The Aerosonde: a Miniature Unmanned Aircraft for = Deploying Lightweight Instruments over Long Range

From: Subject: The Aerosonde: a Miniature Unmanned Aircraft for Deploying Lightweight Instruments over Long Range Date: Sun, 25 Mar 2007 22:58:01 -0700 MIME-Version: 1.0 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Location: http://scholar.google.com/scholar?hl=en&lr=&q=cache:Yh76xDAubhQJ:box.mmm.ucar.edu/uswrp/thorpex/observing/McGeer_Aerosonde.pdf+ X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.3028 =EF=BB=BF The Aerosonde: a Miniature Unmanned Aircraft for = Deploying Lightweight Instruments over Long Range <= /TABLE>

This = is the html=20 version of the file http://box.mmm.ucar.edu/uswrp/thorpex/observing/McGeer_Aeros= onde.pdf.
G o = o g = l e=20 automatically generates html versions of documents as we = crawl the=20 web.

Google is neither affiliated with = the=20 authors of this page nor responsible for its=20 = content.

Page 1

1=20

WIDE-SCALE USE=20 OF LONG-RANGE MINIATURE AEROSONDES

OVER = THE=20 WORLD'S OCEANS

Tad = McGeer=20

The = Insitu Group,=20 Bingen, WA USA

tel 509 = 493=20 8600/fax 509 493 8601

insitu@insitugroup.com<= /DIV>

http://www.insitugroup.com/

Juris = Vagners=20

Department of=20 Aeronautics and Astronautics

University of=20 Washington, Seattle, WA USA

tel 206 = 543=20 7937/fax 206 543 0217

vagners@aa.washington.edu

Abstract

On 20 = August 1998=20 Aerosonde Laima took off from Bell Island, Newfoundland, headed = east=20

through = a stormy=20 night, and landed the next afternoon on South Uist in Scotland's Western =

Isles. = She became=20 the first unmanned aircraft to have crossed the Atlantic, and, at 13 kg = gross=20

weight, = the=20 smallest aircraft by far ever to have made such a flight. Laima = crossed=20 the Atlantic

to = demonstrate in=20 dramatic style new opportunities for economical long-range deployment of =

lightweight=20 instruments. Aerosondes have been designed especially for offshore = weather=20

reconnaissance, in=20 which they have flown hundred of hours of field trials in various venues =

around = the world.=20 Air-traffic and regulatory policy is moving to accommodate these = operations,=20

and = engineering=20 development continues toward fielding of a much more capable, but=20 equally

small, = "second=20 generation". Thus robotic weather reconnaissance, and other applications =

involving=20 lightweight payloads, are poised to become routine within the next few = years.=20

Figure = 1.=20 Aerosonde Trumper leaves her

cartop = launch=20 cradle at Bell Island,

Newfoundland, for=20 the first of four attempts

on the = North=20 Atlantic made by Aerosondes in

August = 1998.=20 Trumper didn't reach Scotland,

but = Laima=20 crossed a few days later, landing

on = South Uist in=20 the Hebrides. She covered

3270 km = in 26 hr=20 45 min, and burned about 4

kg of = 100LL=20 avgas. Altitude was 1680 m

(5500 = ft) most of=20 the way across, but for the

final = 150-odd km=20 Laima flew low over the sea

to = qualify,=20 under UK regulations, as a model

aircraft! She=20 hit every waypoint within a few

minutes of=20 estimate, thanks to a superb wind

forecast by the=20 US National Weather Service

and = careful=20 routing by Cliff Mass and Scarlett

Benda=C3=B1a of the=20 University of Washington. A

map of = Laima's route, overlaid on the satellite

photo = of the=20 rainy enroute weather, can be

viewed = at http://www.insitugroup.com/. Photo by

Keith = Gosse.=20

Page = 2

2 =

Miniature=20 robotic aircraft

Costs = for certain=20 airborne-sensing applications stand on the precipice: they are likely = soon to=20

tumble = thanks to=20 the advent of robotic aircraft with the following key attributes:=20

Aerosonde=20 Operations 1995-1999

complete proposed

Figure = 2. Since=20 its inception the Aerosonde program has emphasised early field trials = with=20

testbed=20 aircraft. Our first trial was flown at the end of 1995, as part of the = Maritime=20

Continent=20 Thunderstorm Experiment off northern Australia (McGeer 1996b), and since = then=20

we = have deployed=20 in diverse jurisdictions and climates in collaboration with the=20 weather

services of=20 Australia, Canada, Taiwan, and the US. Principal results (as discussed = by McGeer=20

et = al.=20 1999) are as follows:

=E2=80=A2 more than 900 flight hours since 1995

=E2=80=A2 deployment with relative ease (especially when we have = traveled with=20 aircraft, ground

station, and=20 launcher boxed up as airline luggage)

=E2=80=A2 7 flights exceeding 24 hr; longest 30.5 hr =

=E2=80=A2 flights in severe tropical thunderstorms (Western = Australia; South=20 China Sea 1998)

=E2=80=A2 flights in midlatitude icing (Vancouver Island, Apr 1998)=20

=E2=80=A2 fully automatic flight from takeoff to landing (Feb 1998)=20

=E2=80=A2 control of multiple aircraft from a single ground station = (Vancouver=20 Island, Mar 97)

=E2=80=A2 inflight transfer between ground stations, and enroute=20 control-by-telephone from

weather=20 forecasting centres (Western Australia; Vancouver Island; North = Carolina;=20

Hawaii)=20

=E2=80=A2 first unmanned Atlantic crossing; 3270 km from Newfoundland = to=20 Scotland on 20-21

August = 1998, in=20 26 hr 45 min using 4 kg fuel

=E2=80=A2 Aerosonde applications well understood and supported by = weather and=20 aviation

agencies in=20 Australia, Canada, UK, Taiwan, and the United = States

Page = 3

3=20

=E2=80=A2 small size, with attendant economies deriving from low = costs of=20 manufacture and use;

easy = shipment;=20 assembly and operation by a single person; flexible basing on ship and=20

shore; = and=20 independence of elaborate facilities;

=E2=80=A2 long range and endurance, sufficient to reach remote areas=20 (particularly over the oceans)

and to = remain=20 there for useful periods of time;

=E2=80=A2 autonomy, with automatic operation from take-off to = landing, and only=20 occasional

messaging while=20 en

route, = such that=20 one

person = could=20 manage

several=20 aircraft

simultaneously;=20

=E2=80=A2 attrition that could be =

high = (perhaps=20 10% of

sorties flown)=20 without

making = safety or=20

economics=20 unacceptable.

Miniaturisation=20 of the necessary

onboard = components,

particularly GPS=20 and satellite-

communications=20 equipment, has

created = opportunity for

development of=20 such aircraft.

Missions that can=20 be flown with

miniature=20 instruments - as listed,

for = example, in=20 Table 1- will be

able to = exploit=20 them.

Our = particular=20 interest has been

in = meteorology,=20 which has a

long-standing=20 need for routine,

inexpensive=20 measurements - of

pressure,=20 temperature, humidity,

wind, = and icing -=20 over oceans

and = remote areas.=20 The

Aerosonde was=20 conceived in 1991 to address this requirement, and The Insitu Group was=20

formed = in 1992 to=20 develop the concept with funding from the US Office of Naval Research = and=20

the = Australian=20 Bureau of Meteorology. In 1995 we were joined by Environmental Systems=20 &

Services, an=20 Australian instrumentation firm, and in 1997 by the University of = Washington.=20 The

weather = services=20 of Canada, Taiwan, and the United States meanwhile added their = sponsorship.=20

Trials = with=20 prototype aircraft began in 1995, and have proceeded through a series of = venues=20 and

technical=20 improvements as noted in Figure 2.

20 = W-hr battery=20 pack (1 hr power)

1.2 kg = avionics=20 set including

two = 68332=20 computers

flight = control=20 sensors

UHF=20 modem

L1, = C/A code=20 GPS

bladder for 5 kg=20 100-octane gasoline

pitot-static=20 pressure ports

flap=20 servo

GPS=20 antenna

oil=20 tank

ignition

engine = and

generator

fixed-pitch=20

propellor AEROSONDE

UHF = dipole=20

antenna=20

in = left=20 tail

graphite/honeycomb shell

graphite=20 sandwich

inner=20 panel

pods = for=20 pressure,

temperature,

humidity=20 sensors

1=20 m

graphite tube=20 tailboom

fibreglass=20 tail

assembly=20 joint

fiberglass=20 sandwich

outer=20 panel

AE RO S O NDE

Figure = 3. An=20 Aerosonde packs the fuel and equipment

necessary for=20 transoceanic flight into a 13 kg package

small = enough to=20 fit on a tabletop.

Page = 4

4 =

The=20 Atlantic crossing

Laima's=20 Atlantic crossing was something of a culmination of the trials = program, as=20 it was the

first = Aerosonde=20 flight to go more than 150 km or so from the launch site. We had=20 previously

excluded flights=20 at longer range for lack of over-the-horizon communications. An=20 Aerosonde-

scale = long-range=20 system will not be useable until new low-earth orbiting networks = (e.g.=20 Iridium,

Orbcomm,=20 and Globalstar) are fully in-service for data transfer, later = this year=20 or in 2000.

These = will be=20 essential for routine service, but we decided on an early demonstration=20 without

over-the-horizon=20 communications in order to alert prospective users, air-traffic = controllers,=20

regulators, and=20 the general public to the level of capability that was quickly coming to = hand.=20

We = chose the=20 North Atlantic, first because the historical significance of Atlantic = crossings=20

would = ensure=20 maximum attention, and second because the distance, and the prevailing = winds,=20

were = just right=20 for the Aerosonde at its current stage of development. The original plan = was to=20

fly = from=20 Newfoundland to Ireland, just as Alcock & Brown had done on the = first=20 nonstop

Atlantic flight=20 in 1919. However the Irish Aviation Authority decided (rather late in = the=20 day)

that in = view of=20 the lack of position reporting enroute it could not authorise our = flight.=20 Perhaps

that = was for the=20 best, since it led to an exhilarating few days in August when we = switched our=20

planning to=20 Scotland and, from a cold start, obtained all of the necessary approvals = from=20 the

authorities in=20 the UK. We are very grateful indeed for the extraordinary assistance and =

enthusiasm=20 offered by the Civil Aviation Authority, the Defence Evaluation and = Research=20

Establishment,=20 and the UK Meteorological Office. (But while their support was=20 truly

Table = 1. Example=20 payloads and applications for miniature aircraft=20

Payload=20

Weight=20

[gm]

Military=20 application

Civil=20 application

Triple = radiosonde=20

instruments

70=20 gm

Ship = -and=20 shore-based

weather=20 reconnaissance

Wide-scale over-ocean= =20

soundings = and = tropical-

cyclone=20 reconnaissance

Nanotesla-level=20 total-field

and = directional=20

magnetometers<= /DIV>

1=20 kg

Magnetic=20 anomaly

detection=20

Geomagnetic=20 survey

Video, in = visible or=20

infrared, with=20 TV or

compressed-data=20 downlink

< 2 = kg

Battlefield=20

reconnaissance;=20 search

and = rescue=20

Fisheries=20 reconnaissance,

search = and=20 rescue; fire

spotting;=20 geodetics; law

enforcement

Synthetic aperture=20 radar

"front = end"

<=20 1kg

Over-the-ridge=20

battlefield=20

reconnaissance=20

Chemical or=20 biological

sensors=20

< 1 = kg

Battlefield=20 toxin

detection=20

Pollution=20 monitoring

Radio = repeater=20

< 1 = kg

Over-the-horizon/over-the-ridge=20 data relay

Radio = detector=20

< 1 = kg

Signals=20 intelligence;

search = and=20 rescue

Search = and=20 rescue

Page = 5

5 =

exceptional,=20 authorities elsewhere - in Canada, Australia, Taiwan, and the US - have=20 likewise

been, = within=20 clear bounds of safety and critical evaluation, very positive about = finding=20 practical

ways to = accommodate Aerosonde operations.)

Our = launch site=20 was Bell Island, near St John's, and we were able to monitor aircraft = through=20

our = line-of-sight=20 data link for about 40 km before losing contact. The subsequent lack of=20

communication=20 made little difference to our aircraft, which operates autonomously = whether in=20

contact = or not,=20 but for the crews on either side of the ocean it made for a long, = anxious=20 wait!

The = tension was=20 all the worse because mechanical reliability was marginal for long = flights. Our=20

emphasis on early=20 field trials had called for some engineering shortcuts, and a number of=20

problems remained=20 to be fixed especially in the powerplant and fuel system. These had cost = us=20

several = aircraft=20 in earlier trials, and we lost three more into the Atlantic. =

At = US$25,000 per=20 aircraft the associated cost was not insignificant, but a gamble is = worthwhile=20

when = early=20 results of novelty and significance are at stake. Of course the high = stakes made=20 for a

tense = game, but=20 Laima, the third Aerosonde to make the attempt, had luck on her = side.=20 (She was

named = for the=20 ancient Latvian deity of good fortune.) She popped up on the Scottish = end right=20

on = time, making=20 contact with the South Uist station at a range of 57 km. An elated trio=20 of

University of=20 Washington engineers brought her in with a gentle plop onto the tall = Hebrides=20

3300=20 km

8200=20 km

ability to=20 reach, loiter, and

communicate=20 wherever observations

are of = greatest=20 value

4=20 km

5.8 = km

greater utility=20 for meteorology and

data=20 relay

54 = kt at sea=20 level

78 = kt at sea=20 level

faster = deployment and less

sensitivity to=20 headwinds

40 = MHz=20 floating point computer

more = autonomy=20 and payload

flexibility;=20 better flight control and

health-monitoring;=20 lower attrition

16 = MHz=20 fixed-point computer

1000=20 flight-hour target

500 = hr target=20 for 1999

$40K expected=20 in 1999

$20K=20 target

<$50/flight=20 hour in wide-scale service

$500/flight=20 hour in small-scale operations

still-air range onboard processing service ceiling maximum

level=20 speed

average lifetime life-cycle cost total flight-hour cost

many=20 applications will depend

upon = economies=20 achieved by

- = lower=20 attrition

- = lower unit=20 cost in production

- = single-person=20 supervision of

multiple=20 aircraft simultaneously

CAPABILITIES=20 OF FIRST- AND SECOND-GENERATION=20 AIRCRAFT

>10 km=20 with supercharging

1.5 = days

5=20 days

sea-level endurance

Figure = 4.=20 Laima, and her stablemates (Figures 2, 3), are relatively = unrefined=20 "first-

generation"=20 Aerosondes developed for early field experience. They will continue to = expand=20

meteorological=20 and other applications over the next couple of years, but upcoming = designs=20

will = offer=20 substantially better reliability, economics, and performance.=20

Page = 6

6 =

grass, = and in due=20 course they joined their British hosts in making full use of Benbecula's = two=20

public = houses.=20 Laima returned to Seattle by air freight, and is now on display = at the=20 Museum of

Flight = on Boeing=20 Field. McGeer & Vagners (1999) give an account of the Atlantic =

demonstration.=20

Toward=20 routine service

The = festivities=20 that followed Laima's flight celebrated not just her own safe = arrival,=20 but

moreover the=20 long-awaited arrival of the Aerosonde concept. Where previously the idea = of=20

long-range=20 operations by miniature robotic aircraft had been a bit of conceptual = esoterica,=20 the

Figure = 5. An=20 important application for Aerosondes is targeted meteorological =

reconnaissance,=20 for example over the northeast Pacific. The NORPEX program in 1998 =

demonstrated=20 that such reconnaissance can substantially improve weather forecasting = (e.g.=20

Szunyogh et=20 al. 1999), and Aerosondes offer about the only fiscally practicable = method=20 in

prospect for=20 routine operations. This would involve cycling several aircraft=20 continuously

though = bases in=20 Hawaii, the Aleutians, and the West Coast (to take best advantage of=20

prevailing=20 winds). First-generation aircraft can begin to nibble at the edges, but=20 second-

generation=20 aircraft will be able to reach, loiter, and communicate wherever = observations=20 are

of = greatest=20 value. For non-meteorological missions, the second generation will = likewise=20

have = improved=20 flexibility for trading range against payload. =

1200 km typ.=20 operating radius

for = 1st=20 generation Aerosonde

3000 km typ=20 operating radius

for = 2nd=20 generation Aerosonde;

>10,000 km=20 missions with

tailwind 0 2000 4000 6000 8000 10000 0 2 4 6

Payload=20 [kg]

Ran g

e = to zero=20 fu

el in s t

ill = a

r=20 [

k m ] 2 nd

gen = Aerosonde

1 st

gen = Aerosonde

Page = 7

7 =

Atlantic crossing=20 made it a well-known fact on the ground. And the set of field trials as = a=20

whole = have pretty=20 well answered the fundamental questions of feasibility. Certainly = aircraft of=20

this = scale are=20 capable of flying long distances and durations - indeed much longer than = has=20 been

achieved by our=20 "first-generation" prototypes (Figures 4, 5). They can be operated =

autonomously.=20 They can fly in, and retrieve data from, severe weather - including = conditions=20

comparable to=20 those in tropical cyclones (Figure 7), which are prospective Aerosonde = targets=20 of

especially great=20 interest to the sponsoring weather services (Holland et al. 1992; = McGeer=20 1993).

Furthermore=20 Aerosondes can be operated safely and accommodated by air-traffic and=20 regulatory

agencies, as we=20 will discuss presently.

With = questions of=20 feasibility substantially addressed, the main question remaining is one=20 of

timing = - or=20 perhaps of funding, which is roughly the same thing. Wide-scale service = within=20 five

years = is a=20 practical possibility. Indeed even the current Aerosonde design, = although=20 developed

as a = testbed=20 rather than a production system, has some capacity for service in the = near term;=20 a

rough annual=20 costs

for = routine=20 Aerosonde reconnaissance

$2,000,000 $4,000,000 $6,000,000 $8,000,000

50,000 100,000 150,000 200,000

annual=20 flight-hours

t o t a

l = annual=20 c

o st [ U S$ ]

100

125

dialy=20 soundings @ 4 flt-hr/sounding

3 = bases/1=20 control station

600 = flt-hr=20 average A/C life

US$30K = lifetime=20 A/C cost

1200 = flt-hr=20 average A/C life

US$20K = lifetime=20 A/C cost

Figure = 6. A=20 reconnaissance program along the lines illustrated in Figure 5 would = cost a few=20

million dollars=20 per year, depending upon the scale of operations and the aircraft = lifetime=20

actually=20 achieved. A few tens of millions of dollars per year would fund a set of = such=20

programs=20 extending cyclone reconnaissance and targeted observations over the = oceans=20

worldwide. The=20 figures here bracket the likely range of costs and lifetimes. To=20 determine

actual = average=20 lifetime - and to optimise it - there is really no alternative to = extensive=20 service

experience.=20

Page = 8

8 =

new = company,=20 Aerosonde Robotic Aircraft (http://www.aerosonde.com/), has been formed in =

Australia to=20 explore its potential. Meanwhile Insitu and the University of Washington = are

moving = on to=20 development of "second-generation" Aerosondes (Figures 4, 5), offering = the=20

reliability,=20 economics, and performance necessary for routine long-range operations = in=20

meteorology and=20 other applications. The full development program through field testing = and=20

setup = for=20 production could be completed in about two years at a cost of about $4M = dollars.=20 A

subsequent annual=20 expenditure of a few tens of millions of dollars per year would be = sufficient=20

to = maintain=20 comprehensive sampling over the world's oceans. This level of = expenditure looks=20

fiscally=20 realistic compared with current weather-balloon spending of about $100M=20 annually.

Mission=20 characteristics in routine service

We = envision that,=20 in routine weather reconnaissance over the oceans, Aerosondes would=20 fly

about=20 10

flight-hours per=20 year worldwide. At a loss rate of, say, one aircraft every ten =

missions, there=20 would be about 100 Aerosondes flying weather reconnaissance at any given =

time. = Likely=20 characteristics of these missions are as follows.=20

horizontal=20 acceleration while flying through a thunderstorm

at = lowlevel;=20 22 m/s commandedTAS

-3

-2

-1

36550 36600 36650 36700 36750 36800

sec = after=20 midnight UTC 24 Jan 98

ho ri zo nta l a ccel era ti o n [g ] east north

10.3 = kg gross=20 weight

20-40 mm/hr

10=20 m/s

sink

25 = m/s=20 (?)

gust on=20 ground

Mr = Freeze=20 at Port Hedland

winds=20 measured by Mr Freeze

-20

-15

-10

-5

35600 35800 36000 36200 36400 36600 36800

sec = after m=20 idn ight UTC 24 Jan 98

nd=20 [

m / s ]

Figure = 7.=20 Hurricane reconnaissance has been a major objective of Aerosonde = development=20

since = early in=20 the program (Holland et al. 1992, McGeer 1993). There was some=20 early

scepticism about=20 the theoretical case for survivability, but actual experience has been=20

reasonably=20 convincing. The plots here illustrate one of several encounters with = severe=20

thunderstorms=20 during 1998 field trials, this one by Aerosonde Mr Freeze. With = the=20 aircraft

orbiting just a=20 few hundred metres above the launch site in Western Australia, we were=20

treated = to a=20 harrowing ten minutes of rapid climbs and descents, phenomenal = overspeeds up to=20

45 m/s, = and=20 dizzying accelerations (indicated here, somewhat qualitatively, by=20 finite-

differences of=20 N/S and E/W groundspeed). The Aerosonde=E2=80=99s = windfinding-by-maneuver

algorithm (McGeer=20 1996a) gets confused in very small-scale gusts, so the reported winds = (o

for S, = x for W) are suspect, but certainly = they were=20 strong. At one point there was a sudden

WHOOSH=20 while our ground-station shed, weighing a tonne or so, jumped = sideways in a=20

microburst.=20 Thankfully we survived the ride, as did Mr Freeze - only to be = struck=20 down in the

end by = water=20 ingestion through the carburettor. An intake shield fixed that problem = for a=20

subsequent storm=20 encounter over the South China Sea.

Page = 9

9=20

=E2=80=A2 supervisory control from a remote station via a low-rate = data=20 link

In = routine=20 oceanic operations, the current line-of-sight link for enroute control = would be=20

replaced by a=20 low-rate satellite link. Data would be routed by telephone from = satellite=20 gateways

to a = central=20 station, which would be responsible for supervisory control of many = aircraft=20

simultaneously.=20 This station would pass position and other data to air-traffic services, = and=20

weather data to=20 users. Data would be in the form of brief messages sent a few times per = hour,=20

2000 4000 6000 8000 10000 12000 86500 87500 88500

seconds after=20 midnight UTC 23 April 1998

icing = sensor=20 frequency [H

100

120

relative=20 hum

idity=20 [

base=20 frq

(bare=20 &

dry)

0.5mm=20 ice

ice = chips=20 lodged

on = edge=20 of

membrane? melting

<base frq due=20 to

water=20 loading

0.2mm=20 rapid

accretion in=20 climb

above = freezing=20 level

base=20 frq

(bare = &=20 dry)

slight = frq drop=20 due

to = mass loading=20 by

a few=20 supercooled

droplets?

comm=20 fade

2000 4000 6000 8000 10000 12000 86500 87500 88500

seconds after=20 midnight UTC 23 April 1998

ing=20

y = [

] 300

600

900

1200 1500 1800

lti

[

]

attempt=20 to

climb = above=20

icing

clim b = abandoned

due to = icing=20 rate

second = attem pt=20 to

clim b = above=20 icing

descent from=20 sounding

2000 4000 6000 8000 10000 12000 86500 87500 88500

seconds after=20 midnight UTC 23 April 1998

icing = sensor=20 frequency [H

-3

-2

-1

sonde = temperature [=C2=B0

]

freezing level=20 @1000 m

2000 4000 6000 8000 10000 12000 86500 87500 88500

seconds after=20 midnight UTC 23 April 1998

ing=20 s

[

]

[

]

overspeed=20 in

turbulence

high = speed for=20

rapid = descent

minimum=20 speed

for = best clim=20 b

Figure 8.=20 Attrition in Aerosonde field trials has been high, but mainly because of = design

weaknesses=20 which will be eliminated in a version for routine service. We believe = that=20 icing

will = then=20 remain as the dominant factor in attrition, as it was (we suspect) for = the two=20 losses to

date = that=20 cannot be attributed to technical faults. Those two might have been = saved had=20 they

been = fitted=20 with icing sensors, as were Aerosondes flown off Vancouver Island in a = 1998=20 trial

for = Environment=20 Canada. Here the sensor - a peizoelectric membrane ringing at its = natural=20

frequency -=20 measured rapid accretion on Aerosonde Fester. (This occurred upon = descent=20 into

cloud, as=20 indicated by the jump to 100% humidity in the upper-left-hand plot.) The =

information was=20 used to alert the ground controller, but aircraft in routine service = would have=20

to = respond on=20 their own with some combination of reversing course, changing altitude, = and=20

then = trying the=20 original course again. Appropriate techniques will be worked out through =

simulation and=20 trials in the second-generation development = program.

Page = 10

10=20

or = otherwise=20 when triggered by events on the aircraft or by the need for a command = from the=20

ground.=20

=E2=80=A2 round-robin = basing

Aerosondes are=20 slow, with a cruise speed of 40 kt or so, and consequently can cover = much more=20

ground = when=20 routed downwind than otherwise. Consider for example reconnaissance in = the NE=20

Pacific, which=20 is of considerable interest to the US National Weather Service. Aircraft = flying

out-and-return=20 from the west coast would usually face strong headwinds on the outbound=20 leg,

and so = be=20 relatively limited in operating radius. A more attractive possibility = would be=20 to

circulate=20 aircraft among bases on the west coast, Hawaii, and the Aleutians = (Figure 5).=20 Typical

winds = are such=20 that the whole NE Pacific would usually be within reach of aircraft = flying=20 this

circuit;=20 out-and-return and reverse-circulation missions would only be used = infrequently.=20

=E2=80=A2 automatically-generated diverse=20 routings

Areas = of most=20 interest to weather forecasters vary from day to day. Contemporary=20 forecast

models = can=20 calculate the sensitivity of a forecast to data from any given area = (e.g.=20 Szunyogh et

al. = 1999), and Aerosonde routes would be varied from day to day to = maximise the=20 value of data

returned. We=20 envision that these routes would be planned automatically (on the = ground),=20 using

optimisation=20 algorithms accounting for forecast sensitivities, aircraft performance, = icing,=20 and

air-traffic=20 constraints.

=E2=80=A2 altitude cycling

To = afford a=20 significant forecast benefit, Aerosondes would have to collect data = through the=20

atmospheric=20 column rather than at a single level. Hence flight profiles would often = involve=20

continuous=20 cycling from minimum altitude (say 200 ft to stay clear of ship masts, = icebergs,=20 etc.)

to = 20,000 ft or=20 so, and in some applications perhaps to more than 40,000 ft. These = cycles would=20

be = constrained=20 by air traffic and icing.

=E2=80=A2 autonomous re-routing (within specified=20 bounds)

Aircraft would=20 have some limited ability autonomously to adjust routes, for example = upon=20

encountering=20 icing or a strong headwind. In the case of icing, the re-routing would = entail=20 some

combination of a=20 backtrack and altitude change before proceeding en route (cf. = Figure 8).=20

Another routing=20 problem would arise in tropical-cyclone reconnaissance. Aircraft=20 flying

cyclone=20 reconnaissance would be programmed to keep station relative to the storm = using=20 their

onboard wind=20 measurements. Stationkeeping might be done in the eye or while orbiting = the=20

storm = at some=20 radius. (McGeer 1996b, Tyrrell et al. 1999)=20

=E2=80=A2 the =E2=80=9CAerosonde = rain=E2=80=9D

We = noted earlier=20 that high attrition would be expected in routine service, that is, most = aircraft=20

would = be lost at=20 sea after a few missions (mainly due to icing, we suppose). One would = have to=20

ensure = that the=20 associated hazard to ships, etc., would be acceptably small. It = turns out=20 (McGeer

1994, = Vagners=20 et al. 1999) that, with a =E2=80=9Crain=E2=80=9D of 1000 = Aerosondes per year, the=20 collision rate

Page = 11

11=20

with = ships in=20 the open sea would be about one every 200 years or so. This seems=20 acceptably

small = relative=20 to other oceangoing hazards.

Regulatory=20 issues

Aerosonde=20 operations, like operations proposed for other new unmanned aircraft, = raise=20 novel

issues = of=20 regulatory policy, and discussions of these issues in various = jurisdictions over=20 the last

decade = have yet=20 to lead to firm standards. The concept of acceptably small risk, in our = view,=20

should = serve as=20 a point of departure.

Familiar=20 manned-aircraft regulation is designed essentially to achieve an = acceptably=20 small risk

to = people and=20 property

=E2=80=A2 onboard the aircraft;

=E2=80=A2 in other aircraft;

=E2=80=A2 on the surface.

The = majority of=20 existing regulation addresses the safety of people onboard, and in = taking care=20 of

them = it takes=20 care of those on the surface as well. But with autonomous aircraft the = onboard=20

problem=20 vanishes, and consequently a lot more diversity is possible than with = manned=20 aircraft.

That = is, with a=20 manned aircraft you have to build to the same standard no matter what is =

underneath you,=20 but among unmanned aircraft, acceptable safety for flights exclusively = over=20

oceans = can be=20 achieved with rather more rickety machines than would be fit to fly over = a city.=20

Hence = the=20 abundance of possibilities which everyone recognises and is struggling = to=20 manage.

Diversity calls=20 into question the conventional approach of mandating specific=20 design

techniques, load=20 cases, redundancy levels, equipment, and so forth. An alternative might = be to=20

specify instead=20 acceptable levels of risk, while being receptive to diverse = solutions for=20 achieving

those = levels.=20 The authorities in Canada

and=20 Australia

have = circulated=20 preliminary ideas

consistent with=20 this concept. Over the next few months, we contemplate developing a = thorough=20

risk = analysis=20 for Aerosondes to serve as a test case and model (Vagners et al. = 1999).=20 This

analysis would=20 aim to define limits on density of aircraft and surface = =E2=80=9Ctargets=E2=80=9D, below which=20

operations could=20 be conducted with acceptable safety.

Elementary=20 midair collision risk and mitigation

We = have=20 mentioned a surface-hazard result from an elementary risk analysis. We = can also=20 add

some = elementary=20 midair collision results (McGeer 1994). It turns out that if Aerosondes = were=20

distributed=20 through some region at the density of radiosonde balloons now flying = over the=20

continental US -=20 and distributed at random, that is, with absolutely no effort at=20 separation - then

from = the point=20 of view of a transiting aircraft the collision probability would be (as = it is=20 now for

radiosondes)=20 about 10

-8

per=20 flight-hour. Meanwhile for transport-category aircraft a=20

Minutes of=20 Non-Piloted Aircraft meeting chaired by Arlo Speer, Chief, Recreational = Aviation=20 & Special Flight

Operations,=20 Transport Canada, Ottawa, 22 May 1998.

Guidance for=20 unmanned aerial vehicle (UAV) operations, design specification, = maintenance and=20 training of

human=20 resources. Civil Aviation Safety Authority draft, September 1998.=20

Page = 12

12=20

catastrophic-failure=20 probability worse than 10 -9

per = flight-hour=20 is generally considered

unsatisfactory,=20 so 10

-8

/flight-hour=20 collision risk is too high - but only by an order of magnitude. =

The = necessary=20 order-of-magnitude improvement - plus several orders of magnitude more - = can=20

readily be=20 achieved by making the Aerosonde distribution non-random.

The = basis for=20 effective separation is precise knowledge of aircraft position, which is =

continuously=20 determined onboard by GPS. At present it is also reported continuously = to the=20

ground = station.=20 Reporting will be reduced to several short messages per hour on = long-range=20

missions with=20 satellite relay, but the ground-station software will run an estimator = during=20 the

intervals=20 between messages. Flexibility of operation will depend upon how = effectively the=20

position=20 information is applied. We foresee application in three levels:=20

=E2=80=A2 Avoidance of busy = airspace

Aerosondes are=20 primarily required over oceanic and remote areas, where low-altitude = traffic is=20

essentially nil,=20 and high-altitude traffic operates on well-defined IFR routes. These = routes=20 can

be = avoided=20 reliably, without undue complication or restriction in flight planning.=20 Operations

near = coastal=20 bases are likely to be more constrained, but, as in our work to date, = bases can=20 be

sited = where=20 activity is light and transit corridors easily arranged.=20

=E2=80=A2 Data communication with=20 ATC

A = real-time data=20 link for position reports and traffic advisories can be established = between=20 the

Aerosonde ground=20 station and the ATC computer network. Aerosondes could then be =

controlled much=20 like manned aircraft, and separation reliably maintained in areas = wherever=20

ATC = tracks the=20 large majority of traffic.

=E2=80=A2 Independent = surveillance

The = definitive=20 technical solution will be active collision avoidance independent of = ground=20

control. A=20 Swedish system has demonstrated mutual avoidance based on local = broadcasting of=20

GPS = position=20 (Nilsson 1992), and similar work is underway in the US (Proctor 1999). = The=20

technical and=20 economic strengths of such an approach make a compelling case for=20 universal

adoption. (The=20 cost might be a few hundred dollars per aircraft, and arguably it would = be=20

economical for=20 government, or industry, to pay for installation on every aircraft = flying rather=20

than = to spend=20 large sums on maintaining TCAS or developing an alternative system.) = Such a=20

system = will not=20 be operational for some years, but once in place it should allow = operations=20

almost = everywhere.

Safety=20 benefits from wide-scale use

In = weighing an=20 acceptable level of risk for Aerosonde operations, one must also = consider the=20

risk = of not=20 doing operations. In particular, we are now suffering a safety = penalty - not=20 least in

aviation, but=20 also in many other fields - through lack of offshore data for weather=20 forecasting.

Thus = if, for=20 example, it was proposed to exclude Aerosondes from some oceanic = airspace due to=20

some = potential=20 for traffic conflict, the risk of conflict would have to be compared = with the=20 risk

associated with