J.L.

Gertz
I

Multisensor Surveillance for

Improved Aircraft Tracking

Cross-range measurements of aircraft travelling a t distances of 50 to 200 miles include

significant errors. Therefore, heading estimates for medium-to-long-range aircraft are
not sufficiently accurate to be useful in conflict-detection predictions. Accurate cross-
range measurements can be made-by using two or more sensors to measure aircraft
position-but such measurements must compensate for the effects of system biases and
aircraft turns. A set of algorithms has been developed that are resistant to system biases,
that detect turns, and that track successfully through both biases and turns. These
algorithms can be incorporated into a complete multisensor system, with good intersen-
sor correlation of aircraft tracks and no added delays to the air traffic control processing
chain.

Single-Sensor Tracking reports are assumed to be measurement noise

and, even after a turn has been identified, the
The Mode S sensor has brought major im- true turn rate cannot be accurately estimated.
provements to the accuracy of aircraft surveil- Figure 2 shows how a typical linear Kalman
lance reports [I].With this sensor, the expected filter handles turning data. Note how severely
measurement noise has a standard deviation (0) the predicted aircraft heading lags the true
value of 25 feet in range by 1 rnilliradian in heading.
azimuth (if the constant bias offsets are ig- These tracking problems are exacerbated
nored). However, this high degree of accuracy when the aircraft enters a sensor diffraction
still introduces considerable error in the mea- zone. In such a zone, the radar signal is dif-
sured cross-range (position component in the
azimuth direction). The 1-mrad noise in azi-
muth translates into a cross-range o many
50

..
I I I
times larger than the range ofor distant targets.
For a range measurement with a standard
deviation of 25', for example, the standard devia-

..
@@*a
tion of the cross-range can be fifty times greater: -
-1 4 9 -
@*.a.
Range of target 1- o range 1- a cross-range
10 nmi 25' 60' 5
- @@a
50 nmi
100 nmi
200 nmi
25'
25'
25'
300'
600'
1,200'
.-w8
i
48-
@@a
@a
@@ .
@a
@a:a
-
Large cross-range errors cause very noisy
tracker inputs. The sample data of Fig. 1 illus-
trate the severity of the problem. The data pro- I I I
47
duced with the current, single-sensor system -25 -20 -1 5
contain a great deal of noise. To produce stable
Nautical Miles
heading estimates, therefore, the data must be
heavily smoothed by a Kalman filter. But such a Fig. I-The current single-sensor sun/eillance system
tracker follows turns very poorly-initial turn produces extremely noisy data.

?he Lincoln Laboratory Journal, Volume 2, Number 3 (1 989)

Gertz -Multisensor Surveillancefor Improved Aircraft Tracking

multisensor measurements to improve surveil-

lance and tracking. The current air traffic con-
trol system makes use of multiple-sensor cover-
age, but only a s a backup mode when primary-
sensor coverage is missing. That is, the current
mosaicking system switches its data source
from preferred to supplemental only when re-
quired; only one sensor's reports are used on
any given scan, and other available data are
ignored.
74 75 76 77 78 A true multisensor system would provide far
Nautical Miles better surveillance and tracking of aircraft than
the current system. For this reason, an ongoing
program at Lincoln Laboratory has developed a
Fig. 2-Kalman filter performance during a turn. The "1 '3 set of algorithms and has implemented them in
show the true aircraft position. Performance of the filter is a system that offers multisensor processing of
poor: the head of each arrow should touch the tail of the
succeeding one.
aircraft reports.

fracted by a narrow object (such a s a smoke- Multiple-Sensor

stack).The resulting beam curvature introduces Tracking Approaches
verylarge errors in the azimuth estimates. In the
worst case, in fact, the azimuth value read by the Multiple-sensor tracking can be imple-
sensor is randomly distributed over the full mented in one of two different ways. The sim-
antenna beamwidth. plest method is to feed unmodified reports from
In a real-life application, diffraction errors the sensors in time order into a common Kalman
can be dangerously misleading. The results of filter. Each report then improves the tracker
the conflict-detection experiment shown in Fig. covariancematrix (thematrix of standard devia-
3 clearly demonstrate the hazard of diffi-action tions of all pairs of measured variables) along its
errors. The measured positions and headings range coordinate. If the sensors are at different
look safe; the actual trajectories are heading for aspect angles relative to the aircraft, the entire
collision. covariance matrixwill tighten and good tracking
Since aircraft are typically observed by two or will result.
more sensors, it should be possible to use The second approach is multilateration. In
this method, illustrated in Fig. 4, range mea-
23 surements from the sensors are time aligned to
I I I I
a common time, and position is determined by
Aircraft Azimuth - 297"
the intersection of the range arcs. Assuming dif-
i Obstacle Azimuth - 295.9"
ferent aspect angles, the measurement error el-
g 22- J0
+
(I) lipse becomes nearly circular. The improved
CC

c
0 reports are then fed into a Kalman filter.
r Multilateration offers four advantages:
2
.- 21 - \ (1) improved accuracy in measurements
Actual Track
E DAB 101 Actual Track shown on the controller display,
DAB 601 (2) quicker detection of aircraft turns,
2o 4; I 1 I (3) no diffraction errors, and
-42 4 1 4 0
(4) estimates of aircraft altitudes.
nmi West of Sensor
For these reasons, Lincoln Laboratory has
pursued most strongly the multilateration
Fig. 3-Surveillance degradation due to diffraction. approach.

382 The Lincoln Laboratoy Journal,Volume 2,Number 3 ( I 989)

 Gertz -Multisensor Surveillancefor Improved Aircraft lYacking

ample of an error due to linear extrapolation

Cross- Range
during a turn, an error so large that it nullifies

/
' )'/
Error

Resultant
Positional Error
the improvements of multilateration. Thus
turns must be identified prior to time alignment,
and circular interpolation and extrapolation
must then be used during turning periods.

Turn Detection
Turns are usually detected by observing a
heading change in the aircraft's trajectory.
However, the heading measurement depends
upon the azimuth, and severe noise in the
Sensor 1 Sensor 2 measurement of the azimuth is the reason that
multisensor surveillance is needed. Thus the
Fig. 4-Multilateration error ellipse. turn-detection technique for multilateration
must use only measured ranges. The most
successful such technique uses the second dif-
,Worst-Case Interpolation Error ference of successive range measurements, as
shown in Fig. 6. The actual observed A(Ap) is

NAP), = (P, - pn-1) - (P,-1 - pn-2)

where nis current scan. This value is compared

Worst-Case Extrapolation Error 1 with the predicted one for a straight trajectory.
If the difference exceeds a predefined parameter,
a turn is declared to be in progress.
This technique was applied to a sample of real
Speed Int. Error Ext. Error
data; the results are shown in Fig. 7. Note that
true turns were detected, but a false apparent

Straight I

Fig. 5-Interpolation and extrapolation accuracy during

3"/s turn.

Time Alignment for Multilateration

Both interpolation (estimating between two
data points) and extrapolation (estimating be-
yond the last data point) are used to time-align
sensor measurements. Interpolation is always v* 7 2
more accurate than extrapolation. A ( Ap) Expected = Ap2 - Apl = 7 sin (h- 8 )
Interpolation and extrapolation are both A (Ap) Actual = dp2- dpl
accurate when aircraft are flying straight, but Turn: I A ( Ap) Expected - A (Ap) Actual I >Parameter
linear extrapolation has severe error potential
during aircraft turns. Figure 5 shows an ex- Fig. &Turn-detection algorithm.

The Lincoln Laboratory Journal, Volume 2, Number 3 U 989)

Gertz -Multisensor SurueUlancefor Improved Aircraft Tracking

turn caused by noise did not trigger the detector.

This technique cannot detect all turns. In
particular, it can be insensitive to radial turns.
However, turn detection is needed only to pro-
duce accurate range extrapolations. The turns
that this approach misses are just the ones in
which straight and circular extrapolation pro-
duce nearly the same result, and thus no harm
results.

urn
A = ~etected

.
= No Turn Detected I
1
em. 0 .
I . , .
A Fig. 8-Coordinate-transformation cases: (1) Conversion
A Apparent A from 6 , 4 top2, ;(2)Determination of d for mulfilateration.
8
Turns A

earth transformation formulas, using the aver-

age earth's radius over the region of interest.
Our studies, shown in Table 1, have shown that
this approach is incorrect. The proper radius to
use in the spherical equations is the local radius
of curvatureof the earth. This quantity matches
the earth surface shape rather than its distance
Nautical Miles
to the true earth center and is largest at the
poles. By contrast, the local earth radius is
largest at the equator.
Fig. 7-Single-sensor range-only turn detection. Accuracy can be further increased by per-
forming multiple transformations. In Fig. 9 the
measurement from a distant sensor is trans-
Coordinate Transformations formed step by step at a constant longitude
(variable latitude) to the latitude of the near
For two reasons, multisensor processing sensor, and then transformed in one longitude
requires very accurate coordinate transforma- step over to the near sensor location. Each step
tions. First, if the preferred-sensor report is uses the local radius of curvature applicable to
absent, reports from supplemental sensors the latitude at the center of the step.
must be transformed into the primary-sensor Mathematically, the multiple steps are a
coordinate system. Second, the distance from multiple convolution of the transformation
one sensor to another is the base of the compu- equations. Thus a single transformation matrix
tation triangle for multilateration. Figure 8illus- can be defined for each sensor pair that de-
trates these two tasks. scribes the overall result in one step, as illus-
Since the transformation procedure requires trated by the equation in Fig. 9. In actual appli-
knowledge of the earth's radius, it must be cation, run-time conversions are always single
known exactly.The earth is not a true sphere; its operations, independent of the number of steps
radius varies with latitude. Most coordinate- employed. The 1 1-step transformation shown in
transformation procedures apply the spherical- Table 1 demonstrates the success of this ap-

7he Lincoln Laboratory Journal, Volume 2, Number 3 (1 989)

 Gertz -Multisensor Surveillancefor Improved Aircraft7kacking

look at the built-in biases of the system.

All sensors have intrinsic sources of mea-
42.48" N surement bias. The measured aircraft range
depends on delays in the sensor's electronics, in
clock errors, in signal refraction, and by the
transponder turnaround delay error. The air-
craft azimuth can be corrupted by north-mark
error, antenna tilting, or signal diffraction. In a
single-sensor system, only the transponder
delay causes a relative error between two air-
craft positions; the rest of the errors simply shift
the positions of all aircraft equally and therefore
..Site B
..X2
do not affect separation monitoring. Since the
transponder bias only affects range and is al-
ways small, system biases are of no conse-
quence in a single-sensor system.
The situation changes markedly when
multisensor data are used. First, a new class of
biases appear: sensor-location errors and other
errors that directly produce intersensor regis-
tration errors. Second, the biases of different
Fig. 9-Three-step coordinate transformation. sensors will produce different results, so that
when different sources supply data for two air-
proach in reducing the coordinate-transfonna- craft, the relative aircraft separation is compro-
tion error to essentially zero. mised. Finally, the transponder-delay bias has a
greater effect when multilateration is used.
Range errors are now transformed directly into
System Biases
azimuth errors, as shown in Fig. 10.
An accurate cross-range measurement sys- The graph in Fig. 11 illustrates the effect of
tem must compensate for the effects of both these biases. These measurements were taken
turns and biases. Now that we have addressed from the current ATC mosaicking system-the
the problem of turn recognition, we must take a hops due to changes in the data source are so

Table 1. Location after Transformation from Site A

to Site B Coordinates

11 Steps 1 Step

Local-
Radius A = -3993nmi A = .3993nmi
Method

Radius-of-
Curvature A = .0002 nmi A = .I823nmi
Method

The Lincoln Laboratory Journal. Volume 2. Number 3 (1 9891

Gertz -Multisensor SurveilZancefor Improved Aircraft nacking

126 I I I I I
380 400 420 440 460 480
Nautical Miles
Sensor Measurements
Do Not Meet Fig. 1 I-Mosaic-system registration errors.

Incremental Bilateration
The standard scenario for a two-sensor sys-
tem with biases is illustrated in Fig. 13(a).Each
sensor is located at a known fixed position, and
it can specifjr the location of the aircraft. The
difference in the two aircraft locations results
from the biases of the sensors. When bilatera-
tion is used, the calculated aircraft position
(where the range arcs intersect) will be at yet a
third location. Thus, whenever the measure-
Adjust by Changing ment source changes from one scan to the next
the Target Azimuth (sensor 1 only, sensor 2 only, or bilateration),
the aircraft position will hop.
An alternate view of this scenario is presented
Fig. 10-Multilateration resolution of range biases.
in Fig. 13(b).In this method, called incremental

large that aircraft tracking has become nearly I I I I

impossible. 24 I 4 I4
1 : Claxby
Procedures to remove registration biases by 2 2: Debden
4
fitting large quantities of aircraft data to bias : Pease Pottage
l 2 2
models have been developed. However, no such - 1
procedure can remove all bias effects, because 1 *4
not all biases can be modeled. Figure 12 pres- 1 24
ents data from one system after bias removal. 4
4
The three sensors clearly give different results - 1 2
for the single aircraft. 1 2
Biases in a multilateration system can easily I I i I I I
produce aircraft azimuths less accurate than -79 -78 -77 -76 -75 -74 -73 -72
those of the original single-sensor system.
Nautical Miles
Thus a multilateration algorithm must be bias
resistant. Fig. 12-Raw sensor reports.

7he Lincoln Laboratory Journal, Volume 2, Number 3 (1 989)

 Gertz -Multisensor Surveillancefor Improved Aircraft Tracking

bilateration, sensor 1 is the master sensor. Its

location is fixed as before, and its aircraft mea-
surement location is taken to be the true loca-
tion. Sensor 2's measurements are used to
locate sensor 2; that is, they are reversed, start-
ing at the aircraft location and going back to
earth. The resulting apparent position for sen-
sor 2 differs fi-omthe real one by a n amount that
compensates for the system biases.
Incremental bilateration offers data consis-
tency-as long a s sensor 2 is assumed to be at
its apparent location, sensor 1 measurements,
sensor 2 measurements, and bilateration will all
agree on the aircraft location, and data switch-
ing will not produce hops. The aircraft location
may, in fact, be in error because of sensor 1
biases, but all aircraft that have sensor 1 as
the master sensor will be in relative agree-
ment. Thus this method removes all system
bias effects.
Incremental bilateration is bias resistant.
That is,
(1) the measurement positional bias is iden-
tical to that seen by using the master
sensor as a single sensor, independent of
any intersensor registration errors; and
(2) the azimuthal measurement noise with
incremental bilateration is less than that
seen with single-sensor surveillance.
The major problem with incremental bilatera-
tion is the requirement that the secondary-
sensor position be calculated for each scan and
for each aircraft fi-om the reverse sensor mea-
surements. Because the earth is a sphere, the
computations are very complex and require an
iterative solution. The next section resolves this
problem.
Also, since the apparent location is a function
of system biases, which are themselves h n c -
tions of geometry, the apparent position will
slowly move during any aircraft flight. A time-
smoothing procedure, called outlier desensiti-
zation, therefore must be used. Figure 14shows
the raw and the smoothed data for the change in
apparent distance between two sensors over
time for a sample flight. As shown in Fig. 15,
Fig. 13-Bilateration scenarios. (a) Standard. (b) lncre-
outlier desensitization weighs outliers (bad data mental. S1 represents the location of sensor I and S2 the
points) less heavily than expected points and location of sensor 2.

The Lincoln Laboratory Journal, Volume 2. Number 3 (1 989)

Gertz -Multisensor Surveillancefor Improved Aircraft Tracking

I I I I
1 - 8 -
8

0.5 - ' 1 :-

* *t
I
I Data Value - Old Estimate I
-1 - b -
I I I L Fig. 15-Outlier-desensitization smoothing algorithm. The
Scan 0 50 100 150 200 red curve is the function, f, described in the equation: new
No. Raw lntersensor estimate - old estimate = f (data value - old estimate).
Range Estimates

because errors from the approximation would

I I I I
1 - - compromise the accuracy of the bilateration. In
addition, the model must allow the use of planar
- - mathematics directly on the p, 8 sensor mea-
0.5
w
C surements; a model requiring measurement
L
transformations would add time-consuming
0 - calculations to the approach. In particular, the
Iij
3 model must permit the use of the representation
-0.5 - - shown in Fig. 16, where

-1 - - pl , 81 = sensor 1 measurements
I I I I p2, O2 = sensor 2 measurements
Scan 0 50 100
Outlier-Desensitization
150 200
p, = ground range = d
m
Estimates zi = height above sensor i (computed
from altimeter reading)
d = apparent distance between the sensors
yo = apparent azimuth of sensor j from sensor i.
Fig. 14-Incremental bilateration smoothing.
Using this model, we can easily compute the
thus produces smoother results. Outliers can- apparent secondary sensor location from the
not be ignored; if they were, and if the smoothed raw sensor measurements:
position drifted far fi-om the true position, all
h t u r e good points would be rejected. But the d = I,/(&sin 4 - p2 sin o2l2 +, ( cos O1 - p2 cos e2I2
desensitization approach always returns to the
true position after an error.

Spherical-Equivalent Flat Earth where the correct sign is a function of the actual
sensor geometry. In actual use, one sign will give
The key to the use of incremental bilateration a physically plausible result and the other will
is the development of a flat-earth model that is be implausible, but the choice of the correct sign
exactly equivalent mathematically to the spheri- can not be predetermined.
cal-earth model. No approximation is possible, Conversely, once the apparent secondary-

'Ihe Lincoln Laboratoy Journal, Volume 2, Number 3 11989)

 Gertz -Multisensor Surveillancefor Improved Aircraft 7Yacking

sensor location is known via the smoothing mine the occurrence of turns; after a turn is
process, the true aircraft azimuth can be deter- detected, the Kalman filter is restarted or
mined by bilateration: adjusted.
An improved tracker has been developed as
part of the multisensor project. This tracker
assumes that aircraft flight can be modeled as
The property of a planar system that permits having a constant turn rate. (Straight flight is
the use of these simple equations is the align- modeled a s a constant turn rate of zero.) The
ment of the x- and y-coordinate systems of the advantage of the constant turn rate is that the
two sensors. This alignment can be expressed filter successfully tracks aircraft through turns
for any aircraft location as without the need for external turn detectors or
filter adjustments. The constant-turn-rate Kal-
man filter has five state variables:
1 1
Ixl - x2 = Ifi sin Ol - p2 sin O2 = x component of d
x = x position
lYl - y2 1 = Ipl 1
cos el - p2 cos O2 = y component of d. y = y position
h = heading
By studying the spherical coordinate-trans-
h = turn rate
formation equations, and seeking to match the
u = velocity.
alignment of the x- and y-coordinates, the
Spherical-Equivalent Flat-Earth Theorem (see The equations of motion that define the filter
the box) can be proved [2]. can be specified in derivative form:
This theorem applies even if the aircraft x = u sin(h)
altitude his unknown. For any estimate of h, the y = ucos ( h )
spherical and spherical-equivalent flat-earth F; = F;
models produce identical results. i;=o
One apparent problem with the theorem is
v = 0.
the need to determine po. A precise calculation of
p, requires spherical-earth mathematics, which The derivative definition leads directly to a
we are trying to avoid, and knowledge of O,, nonlinear extended Kalman filter form of solu-
which we are trying to calculate. Fortunately, po tion. Unfortunately, this approach is very com-
needs only be known approximately to intro-,
duce no altitude error:

North

Aircraft Tracking
To generate expected aircraft trajectories, the
surveillance reports generated by the multisen-
sor-processing algorithms are entered into a
smoothing filter, or tracker. Since the predic-
tions are used for conflict detection, a good
tracker is essential.
Most current FAA multisensor tracking
employs a Kalman filter and assumes straight
s2
flight dynamics. Tight filter gains are used to
prevent bias effects from causing heading vari-
ations. Thus external turn detectors must deter- Fig. 16-Flat-earth sensor location.

l'he Lincoln Laboratory Journal,Volume 2. Number 3 (1 989)

Gertz -Multisensor Surveillancefor Improved AircrajP Tracking

Spherical-Equivalent
Flat-Earth Theorem

Given. a two-sensor system defined by and p, is the slant range of the

target measured from the laca-
(a) A spherical-earth model tion given by latitude (A1 +-Q/2,
(b) Sensor 1 located at latitude A,. longitude (y, + y2)/2. and height
longitude y,, and height h,, h = 0 of the ofigiml spherical
(c) Sensor 1 reading measurements earth model
p l . 8,. and b9 Seneor 2 reading the same
measurements p2, 0,. and 3 as
in (4

(d Sensor 2 located at latitude A2,

longitude y2. and height h,. pro-
ducing a sensor position at range
d, and azimuth ty12 relative to
sensor 1
(e) Sensor 2 reading measurements
p2. 8,. and

Then=an equivaknt two-sensor system

canbe defined by

@ Aflat-earthmodel
@) Sesmt 1located at x,,
=0
- us, = z,,

(h) Sensor 1 reading the same

measurements p,, 8,. m d a, as
(9
(0 Sensor 2 located at xe9 ~IQ,and
9, SO, wit.hx,andgsthedues
that locate sensor 2 relative to
sensor 1 at the wane m u t h yr,,
as in (4, and at the adjusted
-'=

where
Fig. A-(a) Spherical earth. (b) Sphen'cal-equivalentflat
earth.

The Lincoln Labomtory Journal. Volunae 2. Number 3 (1 989

 Gertz -Multisensor Surveillancefor Improved Aircraft ZTacking

plex and time consuming, and can lead to un- where each partial derivative is computed from
stable and divergent results. A linear Kalman the corresponding update equation. For ex-
filter is needed. ample, using Eq. 1, we find that
Fortunately, the derivatives expressed above
are all integrable. For example, the x update
equation can be expressed as
x ( t + T )- x ( t )=

= joT u sin ( h + &)da 6h

=
sin ( hT )
vT[cos(h)-+ s i n ( h )cos(hT)- 1
h~ hT 1
and so on.
= joTv[sin ( h )cos (&) The $ matrix can then be used by the linear
+ sin ( h a )cos ( h )Ida Kalman filter to calculate the covariance matrix
v for one-step prediction. In the absence of noise,
= - {sin ( h )sin ( h T ) the covariance update becomes
h
- cos ( h ) [ c o (s h T )- 11)

The update equations can be expressed as

where Pis the state-variable covariance matrix,

sin ( F;T
sin(h)-
hT
)
-cos(h)
cos ( h ~- )1
hT I

The steps in the recursive Kalman filter

method are then the same for the turn-rate
Kalman filter a s for the standard linear Kalman
filter [3].That is, starting with a n estimate X and
For a linear Kalman filter the update equa- its covariance matrix P, and after receiving a
tions must be expressed in matrix form as new report Ymeas,a new estimate Xsm(smoothed
X ( t + T ) = t) ( t )X ( t ) . values of X) and its covariance matrix Psm are
Clearly the preceding equations cannot be ex- obtained, which are then used for the starting
pressed in this form-x and y are not linear values of the next scan.
functions of t. However, by using partial deriva- To cover aircraft deviation from the assumed
tives, an approximate form of the $ matrix can be constant-turn-rate flight, noise components
built: still must be added to the turn-rate Kalman
filter. The independent noise variables chosen to
represent the accelerations are

h = change i n turn rate

v = change in velocity.

These variables modify the update equation by

increasing the prediction uncertainty.

.lheLincoln Laboratory Journal, Volume 2, Number 3 (1989)

Gertz -Multisensor Surveillancefor Improved Aircraft Tracking

Sensor 1
Mode S
L'SR-9
looooooo Seneor N
/CRBS ,
Beacon Radar Beacon

8
a with
Reports-------------------------.------
Reports--------------...
r------------.--
with
Reports "...---.---------. 8
8
8
8
a TRK Nos. TRK Nos.
8
8
8
8
8
8
8
8
8
8

Intersensor Correlation
Track Number Cross-Referencin

8
8
I8 for Each Aircraft
8
8
8
8
I
I Multisensor Processing
8
8
8
8
8
8
#
I
8
8
8
8
8
8
8
8
8
8
8
8
I
8

Fig. I 7-Multisensor-processing system. Thesystem is divided into threeparts: intrasensorprocess-

ing, intersensor correlation, and multisensor processing.

For the linear filter, the new equation be- and r is analogous to $in the way it relates noise
comes values to state variable updates. A linearized
noise model of I' has been developed.

where Q is the noise covariance matrix:

Multisensor-Processing System
The algorithms presented in this article have
been implemented in a multisensor-processing
system. As shown in Fig. 17, the system is
divided into three segments: intrasensor pro-
The Lincoln Laboratory Jounml, Volume 2, Number 3 (1 989)
 Certz -Multisensor Surveillancefor Improved Aircraft Packing

cessing, intersensor correlation, and multisen- supplemental sensor with a higher rotation rate
sor processing. may produce two or more reports during this
Intrasensor processing correlates and tracks period.)
the reports from one sensor. If the sensor is To meet the delay constraint imposed on the
Mode S, this function has already been per- multisensor system, the release of the report
formed. Other sensors, such as the existing FAA packet when the preferred sensor sees the air-
radar and beacon sensors, provide no such craft is critical: the multisensor system must not
service or only partial service. delay the output of surveillance reports. Since a
Intrasensor processing must also time-con- single-sensor system would obviously send its
trol the output of reports to intersensor correla- report when it sees an aircraft, the rule for the
tion. Therefore, it holds its reports until a fixed multisensor system insures that the output
delay has been reached; this delay is the same as times of the reports match those of current
the delay that would be experienced by reports systems, and add no delay.
from a Mode S sensor, because it has the great- The multisensor-processing subsystem de-
est internal delay. The delay function thus termines multilateration positions by operating
guarantees that the stream of reports from the on the reports in each packet. The subsystem
various sensors entering intersensor correlation then tracks and filters the data, and supplies
is in correct time sequence. accurate position and heading estimates to the
Intersensor correlation has the task of air traffic controllers.
matching the tracks from different sensors that
correspond to the same aircraft. It also creates Intersensor Correlation
packets of reports for each aircraft and passes
them to the multisensor-processing functions. The heart of the intersensor processing sys-
The packets contain all reports from all sensors tem is the intersensor track correlator, the rou-
received during a scan of the preferred sensor. (A tine that determines which tracks from the

Local
Tracked
Report

Use Positions and

Velocities to Find
Potential Matches

Yes
Select
Best Match ,
No
7
Update Initialize
Track

Fig. 1&Intersensor track correlation.

The Lincoln Laboratoy Journal, Volume 2. Number 3 (1 989)

Gertz -Multisensor Surveillancefor Improved Aircraft Tracking

Correlation Box
(c) agree in code (beacon tracks only)with the
local track, and

/ z,". (4 agree in altitude (if known).

All qualifying tracks are then put through
X position and velocity matching tests. The posi-
tion test for radar systems has traditionally used
' a rectangular p, 8 box.-However,a multisensor
track's error ellipse has a completely different
shape and orientation than a single-sensor

OK f'
NO
track. Thus this usual test is a very poor selec-
tion discriminant. Instead, a Kalman position
test has been developed.
The Kalman position test is illustrated by Fig.
19. By implementingthe standard Kalman filter
formulas and the known error ellipses of the
global and local tracks, the system can use the
local track position to update the global track
position. The new position is then scored by its
o-distance from the two tracks, as shown in the
figure. If the score is less than a preset thresh-
old, a position match is declared.
The velocity match test does not compare
speeds and headings, as in the usual procedure.
Instead, both tracks are predicted ahead T sec-
onds by using their own velocity vectors, and the
Score = 40,2 + 022
resultant positions compared by using the
above position test. This form of the velocity test
(b) is simpler than the usual method, adds no new
parameters, and better accounts for measure-
Fig. 19--Kalman position test. (a) Usualcorrelation.(b)Kal- ment errors.
man correlation. If the position and velocity tests are both
passed, a match is declared. If only the position
different sensors correspond to the same air- test is passed, a potential match is recorded.
craft. The key to the routine (outlined in the Such a match is retested on subsequent scans.
flowchart of Fig. 18)is establishing and main- If the velocity test is later passed, or if the
taining a cross-reference array that translates position test is passed for M scans, a match is
local sensor tracks into global system tracks. declared. The latter type of success acknowl-
Whenever a new local sensor track is encoun- edges that successive position matches verify a
tered, the intersensor processor selects the velocity match, even if fishtailing causes each
proper global track (unless it is the first sighting velocity test to fail.
of the aircraft) with which to assign it. The first If one or more matches are found for a new
step in the process is to apply a coarse screen local track, the best match is selected a s the
that filters out noncandidate global tracks. To global track for it to join. If no matches occur, a
qualify a s the global track to which the local new global track is initiated.
track should be assigned, the global track must Since intersensor matching errors can occur,
(a)not already contain a track component either by failing to find a proper match or by
from the sensor, choosing the wrong one, subsequent scans are
(b) be reasonably close in position to the local checked. Failure to find a proper match is recti-
track, fied by reattempting during each scan tojoin any

?he Lincoln Laboratory Journal, Volume 2. Number 3 (1989)

 Gertz -Multisensor Surveillancefor Improved Aircraft nacking

3%
(1) Claxby

Single
Sensor
Standard
IMean

Bilateration
itandard Deviation

Incremental
Bilateration

' (4) Pease Pottage

-150 -100 -50 0 50 100 150

Nautical Miles

Fig. 20-Test of multilateration in Great Britain data 50 100 200 50 100 200 50 100 200
Range Band (nmi)

single-sensor global track to other global Fig. 2 1-Multisensor measurement of aircraftcross-range

tracks. Choosing the wrong track is fixed by
performing a reasonableness check during each position error; o measures the average noise in
scan on the position differences among the the cross-range position error and hence is a
various components of a global track. If diver- measure of heading consistency.
gence is detected, the divergent component is As expected, single-sensor measurements
stripped from the global track and the matching grew linearly less accurate with range, but
process is repeated for it. multilateration performance remained good for
all ranges. This result confinns the tracking
Real-Data Results improvement of multilateration. Also, as pre-
dicted, incremental bilateration had a smaller
Testing of the multisensor-processing system mean error than standard bilateration. This
to date has consisted of analyzing its perform- result confirms the expectation that the algo-
ance on data collected in Great Britain, where rithm is bias insensitive and is unaffected by
seven sensors provide overlapping coverage of registration errors.
the country's airspace. Figure 20 shows the lo- The turning Kalman filter performance was
cation of the three sensors selected for pro-
cessing, and the trajectories of 20 aircraft -1 1 I I I I
studied. - -
-12
The key test was an evaluation of the accu-
racy of the cross-range measurement obtained
through multilateration. Figure 2 1 presents the
-V,
Q)

-
5
-13

-14
-
-
>\t 3
3
-
-
results a s a firnction of range and of algorithm. .-8 -
5 -15 3f
To have more meaning in this application, the a
mean (p)and standard deviation (0)were rede- -16
,3 - 3
-3 3/C -
fined: -
3f
-17 AL<3& -
-1 8 I I I I I I
71 72 73 74 75 76 77 78
Nautical Miles
0 = P I ( e i , meas - ei, true - (6i-1,meas - ei-1,true 11
i Fig. 22-Performance of turning Kalman filter during air-
Thus p measures the average cross-range craft turn.

The Uncoln Laboratory Jounzal, Volume 2. Number 3 (1 989)

Gertz -Multisensor Surveillancefor Improved Aircra. Tracking
8

also tested with these data. Figure 22 presents improvement in aircraft surveillance and track-
filter outputs from the new turning Kalman ing. Future work should extend the performance
filter, which are f a r superior to the results re- benefits to false alarm rejection, especially for
ported in Fig. 2. The new filter quickly matches primary skin radar systems.
the turn rate of the aircraft at turn onset (note
the curved prediction arrows) and just as References
quickly returns to straight smoothing at turn
end. 1. V.A. Orlando. 'The Mode S Beacon Radar System,"
Lincoln Laboratory Journal 2,345 (1989).
2. J.L. Gertz, "Mode S Surveillance Netting," Project Report
ATC- 120, Lincoln Laboratory (4 Nov. 1983). FAA-RD-
Conclusions DOT-FAA-PM-83-17.
3. A. Gelb. Applied Optimal Estimation (MITPress, Carnbr-
The work to date on multisensor data pro- idge, MA, 1974).
cessing has provided a significant performance

-- JEFFREY L. G E m i s a staff
member in the System De-
sign and Evaluation Group.
He received bachelor's,
master's, and Ph.D. degrees
in electrical engineeringfrom MIT in 1965, 1966,and 1970,
respectively. Jeff came to Lincoln Laboratory from Bell
Telephone Laboratoriesin 1973. His work is now focused on
surveillance and tracking of aircraft for the FAA.

The Lincoln Laboratory Journal,Volume 2. Number 3 (1 9891