NATO UNCLASSIFIED

RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

D. Fränken, S. Stroth, O. Zeeb
HENSOLDT Sensors GmbH
Woerthstrasse 85, 89077 Ulm
GERMANY
[email protected]

ABSTRACT
Passive radar (PR) systems rely on third-party transmitters like, e.g., analog or digital audio or video
broadcasting stations to illuminate targets of interest. Lacking own emitting sources, these systems are fairly
lightweight and mobile. They can be set up at arbitrary locations including densely populated areas, and
inherently they possess advantageous stealth properties. The creation of an overall target situation picture is
based on the evaluation of both the reflections from the target and the transmitted signals received directly
from the illuminating transmitters. Signal processing delivers plots with bistatic range, Doppler and
favourably azimuth and elevation, depending on the applied antenna technology. Tracking and data fusion
are subsequently applied to create the situation picture. This paper discusses tracking and data fusion in an
integrated FM/DAB/DVB-T PR system for airspace surveillance tasks and presents typical effects in passive
radar systems that are not familiar from classical monostatic radar tracking.

1 BISTATIC DATA
In classical monostatic radar systems, the emitter and receiver of a signal are positioned at the same location.
Passive radar systems, however, use signals of opportunity from fixed antennas in different locations. Thus,
tracking in passive radar systems differs from classical radar systems. The geometry and the nature of the
bistatic measurements, which are not Cartesian complete, imply several consequences [1, 2].

1.1 Passive Radar Tracking in General: From Bistatic to Cartesian Data.

In a classical Radar, the emission time of a signal is known, so the time until an echo is received defines the
euclidian distance from the target to the radar sensor. Additional measurement of the azimuth and elevation
make an observation cartesian complete, i.e. the 3D-location of the target is known. However, since passive
radar systems rely on third-party transmitters, the signal emission time is unknown and bistatic plots
comprise the Time-Difference-of-Arrival (TDoA), i.e., the time difference between the direct signal and the
reflected signal. The TDoA can be converted into a bistatic range. Such a bistatic range measurement of a
target defines an ellipsoid (or in 2D an ellipse) with the receiver and the transmitter as focal points and the
target on the ellipse, see Figure 1 (red ellipse).

STO-MP-SET-241 4-3 - 1

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

Receiver
Transmitter
Target
Direct Signal
Reflected Signal

Figure 1: Bistatic geometry.

Without further information, the target that reflects the signal could be anywhere on the ellipse, the cartesian
position can not be uniquely determined. If additionally the reflected signal from a second illuminator (blue)
is measured, the intersection of these two ellipses defines possible target locations. Note, that still the target
position is not uniquely determined since in general there are multiple intersection points, see again Figure 1:
the true target location at the yellow marker and the additional intersection a bit below.

The addition of more target measurements originating from different illuminators, i.e., the intersection of
n>2 ellipses, could define the target location uniquely as the single point which all ellipses have in common,
see Figure 2.

Figure 2: Intersection of n ellipses (n=4).

4-3 - 2 STO-MP-SET-241

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

occurring mathematical problems in 2 dimensions, altitude hypothethes can be introduced. Instead of

investigating the underlying 3D-problem, a set of 2D-problems can be extracted [7]. This way, the 3-
dimensional world is cut into “slices” with different altitudes, solutions invoking the intersection of 2D-
ellipses are calculated on these slices and afterwards combined again to 3D target reports.

3 REAL DATA
In this chapter we present measurement data and effects that occur in PR tracking. The underlying data was
captured during a measurement campaign around Munich in 2015. Three PR receivers were involved which
captured data from FM, DAB and DVB-T illuminators. All data was fused in a single PR tracking station.

3.1 Bistatic Plot Data

Figure 7 shows bistatic data (in Cartesian space) accumulated over 10 seconds. It has proven to be useful to
plot not only the locations of the bistatic observations but also small ellipse segments which are defined by
the locations of the illuminator and receiver, the bistatic range and azimuth. This way, intersections of
several ellipses can be identified as target locations. The figure also visualizes with real data what was
already shown in Figure 4: depending on the location geometry of sensor, illuminator and target the ellipse
segments, which originate from the azimuth uncertainty, appear smaller or larger.

2 possible target locations

Figure 7: Bistatic Data.

STO-MP-SET-241 4-3 - 7

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

3.2 R-RR-Space
In Figure 8, the received data in Range-RangeRate-Space accumulated over 30 seconds is shown. Stage 1 of
the tracking algorithm is performed in this space (cf. Chapter 1.2). Traces can be identified and also the sharp
(pseudo)maneuvers mentioned in Chapter 2.1 can be recognized.

Sharp maneuver

Figure 8: Range(y)-RangeRate(x)-Space.

3.3 Generation of 3D-Tracks

As a consequence of the 3-Stage-Tracking algorithm described in Chapter 1.2, one typical effect occurs
when targets are detected for the first time, e.g. after take-off. This effect is visualized in Figure 9: First, R3-
Tracks (black) are initiated (a). When there are enough of them, they can be fused to 3D-Tracks (red) (b+c).
Because of the feedback loop, bistatic plots are then directly associated to 3D-Tracks instead of R3-Tracks.
Since the R3-Tracks are no longer supported by bistatic measurements, they are dropped whereas the 3D-
Tracks are stabilized (c+d).

4-3 - 8 STO-MP-SET-241

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

(a) (b)

(c) (d)

Figure 9: Generation of 3D-Tracks.

3.4 Air Situation Picture

Figure 10 shows the air situation picture accumulated over 120 seconds. 3D-Tracks generated by the passive
radar cluster are shown in black, ADSB reference data is shown in green. Generally, the ADSB-targets are
well-captured. In the lower left corner there are three targets without a 3D-Track, additionally there exist 3D-
tracks without an ADSB-reference. The accuracy of the tracks can be improved further in some cases. These
are ongoing tasks that are under recent investigation.

STO-MP-SET-241 4-3 - 9

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

Figure 10: Air Situation Picture.

4 CONCLUSIONS
In this paper an overview of the principles of passive radar tracking is presented. The effects are shown in
pictures rather than in complicated mathematics and formulas. The basic architecture of the tracking
algorithm is described. Challenges occurring from the Cartesian-incomplete nature of the bistatic data and
additional complications for the data association in single frequency networks are mentioned. After the
theoretical considerations, results from real world data are presented.

5 ACKNOWLEDGMENT
The authors wish to thank Dr. Michael Wiedmann for many inputs to this work and for many valuable
discussions.

4-3 - 10 STO-MP-SET-241

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

6 REFERENCES
[1] Cherniakov, M, “Bistatic Radar: Principle and Practice”, Wiley, 2007.

[2] Willis, N.J., Griffiths, H.D., “Advances in Bistatic Radar”, SciTech Publishing, 2007.

[3] Edrich, M., Kloeck, C., Stroth, S. “A Passive Radar System for Long Range Air Surveillance”, SET-
241 9th NATO Military Sensing Symposium, Quebec, 2017.

[4] Malanowski, M., Kulpa, K., “Two Methods for Target Localization in Multistatic Passive Radar”,
IEEE Transactions on Aerospace and Electronic Systems, 2012.

[5] Wiedmann, M., Deininger, R., Schröder, A., Zimmermann, R., „Multi Transmitter Fusion for Tracking
in a Network of Transmitters of Opportunity“, SET-169-41 Symposium on 8th NATO Military
Sensing Symposium, Friedrichshafen, 2011.

[6] Daun, M., Koch, W. “Multistatic Target Tracking for Non-Cooperative Illumination by DAB/DVB-T”,
IEEE Radar Conference, Rome, 2008.

[7] Daun, M., Kaune, R. “Gaussian Mixture Initialization in Passive Tracking Applications”, 13th
Conference on Information Fusion (FUSION), Edinburgh, 2010.

STO-MP-SET-241 4-3 - 11

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE
 NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE

Tracking and Data Fusion in a Passive Radar System

4-3 - 12 STO-MP-SET-241

NATO UNCLASSIFIED
RELEASABLE TO AUS, FIN AND SWE