer and alarm probability PFa derived from (2) in the case of single
Ultra-wideband (UWB) radar is proposed in [4]. The 35GHz pulse is shown in the Fig. 1 and the results correspond to [8],
FMCW drone detection system is proposed in [5] and many [9].In order to select a proper radar for a particular application,
other applications for small UAV and drone detection by a customer usually requires all information included in
different radar systems are described in many publications. equation (1) for chosen PD and PFa, which is not commonly
available.
This paper presents in comprehensive form the results of
field trials for detection of small UAV using standard pulse- From both formulas (1) and (2) we can get the equation
coherent short range battlefield radar. As the first, the case reflecting mutual relation between the radar energetic
study of small UAV detection by pulse Ku-band radar is given. parameters and the radar probabilistic parameters
The study is oriented on theoretical analysis of the detection
probability versus SNR for the targets with low RCS. The ln ( PFa )
results of field trials with battlefield radar are compared with PT G 2 λ 2 RCS
= −1 . (3)
theoretical analysis for three drone size categories and 3
(4π ) 4
FkT0 BLRMAX ln ( PD )
recommendation for defense applications is included.

c
978-1-5090-5666-8/17/$31.00 2017 IEEE 613
1
10

H [hm]
0.95 RCS = 20 m2 for λ = 0.02 m
PD = 0.8
0.9 5 PFa = 10
-6

0.85

0.8
0 2.5 5 7.5 10
R [km]
0.75
-5
PFa = 10 Figure 2. The radar detection area for chosen battlefield radar
0.7
PFa = 10-6
0.65 PFa = 10-7 B. Theoretical Estimation of Particular Radar and Target
0.6 Parameters
0.55 Let us select for the first radar system particular radar –
0.5
short range battlefield radar with the manufacturer given radar
10 12 14 16 18 20 22 24 26 28 30 system parameters as follows:
SNR [dB]
- RCST =20 m2 for tank;
Figure 1. Relations between SNR and PD for chosen PFa. - Detection range RMAXT =10km;
- Detection probability PDT = 0.8;
We can get the rate between two different radar targets with
different RCS1 and RCS2 using (3) - False alarm probability PFaT = 10-6;
- Output signal (pulse) power PV = 250W;
PT G 2 λ 2
RCS1  ln PFa 2  - Frequency band KU ;
4
3 RCS1  − 1
RMAX 1 (4π ) FkT0 BL SNR1  ln PD 2  - Pulse width PW = 0.4µs;
= =4 (4) - Pulse repetition interval PRI = 250 µs;
RMAX 2 PT G 2 λ 2RCS 2  ln PFa1 
4 RCS 2  − 1 - Antenna main lobe width in horizontal plane Θ0.5 = 3.5°.
(4π )3 FkT0 BL SNR2  ln PD1 
The radar detection area for the given detection parameters
and afterwards, for the expected radar target with known RCS2, of used battlefield radar is in Fig. 2.
PFa2 and PD2, the required radar range RMAX2 is given by
For the second radar system we chose the same radar and
three different drones with theoretically expected RCSDr values
 ln PFa1  1.5 m2, 1 m2 and 0.5m2.
RCS 2  − 1
RMAX 2 = RMAX 1  ln PD1  . (5)
From the equation (6) and the given radar parameters we
4
 ln PFa 2  can get the particular PDDr function values for the particular
RCS1  − 1 drone with respect to the PFaDr. These functions are shown in
 ln PD 2  the Fig. 3.

In case that we want to analyze an influence of RCS2, which


is a parameter of a particular radar target, on the radar range
RMAX2 with defined PFa2, we can derive for PD2 from the
equation (5) as follows

 X ln(PD1 ) ln (PFa 2 ) 
PD 2 = exp  , (6)
 ( X − 1) ln(PD1 ) + ln(PFa1 ) 

4
RCS1 RMAX 2
with substitution X = .
RCS 2 RMAX 14 -6
RCS = 1.5 sqm, PFa = 10
Equation (6) describes relations between parameters (RMAX , RCS = 1.0 sqm, PFa = 10-6
PD , PFa, RCS) for a particular radar system. These relations are RCS = 0.5 sqm, PFa = 10-6
sufficient for an additional indirect comparison of different
radar targets.

Figure 3. Detection probability function for the radar system with PFa =10-6

614
From equation (6) the maximal radar range for drone TABLE II. RESULTS OF RCS MEASUREMENT OF THE CHOSEN DRONES
RMAXDr can be estimated with respect to the chosen drone Drone type RCSDr [m2]
detection probability PDDr and drone false alarm probability Frequency
X8 High One Iris +
PFaDr (see Table I.). 3GHz 0.27 ± 0.39 0.14 ± 0.12 0.07 ± 0.07
9.7GHz 1.15 ± 1,03 0.49 ± 0.56 0.19 ± 0.16
The obtained theoretical results and analysis show, that the 15GHz 1.65 ± 1,78 0.69 ± 0.75 0.27 ± 0.32
used battlefield pulse radar with declared technical parameters 24.3GHz 3.18 ± 4.09 1.12 ± 1.15 0.42 ± 0.45
can be used for drone detection with sufficient detection
probability at the distance more than 3km. RCS High One [dBm ] 9.7GHz
-90
260 5 -80
TABLE I. ESTIMATION OF RMAXDR FOR THE RADAR SYSTEM 250 -70
240 -60
230 -50
RCS [m2] -5
220 -40
PDDr 0.5 1 1.5 0.5 1 1.5
RMAXDr [m] for PFaDr = 10-6 RMAXDr [m] for PFaDr = 10-5 210 -15 -30
0.7 4480 5330 5900 4700 5590 6180 200 -20
0.8 3980 4730 5230 4170 4950 5480 -25
190 -10
0.9 3290 3910 4330 3440 4090 4530
180 -35 0
III. MEASUREMENT AND EXPERIMENTAL RESULTS 170 10

The experimental measurements of RCSDr of chosen drones 160 20


were made for the comparison with theoretically expected 150 30
RCSDr values. The radar signal measurements with stable and 140 40
slowly moving targets according to theoretical presumptions 130 50
120 60
were made for three different drones. In our case the drones 110 100
80 70
were Iris+, X8 and High One (H_O) showed in the Fig.4. 90

Figure 5. The 360°diagram of RCSDr of the High One drone in the horizontal
plane at 9.7GHz.

-15

-20

-25
Power Spectrum Magnitude (dB)

a) b) -30

-35

-40

-45

-50

-55

-60
0 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
c)
Figure 6. Sample of signal processing related with the moving High One
Figure 4. The drones used in the experimental measurement a) Iris+, b) X8
drone.
and c) High One.
In spite of relatively high variation of RCSDr, experimental
A. Drones Radar Cross Sections Measurement
results showed that expectations for RCSDr of our drones were
Chosen drones RCSDr were measured inside anechoic not far from the measured ones.
chamber in the frequency range 3GHz – 24,5GHz. Obtained
results are shown in the Table II. B. SNR Estimation from Drone Radar Signal
For the SNR estimation, we measured “non moving” and
Resulted values are the mean values obtained from the 360°
slowly moving drones at the height 20m above the ground
horizontal plane diagram of RCS (see the example in the
surface in the natural shrubs and small trees environment. The
Fig. 5). One may see that RCSDr of the particular case has very
target distance from the radar was set up to 1km ± 200m and
high variation. The distribution of measured values is in nature
the measured signal was taken from the amplitude and coherent
type and because of that the variation can be greater than
radar channel simultaneously. For further signal processing,
mean value. This leads into situation, that statistical result may
512 pulses with given radar PRI were taken. Due to the nature
differ quite a lot from single practical shot of detection.
of the background clutter we finally used the coherent radar
Relatively high variation of RCSDr comes primarily from the
channel with Doppler resolution 8Hz for SNR calculations. All
drone arms material, structure and their relative angle.

615
signals were digitally processed with respect to the radar signal battlefield radar processing presented in section III showed
parameters and statistically evaluated. good agreement between theoretical analysis and expectations
described in section II. The drones with potential payload
Fig. 6 shows power spectrum belonging to the moving 0.3÷1kg and above e.g. High One and X8 can be detected by
High One drone at the distance 1020m. The Doppler peak at Ku-band short range pulse battlefield radar with sufficiently
200Hz represents target moving speed and some other peaks high probability detection PDDr at the distance from 1km to
are characteristic of the drone propellers. 2km. Detection probability of the smaller drones (e.g. Iris+ or
From the statistics of the coherent radar channel for each the drones with intensively reduced SNR) can be below 0.5
detected moving drone event we estimated particular SNR with with constant false alarm value PFaDr =10-5.
respect to the measured distance [6]

 E (PS ) 
SNR[dB ] = 10 log  , (7)
 E (PN ) 

where E(PS) and E(PN) represents the mean power of signal X8


and the mean power of noise at the receiver – detector output H_O
respectively. Results of the SNR estimation obtained from Iris+
coherent channel measurement are summarized in the Table
III.

TABLE III. STATISTICAL RESULTS OF THE SNR ESTIMATION


Parameters of received signals
Drone
SNR SNR[dB]
X8 231 ± 92 18 ÷ 26
High One 69 ± 21 16 ÷ 20
Iris+ 7±2 6 ÷ 10
Figure 7. The graph of drones PDDr in dependence from the PFaDr

C. Radar Range and Probabilistic Characteristics Estimation ACKNOWLEDGMENT


For the probabilistic characteristic’s estimation we assumed This work is funded by the MoD of Slovak Republic under
Gaussian distribution of the amplitude of reflected signal Grant Agreement 852_08-RO02_RU21-240 “UAV Detection
instead of Rayleigh one and for the span PDDr = (0.6 ÷ 0.95) and Effective Destroying as Protection of Buildings and
SNR[dB] may be increased by δ [10]. Then resulted value for Special Institutions by Armed Forces”.
increased SNRI[dB] is
REFERENCES
SNRI [ dB] = δ SNR[ dB] = ( 3.743 PD2 )
− 4.28 PD + 2.43 SNR[ dB] (8) [1] Drone Detection & Defense Systems, DeTect, Inc., Available at:
http://www.detect-inc.com/drone.html, 04.01.2017
[2] V. Vlasak, J. Pidanic, "The analysis of small RCS target detection in
The drone detection probabilities PDDr calculated from the primary radar system," ELMAR 2016 International Symposium, Zadar,
eetimated values of SNR related to the chosen false alarm 2016, pp. 141-145
probability PFaDr are in the Table IV. [3] R. Yang, Y. Wang, C. Yu and Y. Han, "A probabilistic early collision-
warning scheme for UAVs in 3D space," ISCIT 2016 International
Symposium on Communications and Information Technologies,
TABLE IV. DRONE DETECTION PROBABILITY
Qingdao, 2016, pp. 551-554
Minimum PDDr Maximum PDDr [4] W. Kong, D. Zhang and J. Zhang, "A ground-based multi-sensor system
PFaDr
Iris + H_O X8 Iris + H_O X8 for autonomous landing of a fixed wing UAV," IEEE International
10-5 0.234 0.936 0.966 0.644 0.982 0.997 Conference on Robotics and Biomimetics ROBIO, Zhuhai, 2015, pp.
10-6 0.175 0.924 0.959 0.589 0.978 0.996 1303-1310
[5] J. Drozdowicz, M. Wielgo, P. Samczynski, K. Kulpa, J. Krzonkalla, M.
From the measurement results of the drones and statistical Mordzonek, M. Bryl, Z. Jakielaszek, "35 GHz FMCW drone detection
results shown in Table III and Table IV we can finalize the system", IRS 2017 International Radar Symposium, Krakow, 2016, pp.
graph of PDDr (Fig. 7) calculated using the formulas (2) and (6) 1-4
in spite of its dependence on the PFaDr for the practical usage. [6] M. I. Skolnik, Introduction to Radar Systems, 3rd ed., McGraw-Hill
Inc., New York, 2001.
Comparing achieved results related to chosen drones we [7] S. Spirko, Z. Matousek, “Radar Interception Area Analysis”, Cybernetic
may see a good agreement with theoretical expectations in Fig. Letters, Vol. 1, 2008, pp. 1-10.
3 especially for the drones High One and X8. [8] D. Barton, S. Leonov, Radar Technology Encyclopedia, Artech House,
Inc. Boston, 1997.
IV. CONCLUSSIONS [9] H. W. Cole, Understanding Radar, Second Edition, Blackwell Scientific
Publications, Oxford, 1992.
The experimental RCS measurement of the drones Iris+,
High One, X8 and received signal from short range pulse [10] J. Ochodnicky, Z. Matousek, Radar System Analysis, Armed Forces
Academy, Liptovsky Mikulas, 2013 (in Slovak).

616

You might also like