3.1.6.radaraUAV

7/29/2019 3.1.6.radaraUAV

http://slidepdf.com/reader/full/316radarauav 1/26

C.E. Schwartz, T.G. Bryant. J.H. Cosgrove, G.B. Morse, andJ.K. Noonan

A Radar for Unmanned Air Vehicles

Over the years airborne radars have proven theirvalue as wide-area, nearly all-weather

surveillance tools. Typically, airborne radars are large systems mounted in mannedaircraft. Lincoln Laboratory, however, has built a very capable radar system that is

compact and lightweight; the radar has been integrated into an unmanned air vehicle

(UAV). The work is sponso red by th e Army's Harry Diamond Laboratories and the

Defense Advanced Research Projects Agency (DARPA). A significant component of the

radar is a Lincoln Laboratory-designed programmable processor that performs

moving-target detection on board the UAV. The onboard processing permits the use of

a UAV data link that transmits kilobits per second of moving-target reports instead

of tens of megabits per second of raw radar dat a. The system-the airborne por

t ion ofwhich weighs only 110 lb-detects and tracks moving vehicles s uch a s tanks,

trucks, and low-flying helicopters out to a range of 15 km, and classifies them at

shorter ranges.\.

Unmanned airvehicles (UAV) have proven to

be usefu l for observing activity on th e ground

without placing an air crew at risk. Using

television c amera s a bo ar d small UAVs, the

Israelis have successfully penetrated hostile air

defenses and observed ground activity in

enemy territories.

UAVs have several advantages over their

manned counterparts. In addition to the obvious safetybenefit, UAVs are relatively small and

are thus difficult to detect eithervisually orwith

radars. Propeller-driven UAVs are also difficult

to detect with an infrared (IR) sensor because

their engines ru n a t much cooler temperatures

t han j et engines. This combination of fac tors

makes UAVs more survivable in a hostile

environment than manned aircraft. The ve

hicles are also less expensive than their

manned counterparts.

Current UAVs carry optical-sensor pay

loads-such as TV cameras and forwardlooking infrared (FUR) sensors-which are

less susceptible to detection than active devices

such as radars. Under favorable conditions, op

tical sensors can supply high-quality images of

the ground for human interpretation. How

ever, optical sensors suffer from a limited

field of view and f rom severely reduced per-

The Lincoln Laboratory Journal. Volume 3. Number 1 (1990)

formance in adverse weather and battlefield

smoke and dus t conditions.

Radars, on the other hand, can be designed to

rapidly scan large a re as and they are less af-

fected by weather, smoke. and dust. Without

c ross ing interna tiona l bo rder s, long-range

standoff surveillance radars, such as the Jo in t

Surveillance and Target Attack Radar System

[Joint STARS), can provide rapid surveillanceof large areas of foreign terr itory. (Standoff ra

dars are radars whose long range permits

them to remain at a distance from the area

under observation.)

In such militarily strategic regions as Central

Europe and Korea, terrain and foliage masking

can limit the view of standoff radars. Measure

ments taken in Central Europe (Fig. 1) [1) and

masking studies in the Fulda region in West

Germany [2] indicate that a radarwith a depres

sion angle of 11°can detect moving vehicles on

roads radial to the sensor with about twice theprobability ofa radar with a depression angle of

3°. [A depression angle is the angle of the radar 's

line of s ight below a horizontal plane.) Larger

depression angles would r esul t in even better

visibility of the ground.

To avoid enemy air defenses, however.

manned airborne systems must stay a con-

119

7/29/2019 3.1.6.radaraUAV


Schwartz et at. - A Radarfor Unmanned Air Vehicles

100 . . . . - - - - - - - - , - - - ~ - - - _ _ _ _ r - - - _ _ ,for radars aboard manned aircraft to achieve

large depression angles to the area of interest.

Combining the attributes ofamodern radarwith

the UAVs abili ty to penetrate hostile ai r space

creates a valuable complement to Joint STARS.

During a time of hostilities, Joint STARS could

direct a UAV to explore critical areas blocked

from view by terrain or foliage, or it could cue the

UAV to provide a closer look at the activity in a

particular area of interest.

If survivability becomes more ofan issue, the

sUrvivability of UAVs could be increased with a

number of countermeasures. For example ,

cheap radar decoys could be used to t rick the

enemy into firing expensive missiles at th e de

coys, or inexpensive escort UAVs could be used

to attack enemy search radars.In the earlier Netted Radar [3J and Advanced

Airborne Radar [41 programs, Lincoln Labora

tory demonstrated a network that provided

accurate real-time display ofbattlefield activity.

The network consisted of ground and airborne

Moving Target Indicator (MTI) radars with

modern programmable signal processors. The

205 10 15

Depression Angle (deg)

O L . . . . . . J ' - - - ' - - - - ' - - - - - - -L - - - - - - ' - - - - - - - '

o

Fig. 1-Visibilityof moving vehicles on radialroads versusradar depression angle. Rad ia l roads are those roadswhich are radial (±200) to thesensor. The depressionangle

is the angle of the radar's line of sight below a horizontalplane.

siderable distance behind the forward line of

troops. This requirement makes it difficult

80

.£:0 60'Vi

:>

cQ)

u 40 UAV Radarja.. :... ~

20

Fig.2-TheAmberunmannedair vehicle (VA V), whichis manufacturedby Leading Systems,Inc.

120 The Lincoln Laboratory Journal. Volume 3. Number I (1990)

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A Radarfor Unmanned AirVehicles

---

Fig. 3 - U A V radar f lightgeometry. The UA Vradarsystem hasa 360°wide-area surveillancemodein which

the radar antenna sweeps ou t an annular5-km-to-15-km range swath every 18 s. At an operationalaltitudeo f3 km, the swathcorresponds to depression angles of 11°o 37° thus affordingexcellentvisibilityof ground

activity.

airborne radar system, however, was not prac

tical for UAVs: the radar weighed 1200 lb and

was connected via a wideband lO-Mb/s data

link to a 40-ft g round van that contained a

general-purpose minicomputer and a 1400-lb,

3-kW Westinghouse Programmable Signal Pro

cessor. The challenge of the current program

was to perform th e same functions as the earlier

radar within the size, weight, power. and other

constraints imposed by th e UAV platform.

Lincoln Laboratory ha s built a compact MTI

radar system configured as a payload for th e

Amber UAV (Fig. 2). (Amber was sponsored by

th e DefenseAdvancedResearch ProjectsAgency

[DARPA] and is manufactured by Leading Sys

tems. Inc.) Although th e radar is specifically

designed forAmber it could be reconfigured to fit

into other UAVs [5J that have the sufficient

payload capacity.

The UAV radar system can detect and track

moving tactical vehicles s uch a s tanks . t rucks ,and low-flying helicopters out to a range of

15 km. The r adar can also classify targets as

tanks or trucks at shorter ranges when enough

of the vehicle is visible. To reduce the data-link

bandwidth from tens of megabits pe r second to

less than ten kilobits per second, thesystemwas

designed with the signal processing performed

on board by a high-speed programmable proces-

The Lincoln Laboratory Joun1al. Volume 3. Nwnber 1 (1990)

sor that currently operates at 108 million fixed

point opera tions per second (MOPS). Narrow

band data links can be made more robust and

jam-resis tant than wideband data links. Anonboard Inertial Navigation System (INS) accu

rately determines the UAV's position and atti

tude. This information plus th e radar measure

ments on each of th e detected vehicles permit

the est imat ion of the locations of the detected

vehicles.

The UAV radar system has a 360° wide-area

surveillance mode in which th e radar antenna

sweeps out an annula r 5-km-to-15-km range

swath (Fig. 3) every 18 s. When the UAV radar is

flying at an operational altitude of 3 km, the 5

km-to-15-km swath corresponds to depression

angles of 11° to 37°, thus affording excellent

visibility of ground activity. The UAV radar is

designed to penetrate hostile airspace; th e op

erational alti tude of the radar puts it out of the

range of the most common air defense systemssuch as antiaircraft guns and shoulder-fired IR

missiles.

For initial tests and demonstrations, th e UAV

radar is being flown in a captive-flightmode in

which the UAV fuselage i s attached to and

carried by a De Havilland Twin Otter airplane.

All of the components needed to support free

flight are in the UAV fuselage, but theTwin Otter

121

7/29/2019 3.1.6.radaraUAV


Schwartz et at. - ARadarJor Unmanned Air Vehicles

fable 1. UAV Radar Weight and Power BudgetWeight Power

Radar

Transmitter 21 355Receiver and exciter 10 95

Processor 55 400

Antenna system 14 75

Cables and connectors 1Q 25-

110lb 950W

Support Equipment

Inertial Navigation System (including heat sink) 17 50

Data link 12 185

Altimeter and GPS receiver 9 20

Support structure 25 n/a

Cooling fans -.2 lli2

651b 405W

Total 1751b 1355W

contains th e instrumentation and display

needed for system checkout.

An operator in a ground van control s the

radar with commands via th e data uplink.

Moving-target reports are sent down th e data

link in real time for viewing on a display in th e

ground van. For instrumentation reasons, a

wideband data link i s being employed to com

municate between th e manned testbed and th e

ground van.

The radar is being flown and tested in an area

west of Boston, and the moving vehicles on th e

roads are being used as test targets. Preliminary

test results are encouraging: th e locations ofthe

Table 2. UAV Radar Parameters

122

Radar type:

Frequency:

RFbandwidth (instantaneous):

Receiver noise figure:

PRF (variable):

Range resolution:

Linear dynamic range:

AID quantization of I/O video:

Antenna reflector type:

Rotary joints:

Elevation pattern:

Azimuth beamwidth:

Scan speed (variable):

Azimuth sidelobes:

Peakgain:

Polarization:

coherent-pulseDoppler

Ku-band

10 MHz

7dB

3 to 10 kHz

15,30, and 50 m

40 dB minimum

8 bits/channel

parabolic 18 in x 8 in

azimuth and elevation

cosecant squared

3°

OO/s to 48°s

-28 dB (maximum), -35 dB (average)

30dBi

horizontal

TIle Lincoln Laboratory Journal. Volume 3. Number 1 (1990)

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A Radarfor Unmanned AirVehicles

Table 3. Characteristics of the UAV

Radar Basic Waveforms

1. Wide-area surveillancewith moving-target detection

360°scan every 18 s5-to-15-km range swath

50-m-resolutionwaveform

7-kHz maximum pulse repetition frequency

(PRF)

2. Moving-target tracking

30°scan every 5 s

6-km range swath broken into two 3-km intervals

15-m-resolutionwaveform

404-kHz maximum PRF

3. Helicopter tracking

25°scan every 2.5 s

4-km range swath

30-m-resolutionwaveform1O-kHz maximum PRF

large numbers of detected targets correlate well

with known roads.

General Description of theUAV Radar System

Capabilities, Specifications,

and Features

The UAV radar is designed for th e automatic

detection and tracking ofmoving tactical ground

vehicles as well as low-altitude, slow-flying air

craft such as helicopters out to a range of 15 km

(Fig. 3). At that range, the Ku-band frequencies

used have the capability to permit target detec

t ion in more than 90% of all weather conditions

in Central Europe. Via a narrowband data link,

the UAV radar can report moving targets to a

ground van that is up to 30 km away. In an

operational mode, the target data would be

communicated via the UAV's standard data link,which might employ a radio relay to permit very

long-range operation. The UAV platform can

function either in a standoff mode, or in a

penetratormode over hostile airspace; th e oper

ating alti tude of 3 km puts th e UAV out of the

range of inexpensive short-range air defense

systems.

The total payload of th e UAV radar requires


less than 1400 W of prime power and weighs

I75lb (Table 1). The radar system alone weighs

110 lb; the remaining 65 Ib are contributed by

support equipment such as th e INS, data link,

and altimeter. In a free-flight configuration,

the support equipment would normally be

shared with th e flight control system. The an

tenna subsystem, which weighs 14 lb and

requires 75 W, is a mechanically rotating antenna with an I8-in-by-8-in dish, a 3 0 azimuth

beamwidth, and a cosecant-squared-weighted

elevation beam between -1 1 0 and -40 0• Table 2

l is ts the UAV radar parameters.

The UAV ra dar h as t hree bas ic l in ea r FM

pulse-compression (PC) waveforms and oper

ates in a variety of modes to match different

requirements for detection, tracking, and classi

fication (Table 3). For th e primary wide-area

surveillance mode, th e radar performs a 3600

scan and detects moving targets in a 5-km-to

15-km range swath. A 50-m resolution waveform is used and the radar has a high probability

of detecting targets in this mode.

In themOving-target trackingmode, the radar

operates in a track-while-scan mode and a sec

to r scan is chosen to focus on a specific geo

graphic area of int erest. The sec to r scan is

centered on a specified coordinate in th e Univer

sal Transverse Mercator (UTM) coordinate

123

7/29/2019 3.1.6.radaraUAV


Schwartz et aI . - A Radarfo r Unmanned Air Vehicles

RadarTransceiver

Figure 4 shows the Lincoln Laboratory radar

system that was configured to fly in th e Amber

UAV. which is 18 ft long and has a wingspan of

37 ft. The takeoff and landing gear of the Amber

platform is retractable, which allows both an

enlarged nose section for the radar payload and

a radome below th e payload . Working with

Leading Systems, Inc., Lincoln Laboratory suc

cessfully addressed such system issues as

weight distribution and balance, power require

ments, cooling, and the implementation of a

unique radome.

A major challenge was th e construction of a

radome that would satisfy various electrical,

aerodynamical, mechanicaL and fabrication

requirements and constraints . I t was critical

that th e UAV radome no t significantly degrade

the antenna pattern of th e radar. But a conven-

The Amber UA V Platform

Epoxy-Glass Skins

Honeycomb Cores

Epoxy-Glass Spacer

ReceiverExciter

Fig. 5-Cross section of the low-drag, low-distortion VA Vradome.

Fig. 4-Configuration of radarpayload in AmberVAV.

sys tem used by th e U.S. Army. The 50-m separation of indiVidual vehicles.resolution of th e primary wide-area sur- A high pulse-repetition frequency (PRF) was

veillance mode is not sufficient to resolve designed for th e helicopter-detection mode to

closely s pa ce d t arg ets. As a result, a distinguish between helicopters and moVing

higher resolut ion 15-m waveform is utilized targets on the ground. The high PRF is intended

in th e mOVing-target tracking mode for better for detecting indiVidual flashes of th e main rotor

blade. A limited number of measurements on a

Bell Je t Ranger helicopter showed that the main

blade's flash could at t imes be detected. How

ever. more data are requi red to quant ify the

helicopter-blade-detection performance that

th e UAV radar can achieve.

124 The Lincoln Laboratory Journal. Volume 3. Number 1 (J 990)

7/29/2019 3.1.6.radaraUAV


Schwartz et al . - A Radarfor Unmanned AirVehicles

- Without Radome

- With Radome

co -2 0:s-a>

-0

C0>

~ -30

control equipment and the data-link tracker.

The following sections describe the important

components in detail.

Transceiver

The radar transceiver and antenna sub

assembly were built by theAlL division ofEaton

Corp, and integratedwith a traveling-wave-tube

amplifier (lWTA) supplied by the ElectronTech

nology Division of ITT. Lincoln Laboratory con

tributed to the des ign of a power supply and

modulator that insures low spurious output.

This requirement supports target-classification

efforts tha t use low-level spectral signatures

that can occur in the presence of strong clutter.

As mentioned earlier. the transceiver has 50-,

30-, and 15-m-resolution waveforms to support

the different MTI surveillance, tracking. and

classification modes.

Figure 9 is a block diagram that highlights the

major components of the Ku-band transceiver.

Using one of the three pulse-compression (PC)

networks, th e system generates a linear FM

transmit waveform in th e exciter unit. The

waveform is then upconverted to Ku-band.

- Without Radome

-2 0

-30

-1 0 -8 -6 -4 -2 0 2 4 6 8 10

Angle (deg)

Fig. 6-Azimuth pattern of radar with and without radome.

tional spherical radome could no t be used be

cause such a structure would add a substantial

amount of drag to th e UAV platform and would

thus significantly reduce the UAV's endurance,

Lincoln Laboratory designed a low-drag radome

in which the radar antenna pattern is maintained at incident angles that vary from a direc

tion normal to the radome skin to a direction 75°

off the normal. Figure 5 is a cross-sectional

sketch of the des ign , which incorporates a

multilayer sandwich configuration for the struc

ture. Figures 6 and 7 show that the radome

fabricated did not significantly a lt er the bas ic

antenna patterns; note that the-28-dB azimuth

sidelobe levels are unchanged when the radome

is present.

System Components

Figure 8 is an overall block diagram of the

basic components of th e UAV radar system. The

upper half of the figure shows the airborne

components that are housed inside the UAV

fuselage, including the processor, position

location equipment. a nd data link; the lower

half of the figure shows th e ground-based

-5 0 -40 -30 -20 -1 0 0 10 20 30 40 50

Angle (deg)

Fig. 7-Elevationpattern of radar with andwithout radome.

The Uncoln L.aboratoryJoumal. Volume 3, Number 1 (/990) 125

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A Radarfor Unmanned Air Vehicles

passed through the lWfA and duplexer, and

sent out to the antenna. Phase stabili ty is im

portant for any Doppler radar; the single

sideband (SSB) phase noise characteristics of

the UAV radar's L-band stable local oscil

lator (STALO) ar e -65 dBc/Hz at 100 Hz offset,

-97 dBc/Hz at 1 kHz offset, and -130 dBc/Hz

at 30 kHz offset.

The received signal comes into the duplexer/

limiter and is downconverted, attenuated by the

digitally controlled attenuator, and then passed

to the appropriate PC network. Finally, the

coherent detector produces in-phase and quad

rature (I/Q) signals that are sent to the 8 -b it

flash A/D converters.

Figure 9 also shows other features of the

transceiver implementa tion, inc luding th efrequency agility provided by th e STALO, th e

voltage-controlled crystal oscillator for shift-

ing th e mean clu tter frequency, the antenna

control interface, and the signal processor inter

face. To prevent receiver saturation while

maintaining optimum A/D input signal levels,

the transceiver has a digital ly controlled IF

attenuator in addition to th e sensit ivity t ime

control (STC) provided by the duplexer/limit

er . The limiter can provide attenuation from

o to 50 dB.

Inertial Navigation System

The UAV radar system includes a small,

commercial, lightweight INS. The UAV INS

(Fig. 10), which is built by Litton, provides the

platform-location and attitude information re

quired for accurately locating detected targetsin th e UTM coordinate system. The INS is a

strapdown version in that accelerometers and

Platform-

!itadar Signal/Data Location

Transceiver Processor and Data-Link

Subsystems....

Receiver AIDSignal Navigation Inertial

Exciter I - - .. Converters ~ Processing Processor NavigationElements 68030

--System

\0 •.- Altimeter

, ,TWT Radar PostprocessorCommunication~ k - Processor GPS

Amplifier - Controller 68030 ..68030 Receiver

Radar 1ntenna Data Link

~ ~ =round Terminal Van

Display DataData Link

and f . - . Processors ~ f. -I--.Control 68020

Interface Autotracking

Data

o Recording

Link

Fig. 8-Block diagram of VAV radar system. The upperhalf of the figure shows the airborne components that are housedinside the VA V fuselage, while the lower half shows the ground-based control equipment and the data-link tracker.

126 The Lincoln Laboratory Joumal. Volume 3. Number 1 (1990;

7/29/2019 3.1.6.radaraUAV


Schwartz et a1. - AR adarfo r Unmanned A ir Vehicles

Pulse Width EXCITER

EI

Encoder

AntennaPort

r - - - - - - - . Az Drive

EI DriveServo AzControl EncoderModule 1 - + - - -

WG

Ku-BandDownconverter

Limiter

Circulator

DUPLEXER

STC

-

L-Band Freq Agile Sources

RECEIVER

Standby/

On TWTAIPSSync

Mod Gate

+18V

-18V

+8V

VCXO

o±10 kHzOffset

Pretrigger

Quad Video

PC MODULE +28V Power-Supply

Expansion - ModuleNetwork Sync L...- - '

16,6.7,10ps

CompressionNetwork

0.1, 0.2, 0.3ps

External Clock

~ Q ~ ~ ~ ~ T o jreq Sel Signal

Radar-Controller Digital Control A or B Freq Source ProcessorInterface of .......... - . ~ I Interface VCXO

Signal Processor (DCI) Rcvr Gain

PCMode

Az

BPF

COHODCAEI

Mod Gate

N

NF

PC

azimuth

bandpass filter

coherent oscillatordigital control attenuator

elevation

modulation gate

number of frequencies

noise figure

pulse compression

Legend

PO

STALO

STCTWTA

PS

VCXO

WG

X

XTAL

power divider

stable local oscillator

sensitivity time controltraveling-wave-tube amplifier

power supply

voltage-controlled crystal oscillator

waveguide

multiplier

crystal

Fig. 9-Block diagram of VA V radar transceiver.

gyros are effectively strapped to th e UAV's

frame; i.e.. the gyro-stabilized platform has been

replaced by a gyro-stabilized direction-cosine

computer program. Although the flight control

system is separate from the radar payload.

both would share the INS and data link in

an operational configuration.

The INS accuracy is a function of the dy

namics of the flight and th e accur acy of

position updates. Simulation of a typical flight

indicates that the he ad in g er ror will be ap

proximately 0.15°. and the pitch and roll

T he Li ncol n Laboral Ory Journal. Volume 3. Number I (J 990) 127

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A RadarJor Unmanned Air Vehicles

Fig. 1O-Inertial Navigation System (INS): 7.5-in x 6. 7-in x3.2-in Inertial Measurement Unit (IMU) processor (upperleft), 7.5-in x 6.7-in x 3.2-in navigation processor (upperright), and6.3-inx2.5-inx 3. 1-insensor assembly (bottom).The INSweighs 12.Bib and requires 50 Wofpower. Cooling is by conduction and the system mean time betweenfailure is 5200 h.

errors will both be 0.03°. If position updates

ar e available and accurate to about 50 m,

t he n th e corresponding INS position error will

be about 32 m.

Platform-Location Systems

for INS Updating

Because the INS has inherent drift and preci

sion er ror s, it must periodically be position

corrected. We considered three methods for

th e calculation of accurate position estimates

for INS updating. The methods used (1) range

and bearing position estimates. (2) Global

Positioning System (GPSj receiver posit ion

estimates, and (3) multilateration position

estimates.

Range an d bearing position estimates. Most

current systems estimate the location of a UAV

in flight from th e bearing information provided

by the ground-based data-l ink tracking unit

and a range measurement to th e UAV. The

accuracy of the technique, however, suffers

from cross-range error, which increases as th e

range increases.

CPS-receiver position estimates. The best

method of providing position updates to INS is

with an onboard GPS receiver. This approach

provides posit ion information with a spherical

error probability (SEP) of about 15 m, independ

ent of th e distance between th e UAV and the

128

ground-based terminal station. Greater accu

racy could be achieved by placing a second GPS

receiver at th e ground station's known location.

Differential operation could then provide an SEP

of about 5 m by using th e clear acquisition code.

The new generat ion of integrated INS-GPS re

ceiver packages that will soon be commercially

available should make the GPS approach even

more attractive. However, the GPSmethod is not

currently feasible because exist ing GPS satel

lites only cover a certain geographic region for a

few hours every day. About five GPS satellites

were put into orbit from launches preceding

th e 1986 Challenger disaster. After a long

hiatus dUring which no new GPS satellites

were launched. five more were put i nto orb it

in 1989 , and the satelli tes are currently beinglaunched at a rate of one every 80 days.

Nonetheless, 21 GPS s ate llit es a re re

qUired for 24-hour coverage. After enough

GPS satelli tes are available, we will evaluate a

four-channel Motorola Eagle GPS receive r

Fig. 11-Lincoln Laboratory's UAV-radar signal/data processor. The processor is programmable, weighs 55 Ib,

requires 400 W of power, occupies 1.6 ft 3, and currentlyperforms 1DB MOPS.

Th e L in co ln L a b or a tor y Jo u rn a l. Volume 3, Number 1 (1990)

7/29/2019 3.1.6.radaraUAV


Schwartz et a1. - A RadarJo r Unmanned Ai r Vehicles

XBus g g gRadar

Data ControlDual Dual Dual

and..

Element

Processing Processing000 Processing

ControlElement Element Element

~ ~ ~ ~Data-Stream Array Processor (DSAP) Host Bus

~DSAPNME

Interface

~VME Bus

~ ~ ~ ~ ~68030 68030

RAM Interface EthernetProcessor Processor

Fig. 12-Block diagram of the VA V radar programmable signal/data processor.

under typical UAV flight conditions.

Multilateration position estimates. Until more

GPS satellites become available, we are emulat

ing the accuracy of GPS position updates by

using multilateration. With this technique, th e

platform carries a beacon interrogator that ob

tains range information from ground-based

beacons placed at surveyed locations. The

combination of beacon-based posit ion est i

mates with INS filtering simulates the perfor

mance of th e GPS receiver in thata 15-mSEPfor

th e UAV location is provided.

UAV Signal/Data Processor

Amajor component of the radar development

was a state-of-the-art, programmable, compact,

high-speed signal/data processor [61 built by

Lincoln Laboratory. The onboard processor

(Fig. 11) converts tens ofmegabits per second of

raw radar data into less than ten kilobits per

second of moving-target reports. Because the

processor is programmable, it can support a

variety of modes and permits th e easy addition

of new algor ithms. In th e UAV radar system's

The Lincoln LaboralOlY Journal. Volume 3. Number) (1990)

general surveillance mode, th e processor per

forms moving-target detection on a IO-km

swath (250 40-m range cells) at a 6250 PRF

( l ,562,500 samples per second).

The processorweighs 551b, requires 400W of

power, and occupies 1.6 ft3 . The signal process

ing portion of the processor currently performs

108 MOPS. To achieve the small size and low

power consumption. two custom VLSI chips

were used and all of the arbitration logic be

tween the custom chips and dat a memory were

implemented on a custom gate array.

The processor is a complete system in th e

sense that it prOVides not only for signal

processing, but also for th e acquisition of radar

d ata and the generat ion of radar control and

timingsignals. In a single chassis, the processor

incorporates the radar's analog-to-digital (A/D)

converters and interfaces to th e data link, INS.

and altimeter.

Processor Hardware

Figure 12 is a general block diagram that

sh ows ea ch type of board in the processor.

129

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A RadarJor Unmanned Ai r Vehicles

Custom dual-processing-element (PE) boards

that are 12 in x 6 in perform the high-speed

signal processing. If more signal processing

capability is needed, the architecture allows

additional dual-PE boards. The custom control

element (CE) board dis tr ibutes raw radar

samples to the PEs viaa high-speed parallel bus

capable of operating at 10 million 32-bit words

pe r second. The CE board, which uses one ofits

programmable custom chips as the radar con

troller, also serves as th e interface between the

PEs and two commercial VME single-board

computers built by th e Tadpole Co. The Tadpole

boards, which us e Motorola 68030 micropro

cessors, provide general-purpose processing

capability for a number of tasks: postdetection

processing (described in the following section),data communication, navigation functions, and

control of the radar's mode of operation. When

the UAV is in captive-flight operation, a chip

on the Tadpole board supports th e Ethernet

interface to the manned aircraft. The chip also

provides a diagnostic interface that can

be used when the UAV is on the ground and in

free-flight configuration. A commercial 4-MB

RAM board with a bat te ry backup s to res th e

real-time application and diagnostic software

that is needed.

The chassis for the UAV processor (Fig. 11)

contains sixdual-PE boards. Each of the 12 PEs

can be independently programmed to perform

different functions concurrently. In our applica

tion we have chosen to partition the problem so

that each PE is performing th e same moving

target detection or other algorithm on a different

portion of the range swath. Ifa PE fails it can be

turned off and a reduction in the range swath

being processed would result.

It is interesting to note that only about one

third of the processor chassis (and less than

100 W of power) is devoted to the custom programmable signal processing and the remain

ing two-thirds to the VME backplane. It should

also be noted that many of th e VME slots

are not needed. Thus, with no increase in chas

sis size, we could double the signal processing

capability of the processorby doubling the num

be r of PE boards.

Input

AddressGenerator

lAG)

Input Operand Stream

.. '= = ~ = = = : : : . : : = == = = = = = = ~ = ========.==

~ ~ ~ 1 ~ ~ ~ ~ ~· .· ....·

Data

Structures

Data Memory

Arithmetic

Processor lAP)

(x, Y) =F (A, B, C)

Output

Address

Generator

lAG)Output Result Stream

130

Fig. 13-Processing-element (PE) vectorflow of data between the datamemoryand the arithmetic

processor (AP).

The Lincoln Laboratory Journal. Volume 3. Number 1 (J 990)

7/29/2019 3.1.6.radaraUAV


Gate

Arrays

Data

Memory

Address Generators

' - ~ - - - - Y " ' - - - - _ . /PE #1

Schwartz et aI. - A R a d a rfo r Unmanned A ir Vehicles

Arithmetic Processor

• 12-Layer PC Board

• 6 Address Generators

- ASIC (SiliconCompiler)

- 1.5-pm CMOS

• 2 Arithmetic Processors

- ASIC (Silicon

Compiler)

1.25-pm CMOS

• 4 Custom Gate Arrays7000 Gates

- 1.5-J,Jm CMOS

• 2 Data Memories- 64 k x 32 Bits

- 20-MHz Access

Rate (SO-MHz Clock)

' - ~ - - - - y ~ - - - ~ . /PE #2

Fig. 14 -The dual-processing-element (PE) board. The boardoperates at 60 MOPS (80-MHz clockspeed), requires lessthan 10 W, and weighs 24 oz. The speeds andpowerspecifications assume the latest versions of the AG chip (which hasalready been incorporated onto the board) an dAP chip (which has been designed an d is currently being fabricated).

The PE boards contain two types ofprogrammabIe custom VLSI chips: an arithmetic proces

so r (AP) and an address generator (AG). Both

chips have on-chip instruction memory.

Figure 13 depicts th e flow of data between the

PE's 64k x 32-bit data memory and th e AP. The

AP performs algorithmic calculations without

regard for the addresses in th e data memory

of th e inputs or th e outputs. From th e data

memory, an input AG chip fetches th e input

stream of operands for the AP and a separate

output AG chip s tores the output stream of re

sul ts back into th e memory. A third AG chipselects raw radar samples for the appropriate

ranges from the high-speed bus an d places th e

samples in t he memory for subsequent MTI

processing. There are FIFO buffers on th e input

and ou tpu t s ides of the APs so tha t an AG need

only match the average data rate of the AP to

keep the AP running at full efficiency. The sepa

r at ion o f the add re ss and arithmetic calcula-

The Lincoln Laboratory Journal. Volume 3. Number I (1990)

tions into asynchronous processes via theFIFOs greatly facilitates the optimization of

complex signal processing algorithms. This

feature is a significant advance of this s ignal

processor architecture.

Lincoln Laboratory designed the AG and

AP chips and originally implemented them in

3-,um NMOS technology. The chips are currently

being reimplemented in smaller-geometry

CMOS technology. Figure 14 shows the l at es t

version of the PE board.

Processor Software

The signal processor's real-time application

software uses Doppler filtering to detect moving

targets. To remove DC b iases and gain imbal

ances from the I/ Q radar channels, th e signal

processing program generates a table that cali

brates the incoming l/Q 8-bit AID samples in

real-time via a table lookup. In the calibration

131

7/29/2019 3.1.6.radaraUAV


Schwartz et al. - A RadarJor Unmanned Air Vehicles

120.0

2.0.0

Meter/s

-2.0

40.0CDS.r::.

'6>30.0:

(jjQ)

ro20.0j

a:

10.0

noise level in those filters which do not contain

clutter . Detec tions ar e then reported to th e

postprocessor.

Figure 15 is a plot of the fil ters averaged over

range in which the x-axis i s th e 256 Doppler

filters. A target-detection threshold 15 dB above

the average noise level has been plotted. From

this averaged spectrum. those Doppler filters

which contain clutter (in Fig. 15. the filters

represented by the Doppler indexes ranging

from -20.0 to 20.0) are determined. Such filters

ar e ignored in the detection process to insure

that clutter i s not detected.

Figure 16 is a blowup of the bel l-shaped

ground clutter curve of Fig. 15 with the x-axis

converted to m ete rs p er second. The figure

shows the substantial width of clutter velocitythat occurs at broadside. In contrast. Fig. 17

shows the clutter spectrum when the angle be

tween the radar beam and th e UAV velocity

vector is small-a direction in which the spectral

width of clutter is narrowest.

0.0 L....----l._---l._---l._---l._---l._---l._---l.---l

Fig. 16-Blowup of the bell-shapedcluttercurveof Fig. 15.

Note that the x-axis has been converted to meters persecond. The substantial clutterprecludes the detection oftargets whose component of velocity along the radar's lineof sight falls in the clutter spectrum.

50.0

0.0 L.....L.__ ' - -_-- ' - - ' - - - ' -_ ' - - - ' - -_-- ' -_-- - - '......

-120.0

process . which is performed on board th e UAVwhile the vehicle is in flight. 30.000 samples of

radar ground-clutter data are used to determine

the DC biases. gain imbalances. and orthog

onality of th e I/Q radar channels. The calibra

tion process assumes that (1) the amplitudes in

eachchannel should add up to zero if there is no

DC bias. (2) a difference in power indicates a

difference in gain between the channels. and (3)

the I/Q samples should be uncorrelated if th e

channels are orthogonal.

The program then checks for A/D saturation.

performs a Hamming-weighted 64- or 256-pointFIT. estimates th e noise level on a filter-by-filter

basis. and determines the position and width of

clutter in Doppler space. Each Doppler filter is

averaged over range to der ive a noise estimate

for the set ting of detection thresholds and to

determine the location and width ofclutter. The

signal processing program detects targets by

se tt ing thresholds relative to the average

10.0

-40.0 40.0

Doppler Index

Fig. 15-Strength of radar return signal averaged overrange versus Doppler-filter number. A target-detectionthreshold 15dBabove the average noise level anda fourfilter guard band around the clutter have been added.Ignoring the filters within the guard band in the detectionprocess insures that clutter is not detected.

I 1

I 1 Clutter50.0 1

:.--- Guard'11 Gate11

Target- 11

40.0 Detection \ :Threshold I

CDI~ r ~ _ . A - . . . . A . . . - ~ 1 \" "

S I

.r::. I

0>1

c: 30.0 1Q) 1... 1(jj 1

Q) 1

> I

'iiiQj 20.0a:

132 The Lincoln Laboratory Journal. Volume 3. Number 1 (1990)

7/29/2019 3.1.6.radaraUAV


Schwartz et al. - ARadarJor Unmanned Air Vehicles

0.0 L.L..-_...L..-_...L..-_...L..-_...L..-_...L..-_...L..-_..............

INose-Wheel

Blockage

1.2

1.0

1.6

1.4

0.8

0.4

0.6

0.2

100.0 200.0 300.0

Radar Azimuth (deg)

Fig. 18-Clutter-masked ground speeds as a function of

the radarazimuth angle. The noseof theplane isat 00; thus

900

and270 0 represent the radar's lookingat clutterbroadside to the aircraft. In the figure, the curve and the shaded

region under the curve represent ground speeds masked

by clutter. Forexample, if the radarisat anazimuthof 3000,

objects traveling at a ground speed of less than about 1.5m/s will be masked by clutter. The figure was obtainedat aUAV-platform ground speed of about 81 knots.

1.8

0.0 L. . . -_. .L. . . -_. . . .L . . ._. . .J . . . ._-H-_- - ' -_- - - -L_- - - - ' - '

muth of 300°, objec ts t rave ling at a ground

speed ofless than about 1.5m/s will be masked

by clutter. Note that at an azimuth of 200° the

nose wheel of the Twin Otter blocks th e radar

and thus removes th e data used to derive clutter

width .

The velocity extent of clutter at broadside

would have been narrower if the UAV platform

had traveled at a slower speed. The results of

Fig. 18 were obtained with the Twin Otter flying

at ag round speed of 81 kn.

(Note: Wetried

tosimulate free-flight operating conditions as

much as possible. However, because ofphysical

limitations th e Twin Otter could no t be flown as

slowly as the operating speed of the UAV.)

Thus, because the operat ing speed of th e UAV

is 60 to 70 knots (kn), th e performance of th e

UAV radar in free-flight operation will be better

than that indicated by Fig. 18in that th e ampli-

44.00.0 42.0

Meterts

Fig. 17-A veragedclutterstrength versus Dopplervelocity

forthecase inwhich the radarantenna ispointing along theUA Vplatform's velocity vector. Note that the velocity widthof clutterin this figure ismuch narrowerthan that of Fig. 16.

40.0

in~.r::

0>30.0

2!(j jQ)

>.

20.0)a:

10.0

50.0

The spectral width ofclutter results from th e

varying clutter velocities that ar e contained

within the radar' s 3° azimuth beamwidth, Le.,

th e component of th e platform velocity vector at

different angles within the beam. This spectral

width ofclutter varies significantly as th e angle

between the radar beam and th e UAV velocity

vector changes because of th e 360° rotation of

the radar antenna. Adaptive est imation of the

clutter width, as implemented in th e processor,

provides good clutter suppression while allow

ing for low-speed target detection at the more fa

vorab le angles . F igure 18 shows the expected

sinusoidal variation of th e clutter-masked

ground speeds as a function of the radar azi

muth angle. The nose of the p lane is at 0°; thus

90° and 270° represent the radar's looking at

clutter broadside to th e aircraft. In th e figure,

th e curve and th e shaded region below the curve

represent ground speeds that are masked by

clutter and are ignored in th e target-detection

process. For example. if the radar is at an azi-

The Lincoln Laboratory Journal. Volume 3. Number 1 (1990) 133

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A RadarJor UnmannedAir Vehicles

Fig. 19-0e Havilland Twin Otter aircraft carrying VA V fuselage.

tude of th e curve will be reduced by the ratio of

the UAV speed to the Twin Otter speed.

More computationally intensive algorithms

could be implemented to detect low-speed tar

gets that fall within the clutterspectrum. Theal

gorithms would have to separate the slow-mov

ing targets from natural clutter s uch a s wind

blown trees and large man-made structures

such a s buildings and water towers.

During postprocessing. th e mOving-target

detec tions are grouped into centroids based

on their proximity in range. azimuth. and

Doppler velocity. The posit ion of each moving

targetwith respect to the radar is determined by

an average of the ranges and azimuths of th e

detections weighted by th e logarithms of their

amplitudes. The postprocessing program

uses platform heading and position informa

tion from the INS a long with range and azi

muth estimates to determine the UTM posi

tion of each moving-target cen troid. Thi s

general-purpose postprocessing is per

formed on two Motorola-68030-based single

board computers.

In addition to real-time application software.

134

we have developed a significant amount of

support software. including

• a software simulation of th e hardware (the

simulation permitted th e development of

application software and the alteration of

the hardware design prior to the availabil

ity of the hardware) ,

• an assembler to support the development

of application software in a symbolic lan

guage.

• a software and hardware debugger and a

package of hardware diagnostic pro

grams. and

• a real-time UNIX-like operating system to

support multitasking on the multiple

commercial single-board computers.

UAV Captive-Flight Testing

For captive-flight testing of the radar pay

load. an Amber UAV fuselage was attached to a

De Havilland Twin Otter aircraft (Fig. 19). Al

though initial tests have been partially sup

ported by instrumentation aboard the Twin

Otter aircraft. the aircraftwill ultimately provide


7/29/2019 3.1.6.radaraUAV


Schwartz et al. - A RadarJar Unmanned Air Vehicles

Inertial

NavigationSystem (INS)

Fig. 20-Radarmounted in forward payloadbay of VAV fuselage.

only power to th e payload, and physical support

to the UAV fuselage and radar.

All of th e radar system components intended

for UAV free flight-the Ku-band radar t rans

ceiver, th e signal processor, and the var ious

supporting components such as th e UAV INS,

narrowband data link, and antenna sys tem

were assembled in an operational config

uration inside the captive UAV fuselage

(Figs. 20 an d 21). For monitoring and record

ing data from th e radar system, th e manned

aircraft car ries the following instrumentation

and display equipment: a two-way wideband

data link, a DELCO Carousel IV INS, an auto

pilot. and a beacon-interrogation system that

provides the manned aircraft with a complete

and independent system for measuring it s loca

tion and attitude. As mentioned earlier, a

GPS receiver will eventually be used to update

th e UAV INS after a sufficient number of

The Lincoln Labomlory Joumal. Volume 3. Number I (1990)

GPS satellites have been launched.

Radar mOving-target reports as well as th e

Fig. 21-Parabolic dish antennamounted in VAVfuselage.

135

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A Radar ja r Unmanned A ir Vehicles

(a)

(c)

Sun

Development

System

Data Proc.and Tape

Drive

Data

Link

Keyboard

Storage

Closet

Air

Conditioner

Cab

Fig. 22-Experimental groundterminal van forUA Vradar: (a) exterior, (b) interior, and (c) plan view of interior layout.

current radar-mode information and INS data

are sent down the wideband data link to th e

ground terminal van while radar-mode control

commands are sent uplink. Thus an operator in

the van can con trol the radar and view the target

detections on a display in real time. The 10-W

CW FM data link, which is a portable C-band

autotracking unit, was manufactured by

AACOM Co. and adapted for our purposes with

out concern for the link's military specification

requirements or electronic-countermeasure

susceptibility. The data l ink's range and azi

muth coverage are 30 km and 360 0, respectively.

The ground terminal van (Fig. 22) contains adata-link interface, communications equip

ment, a 68020 general-purpose processor, a

digital recording system, a displaymonitor, and

th e operator control terminal. The general func

t ions of the ground terminal equipment ar e to

allow ground control of the radar modes, per

form track-while-scan target tracking, receive/

display target detections and track information,

136

supply UAV data to a general communications

network, and provide playback capability basedon recorded data.

Test Results

We conducted tests to assess the sys tem's

accuracy in determining the posit ions of de

tected moving targets, as well as to assess the

overall system performance. Although th e tests

were done in captive-flight operation with a

DELCO INS and a beacon system for th e deter

mination of platform location and attitude, we

believe the results are indicative of th e UAV

radar's performance in free flight.

Accuracy o j UAV Radar System

For themajor sources oferror associated with

a s ingle detection of a moving target, we evalu

ated a primary error budget. Target-location

errors ar e directly related to th e location error of


7/29/2019 3.1.6.radaraUAV


the UAV itself. an error that is estimated to be

less than 50 m. In addition, when the INS

heading errors (4 mrad), beamsplitting errors

(6 mrad), and other miscellaneous errors were

combined in a root of the sum of the squares

(RSS) manner, th e total angle errorwas found tobe 8.5 mrad. This angle error dominates the

range error and produces a cross-range error

of 85 m at a range of 10 km. Combining th e

50-m and 85-m errors in an RSS manner re

sults in a position error est imate that is less

than 100 m.

Using an electronic mOving-target simulator

(MTS), we tested the system's position accuracy

for mOving-target detections. An MTS is a device

that intercepts a radar signal, amplifies it, and

shifts it s carrier frequency-just as a real mov

ing target would-and then sends the signal

back to the radar. If th e MTS is located at an

accurately surveyed position on the ground, we

can use the measurements to calculate the

Schwartz et al. - A RadarJor Unmanned AirVehicles

range and angle biases in th e UAV radar sys

tem. The measured biases can then be used to

correct measurements of the moving-target

detections.

Figure 23 is a plot of"moving" targets that the

UAV system detected as the radar was flown in

a 10-km circle around a stationary MTS. On a

UTM grid, the figure plots the sys tem' s 354

estimates of the MTS's location. Because the

observations were taken over a full ci rc le, th e

errors in locating th e MTS are indicative of th e

cross-range errors of th e system. The 50-m

resolution waveform was used, and the result

ing one-sigma range error was 16 m. The cross

range errorcorresponded favorably to th e calcu

1ated estimate from th e primary error budget.

However, ou r resultswere obtainedwith an ideal

target; Le., th e MTS had a 30-dB SNR and the

MTS returns were non fluctuating. Further

more, we used a radar scanning r at e of 10°Is .

The use of a faster scanning rate (such as

4,707,675 +• MTS Location

+

+4,707,650

+++

.s 4,707,625

.c~ ++*I -

::::> 4,707,600

+

4,707,575

+4,707,550

280,150 280,175 280,200 280,225 280,250 280,275

UTM East (m)

Fig. 23-Plotof "moving"targets thatthe VAVsystemdetectedas the radar was flown ina 10-

km circle around a stationary moving-target simulator (MTS). The figure contains the VA Vradar system's 354 estimates of the MTS's location. The one-sigma range error is 16 m.

The Lincoln Laboratory Journal. Volume 3. Number 1 (1990) 137

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A RadarJor Unmanned Air Vehicles

(ZOOM INJ (ZOOM OUT) (GRIO ON) (GRIO OFf) [REDRA'tI) [NEW 'tIINDO'tIJ ['olIN Off) [ZO(J.t IN ) [ZO(J.i OUT) (GRIO ON) (GRlD (H ) (REDRA'tI) ~.. ' , . : . . . ~ . _. ' .. - . , .1"

: I) :31:57 ;1t1":[(1 ;f-'f'.I:I hLi irneofj:o("'S I - ItllJlJH I;HIIl '":II/H:UJli (htel.'"'Il" "\ - 111Utl

~ 7 J I 0 0 0 I .. - , - -,

4712011U I

•.....-- .- - ,

.---,;- -,:.:

'--;\'"..' -

4121UUU, ;1":

..

\, ,' . J., , .. j .-=-', ~ 7 1 1 i j U U --~ . .' f:'. ~ ~ ~ . : . - . ~ . /'(

. ":'ff: ~ . . , .,. .-'.-'j ','

411111CllJ _ I ..... /1:. I 4'/I.I1UIHl 1---- - , ·r - -~ ~ ~ . ~ : 1 ·Nr, , .

Leomlnstllr - \.'; I , II ,. .. . . ~ : : " -

\. : .j!. f -.- \ ,

. ' ..

, , ,

\f;.' ' " 1'-"'- ~ 7 U G B U O ,. ::....

\ JiI • ~ f ~ · "- . Y, rtI I - - - -

h ' .......o17HIUUn _.-'r.

Kf,'r,!;

. ,, .. .':.. I (ZU(J.t INJ (ZU(J.t OUT) (GRIO ONI (GRIO Off)IREDRA'tIi em

} . ' • .)00- ,',J f ..'\, .

, ,',' , ' . .. .. ' ~ ~. , .. ,-.

: -4&9 LOOO ____ --'" ~ " ";

' i ~ , • , ~ - - !ir·:rrl SPlIi:TM, (m . l f " . ' ~ .. I = 1 ~ l l j ~ t

.IT

. , .r · .I . - ; "J .

- - ' ..

. '\ i . . --i -, ,..

4GO 1000l

I 1"' ~ \6Q51UU ..:JIo ."

40941UU

4G71 l1 lH l_ - _ .-257000 267000 277000 207ROO 297000 307000 317000 ',' ~

~ f i 9 : ' l t r l n Ie'

{ I ,;::- - ~ ......, ' , ~

:;,..

\:-; -. :-: 4 G ~ l Z 1 0 0 -

" \ih

.... ~ . ~ ~;

:<

Fig, 24-Van operator's display showing3.5min ofmoving-targetreports during a testin which the UAVradarsystem

was flown west of Bostonabove the town of Leominsterand Interstate 495, amajorhighway with three lanes ineachdirection. The operatorhas selectedthree map windows, In the left window the operator is viewing the overallarea.

Some of the major roads in the area have been drawn as line segments, and a UTMgrid is used for reference. Thesmallblacksquares indicate the locations ofmoving-target reports from theUA Vradar. On the rightpartof the display,theoperator haschosen a higher resolution tomonitor two road intersections. Note thatthe moving-targetdetectionscorrelate well with the road locations.

20° Is) or measurements on smaller vehicles

(such as automobiles) would result in larger

location errors. Also, large tactical targets such

as tanks and trucks can cause th e radar returns

to fluctuate.

Nonetheless, further captive-flight tests on

military vehicles indicate that the cross-range

error should be within the estimated error from

th e primary error budget. (The tests ar e de

scribed in greater detail in the following sec

tion.) We estimate that the error will be com

parable after the GPS receiver is incorporated

into the system. As stated earlier. we are

cur rent ly emula ting the posit ion-upda te ac

curacy of a GPS-based system by using

138

multilateration with ground beacon ranges to

determine the platform location in real time.

Further improvement can be obtained by the

correlation of target measurements from scan

to scan.

Test Results on Moving Targets

In captive-flight tests. the UAV radar system

has detected vehicles traveling on the highways

west of Boston in the vicinity of Interstate 495.

Figure 24 is a picture of the ground van's real

time operator display from 3.5 min of moving

target reports. The radar is in the wide-area

surveillance mode and the operator has selected


7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A RadarJo r Unmanned Air Vehicles

t hree map windows on th e display, which is

manufactured by Sun Microsystems. In the left

window th e operator is viewing the overall area.

Some of the major roads in th e area have been

drawn as line segments and a UTM grid is used

for reference. The sma ll black squares indi

cate the locations of moving-target reports from

the UAV radar . On the right part of the display.

the operator has chosen a higher resolution

to monitor two r oad intersec tions. Dur ing

the tests. a large number of moving tar

gets were detected over an area ofapproximately

900 km 2• and the detections correlated well

with the roads, as shown on the display. Note

that an operator can easily discern the varying

densi ty of the traffic. The main road in the cen

te r of the map is I-495-a large highway with

three lanes in each direction. Other smaller tree

lined roads nearby are also clearly visible.

During th e tests, the plane flew at an altitude

of 4500 ft and the radar antenna scanned a 360 0

azimuth sector at a rate of20°Is. The radar was

transmitting a 50-m-resolution waveform. and

data from a range swath between 5 km and 15

km were processed. The PRF used corresponds

to an unambiguous velocity interval of about

±29 mls (±64 mph).

The operator in th e van can also choose to

display mOving-target tracks that ar e generated

by a tracker (Fig. 25). The tracker correlates

GRID SPACING (melers) = seo

47002110/

/I

18:39:2&

II

sync

up on up of f cent . ce.' -deb 4 do . 8 cent cl r sya box

_it on _i t of f trk on Irk o ff

_otI on ao d o ff t • • tn'1t tr k cl r .

do cent don·' cent on hts't o ff . ns t

1.p • 1. , 1 bn • b n 4

1. p 2 '.p 3 bn • hn 1 .

iap 4 1. p 5 on ta g of f la g

vector spat

win"" speed

start step

cip DR <1, of f

4707280

••: . X\1 ••

2007080020079700

4706'700 L ---J

279200

Fig. 25-Real-time display of four truck-size vehicles in track. The large dots represent the most recentlocations ofeach vehicleand the smallerdiamonds depictthe historyofprevious locationestimates. The crosssymbol indicatesthe centerof the currentsector scan. The target numbers are used for referenceby the operator to fetch more dataon a particular track or to select target information that needs to be transmitted to another user. The processingalgorithm automatically counts the number of vehicles in the operator-drawn box and displays the count (fourvehicles) and their average speed (24 mph) in the lower right corner of the screen,

The Uncoln Laboratory JoumaL. Volume 3, Number 1 (1990) 139

7/29/2019 3.1.6.radaraUAV


Schwartz et aI. - A Radarfor Unmanned Air Vehicles

sync ca l AD cent on cent of fu

up of f eent • cent -p on

deb" deb 8 cent cl r By. box

lCJIJ1 t on .1 t of f trk on tr k o ff

.od on .. od of f tr k ln1 t tr k cl r

do cun1 don1 cent on hi &1 of f hut

l. p I l .p 1 On • On 4

IMp 2 iMp 3 On & hn 18

tap 4 iMp 5 on tllg of f tag

vectar spot

..1ndow spend

start stop

ti P on c1 p of f

11 :24:45

. . .

GRlO SPACING (melers) = 500

- \

\\

47873041-- - - - - I - - - - : :=="""" -= : : : := - - - -+- - - - -+ \ - - - - - \ - - - - - -1

\

47078041- -1---.., - /- -++ 1

4-/068041--+----I------+-= ......:::.:::..---I+-_J.-j

4 7 ~ 6 3 U 4 l - ---l -- l. . - - ' - - - - : - _ - - - - '

219350 219050 200350 200050

speedtrks 1 .. v... b p e ~ d £,5 mph

Fig. 26-Real-timedisplayof a low-flying helicopterin track. The large dotrepresents themost recent estimate of the

vehicle's location and the smaller diamonds depict the history of previous location estimates. The cross symbolindicates the centerof the current sector scan. The operator has chosen to display nine previous estimates and thelower right corner of the screen indicates that one target travelingat 65mph has been detected within the operatordrawn box.

moving-target reports from scan to scan to

generate estimates of mOving-target positions

and velocities. For effective track association.

vehicle measurements are needed every 5 s.

The t racker uses the measured Doppler-based

velocities and positions to initiate tracks and

to associate measurements from one scan to

another.

In Fig. 25. the operator has drawn a boxaround a truck convoy on a road westofBoston.

The processing algorithm automatically counts

the number of vehicles in the box and displays

the count (fourvehicles) and their average speed

(24 mph) in th e lower right comer ofthe screen.

The large dots represent the most recent esti

mated locations of each vehicle. The smaller

diamonds dep ic t the history of previous esti-

mates of th e targets' locations. The target

numbers can be used by the operator to fetch

more data on a particular track orto select target

information that needs to be transmitted to

another user.

Figure 26 shows part of the track of a low

flying helicopter that flew into the test area. As

in Fig. 25. th e large dot represents the most

recent estimate of the vehicle's location. Theoperator has chosen to display nine previous

estimates to obtain the flight path of the helicop

ter, and the screen indicates that there is one

target traveling at 65 mph within the operator

drawn box.

Currently, the tracker is configured to main

tain files on 100 tracks. Amuch larger number

of t arge ts cou ld be tracked if the process-

140 The Lincoln LaboraLOry Journal. Volume 3. Number J (J 990)

7/29/2019 3.1.6.radaraUAV


i ng power in t he g round van is substantially

increased.

The UAV radar system's 15-m waveform

provides good resolution in range, but the 2.3 0

two-way azimuth beamwidth results in a 400-m

cross-range resolution at a 10-km range. The

large cross-range resolution limits the system's

ability to separate individual vehicles in the

azimuth direction.

Future Improvements and

Applications

Although system weight was a major concern

in designing th e prototype UAV radar, our pri

mary emphasis was on proving th e feasibility of

the system. A second prototype design could

easily reduce the radar weight by 15 to 20 lb,

which would result in a maximum weight of 90

to 95 lb for th e radar system alone. In addition,

elements of th e support equipment such as the

frame could be lightened.

Enhancements could also be made to in

c rease the sensitivity of the r adar and there

by increase its detection range. Two modifica

tions that would contribute to such an increase

ar e (1) a new low-noise amplifier in th e front

end that would decrease the system noise, and

(2)a new antenna to provide more gain as wellas elevation information for estimating

target-aircraft alt itude. The above hardware

modifications coupled with additional signal

processing could i nc rease the sys tem' s range

to 25 km.

The speed of the signal/data processor is

currently l imited by the speed of the custom

VLSI NMOS AG and AP chips. To replace th e

chips, application-specific integrated circuits

(ASIC) were designed with Seattle Silicon's

ASIC compiler. The new AG chips. which have

been fabricated in 1.5-,um CMOS, will supportan increase in processor speed from 108 to

360 MOPS. The new AP chip is being fabricated

in I .25-,umCMOS and is expected to support th e

same increase in speed. Because the new chips

ar e plug compatible with th e existing ones, the

current processors can be upgraded with simple

one-for-one chip replacements.

The Lincoln Laboratory Journal. Volume 3. Number J (J 990)

Schwartz et a1. - A Radarfo r Unmanned Ai r Vehicles

Use of th e ASIC compiler will ease future

upgrades because once a chip is designed with

th e compiler th e chip can be fabr icated in a

newer, higher-speed technology by a relatively

simple recompilation in th e new rule set . That

is, th e chip would not have to be redesigned.

Seattle Silicon guarantees th e delivery of

packaged functioning chips that will correctly

execute the simulation test vectors used to vali

dat e th e chip design.

Increased processing capability is important

to support new radar applications such as syn

thetic-aperture radar for stationary-target de

tection, adaptive jammer nulling, and air de

fense. UAV r adar s a re not limited to ground

surveillance; in fact, theycould provide accurate

surveillance oflow-flying aircraft that cannot be

detected by ground radars because of foliage or

terrain masking. Or UAV radars could be used

for th e detection and tracking of ships at sea.

Although the prototype system was specifically

designed to detectmoving targets on the ground,

the system is programmable and c an hence be

reprogrammed to handle other types of applica

tions. For example, to detect and track low

altitude, fixed-wing aircraft, th e UAV radar

could be reprogrammed to scan a limited azi

muth sector more rapidly.

Summary

Current technology supports a UAV radar

payload that detects moving ground vehicles

and helicopters out to a range of 15 km. The

payload can supply the battlefield commander

with a real-time display of battlefield activity in

all weather conditions. dayor night. Information

provided would include accurate vehicle loca

tion and velocity data, and th e payload has

limited target-classification capability.

Acknowledgments

The success of the UAV-radar program at

Lincoln Laboratory is th e result ofthe combined

efforts and con tr ibut ions of many ded icated

people. The autho rs thank their group col

leagues Ben Bader, Paula Caban, Cary Conrad,

141

7/29/2019 3.1.6.radaraUAV


Schwartz et a I. - A Radarfo r Unmanned Air Vehicles

Bob Catalan, Elsa Chen, Dave Craska, John

Duncan, Bob Giovannucci. Ed Hall, Don Mal

pass, Gary Provencher, Greg Rocco, Tom Sefra

nek, Keith Sisterson, Mike Spitalere. Sue Yao,

and Alan Yasutovich. The valuable support of

Warren Bebeau, Robin Fedorchuk, Michelle

Hinkley, and Michele Kalenoski in th e Labora

tory's Battlefield Surveillance Group office is

also appreciated. In addition, th e authors ac

knowledge the significant contributions ofMike

Judd and Al Benoit in th e Laboratory's Engi

neering Division and George Knittel for th e

radome design.

The work descr ibed in this art ic le was spon

sored by th e Army's LABCOM. Harry Diamond

Laboratories, and th e Defense Advanced Re

search Projects Agency (DARPA). The authors

thank Jeff Sichina. John David, and Richard

Slife of th eArmy's Harry Diamond Laboratories;

and John Entzminger and Dominick Giglio of

CARROLL EDWARDSCHWARTZ received a B.S.degree in physics from Columbia University. Since

1965 he has worked on thedesign o f rea l- time radar

systems. ranging from large Ballistic Missile Defense measurement radars to smaller tactical airborne and groundsurveillance radars. His field ofspecialty includes real-timesoftware. automatic target detection an d tracking. and

signal processing. Afterjoining the technical staffof LincolnLaboratory in 1972. Ed spent five years working on algorithms for the automatic real-time detection. tracking. anddiscrimination of ballistic targets . For three ofthoseyears.he worked on Kwajalein. Marshall Islands. where he led a

group that tested the algorithms against missiles launchedfrom VandenburgAir Force Base. Ed is currently the Assistant Group Leader of the Laboratory's Battlefield SurveillanceGroup. which is responsible for building the UAV radarand its high-speed processor.

142

DARPA/TTO for their valuable advice and con

tinuing support.

References

I. V. L. Lynn. '"Terrainand FoliageMasking for Long-RangeSurveillance: A Sample o f Measurements in Central

Germany." Project Report TST-35. Lincoln Laboratory (15June 1979), DOC AD-B040205-L.

2. W. L. Post."LineofSightConsiderations in the FuldaGapArea of the Federal Repub li c o f Germany. " TechnicalReport DELCS-TR-82-2. U.S. Army ERADCOM Combat

Surveillance & Target Acquisition Laboratory. Fort

Monmouth, NJ (Sept. 1982).3 . M.l. Mirkin. C.E. Schwartz. and S. Spoerri. "Automated

Tracking with Netted Ground Surveillance Radars."

Record o j IEEE 1980 [nil. Radar Con].. Arlington, VA,28-30 Apr. 1980. p. 371.

4. C.E. Schwartz. "Netted Radar Program Final Report.Volume Ill: Airborne Ground Surveillance Radar," Lincoln Laboratory (30 Sept. 1981). OTIC #AD-B061852-L.

5. E.R Hooten and K. Munson, Jane's Battlefield Surveil-

lanceSystems1989-90

(Jane's Information Group.Alexandria. VA. 1989).6. F.E. Hall and A.G. Rocco. Jr .. "A Compact Program

mable Array Processor." Lincoln Laboratory J. 2, 41(1989).

THOMAS G. BRYANT is astaff member in the Battle

field Surveillance Group atLincoln Laboratory. Duringhis 21 years at the laboratory. which included three

years on Kwajalein in the Marshall Islands. he ha s helped todevelop val;OllS radar systems. includinga millimeter-waveradar for a missile-seeker application. He is currently theproject engineer for the unmanned ai r vehicle (UAV) radarsystem development program. His research interests include radar systems analysis. radar systems for airbornemoving- and fixed-target detections, and associated signalprocessing.Tom received a B.S.E.E. degree and anM.S.E.E.degree from th e University of Maine. He is a member of the

IEEE. Eta Kappa Nll. Tau Beta Pi. PhiKappa Phi, an d SigmaXi.

TIle Lincoln Laboratory Joumal. Volume 3, Number 1 (1990)

7/29/2019 3.1.6.radaraUAV


JAMES H. COSGROVE is an

associatestaffmember in the

Battlefield SurveillanceGroup at Lincoln Laboratory.

Beforejoining th e Laboratory 15 years ago. Ji m worked forth e U.S. Army. He received a B.S. degree in physics from the

University of Massachusetts.

JAMES K. NOONAN re

ceived a B.A. degree in

mathematics from Merri

mack College and an M.S.degree in mathematics from Northeastern University. Heworked for the National Security Agency before joining

Lincoln Laboratory 10years ago. Currently a staff member

of the Laboratory's Battle Surveillance Group. he conducts

research in radar signal processingalgorithmsand real-time

systems. He is a member of Phi Kappa Phi.

The Lincoln Laboratory Joumal. Volume 3. Number 1 (1990)

Schwartz et al. - A RadarJar Unmanned Ai r Vehicles

GERALD B. MORSE is

Group Leader of Lincoln

Laboratory's BattlefieldSur

veillance Group. Hi s re

search interests include radars and signal processing. Before joining th e Laboratory 25 years ago. Jerry worked forVarian Associates. Hereceived a B.S. degree in physics fromNortheastern University and an M.S. degree in physics fromthe Rensselaer Polytechnic Institute.

143

7/29/2019 3.1.6.radaraUAV


Documents

3.1.6.radaraUAV