See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225247482

Aerial Robotics

Chapter · January 2008

DOI: 10.1007/978-3-540-30301-5_45

CITATIONS READS

53 10,820

2 authors, including:

Eric N. Johnson
Pennsylvania State University
327 PUBLICATIONS 7,880 CITATIONS

SEE PROFILE

All content following this page was uploaded by Eric N. Johnson on 12 June 2014.

The user has requested enhancement of the downloaded file.

 1009

Aerial Robotic
44. Aerial Robotics

Eric Feron, Eric N. Johnson

A wide array of potential applications exist for or development in this field. However, many
robots that have the level of mobility offered universities, research centers, and industries
by flight. The military applications of aerial have now met this requirement and are ac-
robotics have been recognized ever since the tively working on the challenges presented

Part F 44
beginnings of powered flight, and they have above. The largest obstacle to the commer-
already been realized to sometimes spectacu- cial development of aerial robots is, however,
lar effect in surveillance, targeting, and even the necessity to comply with and support
strike missions. The range of civilian appli- a regulatory environment which is only be-
cations is even greater and includes remote ginning to address these rapidly developing
sensing, disaster response, image acquisition, systems.
surveillance, transportation, and delivery of
goods.
This chapter first presents a brief history of 44.1 Background ......................................... 1010
aerial robotics. It then continues by describing
the range of possible and actual applications 44.2 History of Aerial Robotics ...................... 1010
of aerial robotics. The list of current chal- 44.3 Applications of Aerial Robotics ............... 1012
lenges to aerial robotics is then described. 44.3.1 Possible Applications
Building from basic notions of flight, propul- of Aerial Robots ......................... 1012
sion, and available sensor technology, the 44.3.2 Current Applications ................... 1013
chapter then moves on to describe some of
44.4 Current Challenges................................ 1014
the current research efforts aimed at address-
44.4.1 Regulations and Certification ....... 1014
ing the various challenges faced by aerial
44.4.2 Human–Machine Interfaces......... 1014
robots. 44.4.3 Navigation ................................ 1015
The challenges faced by aerial robots span 44.4.4 Agile Flight and Fault Tolerance ... 1015
several and distinct fields, including state reg- 44.4.5 Obstacle Avoidance .................... 1015
ulations, man–machine interface design issues, 44.4.6 Aerial Robot Landing
navigation, safety/reliability, collision preven- and Interaction
tion, and take-off/landing techniques. The size with Other Vehicles .................... 1015
of aerial robots can considerably influence their 44.4.7 Multivehicle Coordination ........... 1015
flight dynamics, and small aerial robots can end
44.5 Basic Aerial Robot Flight Concepts .......... 1015
up looking considerably different from their
44.5.1 Aerial Robot Flight
larger counterparts. Similar to their manned and the Importance of Scales ...... 1015
counterparts, aerial robots may enjoy diverse 44.5.2 Propulsion Systems..................... 1017
propulsion systems and operate over large speed 44.5.3 Flight Vehicle Types
ranges. and Flight Regimes..................... 1018
Aerial robots must be equipped with reliable 44.5.4 Lighter-Than-Air Systems............ 1019
position and actuation equipment so as to be
capable of controlled flight, and this constitutes 44.6 The Entry Level for Aerial Robotics:
Inner-Loop Control ............................... 1020
a nontrivial requirement prior to doing research
44.6.1 Sensing and Estimation .............. 1020

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1010 Part F Field and Service Robotics

44.6.2 Estimator Design ........................ 1021 44.7.4 Reactive Flight

44.6.3 Inner-Loop Control..................... 1021 in Cluttered Environments
and Obstacle Avoidance .............. 1025
44.7 Active Research Areas............................ 1022 44.7.5 Path Planning and Higher-Level
44.7.1 Interfacing Planning Capabilities .................. 1025
with the Human Infrastructure: 44.7.6 Integrated Aerial
Meeting the Regulations ............. 1022 Robotic Operations:
44.7.2 High-Agility Flight ...................... 1023 Aerial Robotics Contests .............. 1025
44.7.3 Take-Off, Landing,
and Interaction 44.8 Conclusions and Further Reading ........... 1026
with Other Vehicles .................... 1023 References .................................................. 1027
Part F 44.2

44.1 Background
The term aerial robotics is often attributed to Robert cept. Finally, it could mean a combination of the above,
Michelson [44.1], as a way to capture a new class of that is a description of the robotic platform, together with
highly intelligent, small flying machines. However, it is its robotic mission. In aerospace jargon, robotic flying
clear that the range of systems and activities covered un- machines are commonly referred to as unmanned aerial
der the label aerial robotics could extend much further, vehicles (UAVs), while the entire infrastructures, sys-
and that its roots can be found far back in the beginning tems and, human components required to operate such
of the 20th century, together with the birth of aviation. machines for a given operational goal are often called
Behind the word aerial robotics we can find several unmanned aerials systems (UASs). Finally, it is worth
meanings: it could mean robotic flying machines, that is, noting that many current manned aerial systems defi-
a mission-independent, platform-oriented concept; how- nitely carry relevant features of some robotic systems,
ever, it could also mean robotics that use flying machines, and so much of this discussion is relevant to manned
that is, a platform-independent, mission-oriented con- aircraft.

44.2 History of Aerial Robotics

The history of aerial robotics is very closely tied to by electromagnetic waves, and with enough onboard
the history of flight itself. Indeed, the rate of fatali- logic to recognize and execute remotely transmitted or-
ties associated with early manned flight tests probably ders. However, the concept imagined and engineered by
convinced engineers that there was a need to operate fly- Tesla did not apply specifically to airborne vehicles.
ing machines without the presence of humans on board Using the definition
even before potential applications of unmanned aircraft
An aerial robot is a system capable of sustained flight
surfaced. In 1903, heavier-than-air flight was unambigu-
with no direct human control and able to perform
ously shown to be feasible, following the achievements
a specific task,
of the Wright brothers. The first successful powered
flight was unmanned, presumably to reduce the risk to leads us almost immediately to the Hewitt–Sperry au-
the pilot and to allow a smaller and less expensive vehi- tomatic airplane, developed before and during World
cle (reasoning that is still put forth today) by Samuel P. War I [44.3]. The airplane’s purpose was to act as a fly-
Langley’s Number 5 in 1896 [44.2]. ing torpedo, carrying onboard intelligence to sustain
Finding a truly defining moment for aerial robotics is flight over long periods of time without human inter-
a challenge, with encyclopedias dating the concept back vention. Such intelligence was provided by a complex
to Leonardo da Vinci, while Newcome [44.3], in his his- system involving Sperry’s own gyroscopes, mechani-
tory of unmanned aviation, gives early credit to Nikola cally coupled to the airplane’s control surfaces so as to
Tesla for devising a robotic vehicle remotely controlled stabilize the vehicle. This made the airplane suitable for
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.2 History of Aerial Robotics 1011

Maximum altitude (×1000 ft)

0 10 20 30 40 50 60 70 80 90
0.1
Black Widow
Microstar High alt/UCAV
Medium alt
Tractical
Wasp Mini
Micro
Micro
1

Fig. 44.1 V-1 German cruise missile (1940s) Dragon Eye

Aladin
10 Sender

Part F 44.2
Pointer Mini
Azimut
Extender
Seascan
Luna Sheddon
Shed Mk3
100 Mini-V
Tractical High alt/UCAV
Fox
Pioneer
Phonix
Shadow 200
Solar Bird Pathfinder +
Frowler II
1000 Searcher Gnat 2 Centurion Raptor
Hunter Perseus
Predator Altus
Eagle 1
Hermes 1500
Eagle 2
10 000 Medium alt
Predator B

Global Hawk

Fig. 44.2 QH-50 DASH unmanned helicopter on final ap-

proach (US Navy)
Max take-off weight

human remote control and, eventually, prosecution of Fig. 44.3 Taxonomy of unmanned aerial vehicles (after
distant targets. As discussed later, one of the key char- R. Weibel [44.4, 5])
acteristics of aerial robotics is this particular necessity
for the robot to sustain itself in the air with no hu- devices. Many of the ensuing aerial robotics develop-
man intervention, which requires the early adoption and ments followed the initial idea of defense applications
understanding of the critical role played by onboard in- for unmanned systems, that is, ever more accurate fly-
telligence, much more so than other robotic applications. ing machines for the purpose of either reconnaissance
While a string of inventors in many countries came to or weapon delivery. One notable machine was the US
develop ever more sophisticated machines, credit goes Navy’s Gyrodyne QH-50 DASH, an unmanned heli-
to the German V-1 cruise missile for making a lasting, copter developed in the 1950s and operated from US
and unfortunately deadly, impact on large segments of destroyers, which was able to perform reconnaissance
the population in England. This form of robotic aircraft missions and deliver torpedoes (Fig. 44.2). However,
owed its relative inefficiency (three out of four vehicles these machines remained relatively unintelligent, and
reportedly missed their target – predominantly London) their level of autonomy remained limited to the abil-
to mechanical failures and lack of good navigation capa- ity to sustain flight using complex inertial and other
bilities beyond dead-reckoning assisted by gyroscopic measurement systems.

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1012 Part F Field and Service Robotics

a) b) ability to fly reliably and take off and land from lim-
ited areas, allowing researchers to focus their attention
on developing higher levels of autonomy beyond basic
vehicle navigation and control. Military applications of
unmanned robotics followed the track of the German V1,
with the advent of modern cruise missile technology.
From the mid 1980s on, the development of aerial
robots has followed an exponential pace, with one no-
table trend for operational aerial robots to systematically
find military applications as their most significant mar-
ket.
A snapshot of available aerial vehicle platforms and
Fig. 44.4a,b Autonomous helicopters: (a) Yamaha’s R-MAX systems appeared recently in [44.6], and regular updates
(b) Nascent Technologies Corp’s XS on available platforms can be found in the aerospace
Part F 44.3

literature. The chart in Fig. 44.3 illustrates the number of

From a robotics perspective, probably the next machines currently under development or in operation,
significant technological enabler is the advent of which exceeds 200 vehicle types. From this chart, one
lightweight processors and sensor systems, together with concludes, however, that the vast majority of current,
global navigation satellite systems, which allowed aerial operational aerial robots are fixed-wing aircraft, and that
robots to perform increasingly complex tasks. Japan, they tend to be present at all altitudes.
motivated by a policy of food self-sufficiency com- We must also remark that onboard robotic intelli-
bined with a massive shortage of agricultural workforce, gence has made its way not only into unmanned aircraft
took a lead in aerial robotics by developing highly reli- but also manned aircraft. Many commercial airliners
able helicopters in the 1980s, such as the Yamaha R-50 now have the ability to fly automatically from right after
and subsequent Yamaha R-Max (Fig. 44.4), with similar take-off (a decision left to the pilot) to right after land-
systems developed by other companies such as Yan- ing, by engaging the autopilot and letting the aircraft fly
mar. These robotic helicopters are used primarily for a predetermined profile. In addition, these vehicles are
crop dusting applications, especially over wet rice fields. now able to make systems-management decisions based
These vehicles also turned out to be very popular among on sensor inputs, sometimes escaping the human pilot’s
universities and other institutions for their unmatched ability to understand them.

44.3 Applications of Aerial Robotics

Listing all possible applications of aerial robotics ogy, geology, and agriculture [44.8, 9], as well as
is very challenging. However, there are fewer ac- unexploded mine detection [44.10].
tual implementations, because of the necessity for • Disaster response such as chemical sensing, flood
the corresponding operations to comply with strin- monitoring, and wildfire management.
gent air safety regulations. In the following, a brief • Surveillance such as law enforcement, traffic mon-
description of possible and current applications is pro- itoring, coastal and maritime patrol, and border
vided. patrols [44.11].
• Search and rescue in low-density or hard-to-reach
44.3.1 Possible Applications areas.
of Aerial Robots • Transportation including small and large cargo
transport, and possibly passenger transport.
The list of possible applications of aerial robots is long. • Communications as permanent or ad hoc communi-
According to [44.6,7], such applications fall within nine cation relays for voice and data transmission, as well
categories: as broadcast units for television or radio.
• Payload delivery e.g., firefighting or crop dusting.
• Remote sensing such as pipeline spotting, powerline • Image acquisition for cinematography and real-time
monitoring, volcanic sampling, mapping, meteorol- entertainment.
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.3 Applications of Aerial Robotics 1013

Military applications of aerial robots follow the same from model aircraft operations. Most often, the oper-
descriptive lines, with a particular emphasis on remote ated machines do in fact bear much resemblance to
sensing of humans and critical infrastructure, surveil- radio-controlled model airplanes. Other intermittent ap-
lance of human activity, and payload delivery (bombs, plications involve the use of unmanned vehicles for
missiles, and ad hoc ground infrastructures devoted to specific reconnaissance tasks, such as the detection of
communication and surveillance). fish banks from trawlers [44.12]. Such a task consti-
tuted one of the original purposes for the development
44.3.2 Current Applications of machines such as the Seascan unmanned aerial vehi-
cle.
Current applications of aerial robots are somewhat fewer Long-term scientific applications such as atmo-
and they are at present driven by the military context. spheric sampling experiments [44.13] appear to benefit
considerably from aerial robots. One report [44.13] reads
Aerial Observations
From March 6 to March 31 2006, we probed the
The most important application of aerial robots is aerial

Part F 44.3
polluted atmosphere over the North Indian Ocean
observations, which can then be used for terrain map-
with lightweight unmanned aerial vehicles (or UAVs)
ping, environmental surveys, crop monitoring, target
fully equipped with instruments. This UAV campaign
identification etc. There is a divide, however, between
launched from the Maldives laid a solid foundation
the state of the art for military applications and civilian
for the use of UAVs to study how human beings
applications, detailed below.
are polluting the atmosphere and their impact on
climate, including global warming.
Military Operations. Military and government use of
aerial robots has sharply increased in recent years in war Because such activities naturally require much planning
zones. As a result, dozens of vehicles are now delivered ahead, special permits can be obtained from aviation
every month, and end up flying in “hot” areas around the authorities within time limits that do not significantly
globe, most notably in Southwest Asia. The machines affect the overall experimental project. Other scientific
being flown range from man-portable machines flying missions led with success include [44.14], where the au-
at low altitudes, such as the Pointer or Raven aircraft, thors were able to survey Mount St. Helens (then active)
to mid-sized machines such as the Aerosonde, Seascan, by taking advantage of the temporary interdiction to fly
or Shadow unmanned vehicles, to larger-sized vehicles in the vicinity of the volcano.
such as the Predator or Global Hawk. Their wings span With the progressive introduction of aerial robots
from a meter or so for the smaller vehicles to 35 m for in the regulatory framework of many countries, we be-
Global Hawk (the same as a Boeing 737). Besides their lieve that intermittent applications of aerial robotics in
use in military areas or war zones, these machines now populated areas will eventually become commonplace.
find applications in border surveillance, with a particular However, this requires that flight authorizations be de-
interest in oceanic borders, where vehicles operate in livered within a fraction of the time duration of the
desert or quasi-desert areas. event: for example, firefighting operations are often trig-
gered within a few seconds of the fire alert. Permits
Civilian and Private Applications. Current civilian for aerial robotic support should therefore be delivered
applications of aerial robots for surveillance and ob- about as quickly if they are ever to be embraced by
servation remain sporadic and ad hoc: unlike many firefighters.
other robotic devices, civilian aerial robots do not op-
erate in closed environments but in civilian airspace, Payload Delivery
which is subject to strong safety regulations that do not Under the heading payload delivery, we find the nu-
yet systematically accommodate aerial robots. Conse- merous applications of aerial robots aimed at delivering
quently, the current trends in civilian aerial robotics are solid, liquid or gaseous products in areas that are
as follows. hard to reach for humans. So far, the most suc-
Small-scale, intermittent civilian aerial robotic cessful civilian application has been chemical crop
applications tend to happen in relatively isolated en- spraying using small unmanned helicopters. Leverag-
vironments (e.g., for film making or environmental ing the high costs and prices associated with crop
surveys), and often follow the safety and opera- culture in Japan, several thousand helicopters have
tions rules most familiar to their operators, derived been purchased by farmers, resulting in a profitable

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1014 Part F Field and Service Robotics

operation both for themselves and for the helicopter General Characteristics of Current Applications:
manufacturers, among them Yamaha and Yanmar. How- Level of Autonomy
ever, this application remains unique and involved the Most, although not all, aerial robots currently under
involvement of Japan’s government for it to be success- operation are automatically controlled as far as their
ful. dynamics are concerned. However, higher levels of au-
Besides this particular application, military appli- tonomy, such as path planning, object detection, and
cations form the bulk of unmanned aerial robotics for recognition and mission management involve human
the purpose of payload delivery, beginning in its crudest operators, who always remain in contact with the fly-
form with missiles, and evolving towards cruise missiles, ing machine. Thus not much distinguishes current aerial
able to navigate for thousands of miles and reach their robots from traditional manned aircraft, except that the
targets with high precision. One of the most talked-about pilot sits on the ground rather than in the air.
recent military application of aerial robots for payload As such, most of today’s operational aerial robots
delivery involves the Predator aircraft equipped with may be justifiably called remotely piloted vehicles
Hellfire missiles. (RPVs).
Part F 44.4

44.4 Current Challenges

In the following, we introduce six major challenges for rules by regulatory bodies until recently (2000), many
aerial robotics. This list is not meant to be exhaustive, researchers have operated under the rules of local radio-
but it reflects the current focus of researchers. How these controlled aircraft associations (e.g., the Academy of
challenges are addressed will be discussed later. Model Aeronautics). However, there is a rapidly grow-
ing trend for radio-controlled vehicles to incorporate
44.4.1 Regulations and Certification more onboard electronics, including radio transmitters
and sometimes guidance systems. In this environment,
A big challenge to the success of aerial robots is one can expect regulatory bodies such as the FAA to
doubtlessly their acceptance by certification authorities. continue to evolve their policies.
Indeed, the operation of aerial robots is currently signif- The ensuing challenge is for the research community
icantly limited by regulatory constraints. This is due to is to develop the requirements and subsequent tech-
the complex set of regulations put in place by national nology that meets the constraints set by the regulatory
agencies (e.g., the Federal Aviation Administration in agencies, or to propose and justify alternate constraints.
the US, the National Air Traffic Services in the UK, or In particular, the maturation of aerial robots leading
the Direction Générale de l’Aviation Civile in France), to their everyday use in populated areas will require
whose aim is to maintain very high levels of safety for the development of more reliable components, defined
air traffic. The downside of the excellent safety record maintenance procedures, formal training programs, and
reached by regulatory agencies is their (justified) risk the automation of emergency procedures (such as the
adversity, and therefore a slow acceptance of disruptive forced landing process). The core technology for UAS
technologies such as aerial robots. This is compounded already exists to demonstrate safety concepts. However,
by the current rapid pace of change and related lack of developing highly dependable systems – and making
standards among aerial robotic systems and how they such dependability guarantees acceptable to the regula-
are used. However, with the help of other organizations, tory authorities – is a current and urgent challenge.
such as the Radio Technical Commission for Aeronau-
tics (RTCA), regulatory agencies have moved forward 44.4.2 Human–Machine Interfaces
towards establishing rules for the routine operation of
aerial robots. Such rules include the ability for aerial The pilot interfaces used for manned aircraft have
robots to see and avoid or sense and avoid other traffic evolved continuously since the first manned aircraft.
at least as well as a human pilot. The standards that exist today directly benefit safety
In the recent past, many aerial robotics research and operator costs by minimizing operational errors and
activities and corresponding flight tests have occurred training time when transitioning between aircraft types.
at very low altitude. In the absence of clearly defined Despite ongoing development efforts [44.15], this can-
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.5 Basic Aerial Robot Flight Concepts 1015

not be said of aerial robotic systems, which run cover which must be able to keep operating well at unusual at-
a much wider range of autonomy, mission capability, titudes and under partial system failures such as loss of
and operator skill. Add to this the desire to have single actuation [44.16]. Researchers must develop automation
operators control multiple aircraft, and it clear this area systems that meet this need.
presents an ongoing challenge for researchers.
44.4.5 Obstacle Avoidance
44.4.3 Navigation
The ability for a vehicle to manage its position
Figuring out absolute and relative position is a central is- away from obstacles represents a significant issue and
sue for aerial robots, as it is for other robotics activities. a necessity for low-altitude operations in crowded en-
The existence of a significant manmade infrastructure vironments. One of the key features of aerial robots is
(the global navigation satellite system – GNSS) makes their possibly high speeds, which challenges many exist-
basic navigation easy but remains the subject of an in- ing sensor management and data processing algorithms,
tense debate; indeed, systems that overly depend on such especially their ability to detect hard-to-see obstacles

Part F 44.5
infrastructure lack resilience and tolerance to position- such as suspended cables quickly.
ing services shortage, whether such a shortage originates
from the system itself or from the particular robot config- 44.4.6 Aerial Robot Landing
uration (in cluttered environments such as cities). This and Interaction with Other Vehicles
situation will improve with the development of highly
reliable multimode navigation systems with built-in in- Owing to the finite endurance of aerial robots, land-
tegrity monitors, and with three independent satellite ing and docking are particularly important to them.
navigation constellations (Glonass, Galileo, and GPS) While landing constitutes an important element, dock-
currently deployed or under deployment. ing with other vehicles, such as during aerial refueling,
The challenge for researchers is to develop naviga- is also very important. All operations involving close
tion technologies that allow aerial robots to live without coordination and physical interaction between vehicles
manmade external navigation infrastructure, to handle or between a vehicle and the ground require further
the situations when it is not available. research.

44.4.4 Agile Flight and Fault Tolerance 44.4.7 Multivehicle Coordination

Nearly every aircraft in operational use today has been Several tasks require aerial robots to operate as a group,
challenged to fly far beyond its flight envelope during rather than as individual systems. This happens, for ex-
flight tests, including famous maneuvers such as that of ample, in order to create phased array antennas, or to
the Boeing 707 that, during early flight demonstrations perform object geolocation, or to improve the quality
to customers, performed a full barrel roll. The purpose of a surveillance service (e.g., fire monitoring). Other
of these demonstrations is not only to show the full capa- tasks requiring multivehicle coordination include the
bilities of the vehicle, but also to bring a sense of safety requirement for collision avoidance. More recently, mul-
to the pilots, that the aircraft is still able to perform well tivehicle coordination has been seen as a valuable way
after its goes into some upset condition. What applies to design aerial robotic systems that remain functional
to large, manned aircraft also applies to aerial robots, despite individual vehicle failures.

44.5 Basic Aerial Robot Flight Concepts

44.5.1 Aerial Robot Flight tors). A detailed description of vehicle flight mechanics
and the Importance of Scales is outside of the scope of this chapter; we can, nev-
ertheless, recall a few fundamental and useful notions
Like all flying machines, the performance of aerial critical to successful flight. The reference [44.17] is an
robots depends extensively on: (1) their size and (2) the excellent and entertaining introduction to the subject,
characteristics of their lifting mechanisms (wings, ro- while [44.18, 19] offer a more academic perspective on

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1016 Part F Field and Service Robotics

a factor of 4 only (22 ). So, if we were to fly this smaller

wing at the same speed, same altitude, and same angle
of attack as its big sister, the total generated lift must be
S/2ρV 2 , that is, twice as much as necessary to balance
out the effect of gravity.
Several solutions to this issue are possible: to reduce
the the actual wing dimensions at constant mass, slow it
down, or reduce its angle of attack.

Shrink the Wing

To obtain the proper lift (while keeping the speed and
angle of attack constant), we must shrink the wing area
by another
√ factor of two, or the wing dimensions by
a factor 2. Thus we already see one important conclu-
Part F 44.5

sion, which is that, at equal speed and angle of attack,

Fig. 44.5 Flying wing (Northrop’s YB47) and its shrunk the relative size of the wings with respect to the over-
version flying together all vehicle must shrink as the overall vehicle size goes
down. Borrowing again from [44.17], this explains much
the matter. One important quantity is the mass of a flying of why a Boeing 747 looks, with its large deployed
machine. Roughly speaking, the mass of a flying ma- wings and relatively narrow fuselage, like a condor
chine is proportional to its volume, and therefore grows while a B737 feels more like a puffin, and the smaller
like the cubic power of its size. Another quantity is the Embraer 145 is like a dart, as shown in Fig. 44.6. All
lifting forces that keep a vehicle up in the air; these are
proportional to the pressure exercised on the lifting sur-
face (rotor or wing), times the area of the lifting surface, a)
that is, roughly the second power of the vehicle size. The
pressure itself is proportional to the density of the sur-
rounding atmosphere (it need not be air only, think of
Mars), multiplied by the square of the average velocity
of the gas molecules relative to the lifting surface.
For illustrative purposes, consider the flying wing
shown in Fig. 44.5 and a notional scaled-down version
of it flying together. To make matters simpler, we assume
that the scaled-down wing is about half the size of the
full-sized wing. We now examine the impact of scales
on the way these wings must fly.
Consider for example the lift created by the full-
scale flying wing depicted in Fig. 44.5: it is proportional
to SρV 2 α, where S is its total surface, ρ is the air density,
V is the wing speed relative to the surrounding air, and
α is the angle of attack (roughly speaking the angle
b) c)
between the wing chord and the flow of air).
To get an idea of the importance of scales, and
following arguments developed in much greater detail
in [44.17], we now examine the requirements for the
scaled-down wing to fly at the same speed as the large
wing, assuming all its components are shrunk by a fac-
tor two in size as shown in the picture, and examine the
consequences of having to meet such requirements.
First, the mass of the wing roughly gets divided by Fig. 44.6a–c Relative fuselage and wing sizes for various
a factor 8 (23 ). However, its lifting surface has shrunk by aircraft: (a) Boeing 747: (b) Embraer 145: (c) Boeing 737
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.5 Basic Aerial Robot Flight Concepts 1017

tack. This makes it possible to fly about the same

speed with a scaled-down model of a flying machine.
However, this option comes with significant draw-
backs, especially for fixed-wing aircraft. In particular,
the sensitivity of the lift created by the wing to exter-
nal perturbations (e.g., air turbulence and wind gusts)
would again be higher, creating another recipe for bumpy
rides.
The previous considerations about the forces acting
on aerial vehicles also apply to moments: consider the
Fig. 44.7 Boeing 747 and dragonfly flying wings shown in Fig. 44.5, and assume that their
density (mass per unit volume) is constant throughout.
these aircraft can fly about the same speeds and altitude Their angular inertia about any axis are proportional
ranges. to the fifth power of their size. On the other hand,

Part F 44.5
the forces that apply to the wings are proportional to
Reduce the Speed their area; thus when moments are computed, forces are
If this option
√ is chosen, then speed must be divided by multiplied by distances, and the resulting moments be-
a factor 2 for our scaled model to balance lift and come proportional to volume, that is, the third power
weight. The consequences of reducing speed are many: of vehicle size. Consider then the angular momentum
the time required for mission completion of course in- equation
creases. On the other hand, the drag generated by the
J θ̈ = M , (44.1)
flying machine (and which must be paid for by the
propulsion system) goes down. where J is the moment of inertia of the vehicle and M
Pushed to their limits, the consequences of slowing is the applied torque. The term to the left of (44.1) de-
down the vehicle as it shrinks can be quite dramatic: creases much faster with vehicle size than that to the
consider a dragonfly (one of the role models for micro- right of the equation. As a consequence, we might im-
aerial robots) trying to land next to a Boeing 747 at the mediately conclude that the scaled-down flying wing
same airport. The figure below shows that the two share is inherently much more maneuverable than the larger
(very roughly) the same proportions. one, in the sense that it can change orientation much
For the sake of simplicity, assume that the dragonfly faster.
is the 1/1000 scaled-down version of the Boeing 747. This opens up a wealth of possibilities for robotics:
In order for both to fly level, and√ according to our venturing into the world of small flying robots, enabled
rule, the dragonfly must fly a factor 1000 = 32 slower by improvements of battery power and computation den-
than the B747. Assume the B747 flies at 500 km/h; sities opens new possibilities in terms of defining the
that makes the dragonfly fly at about 17 km/h. Imag- way these vehicles fly and interact with their environ-
ine now that the weather is gusty, with winds topping ment.
30 km/h. The 747 (and its passengers) will see little
variation in airspeed (from 470 to 530 km/h), and the 44.5.2 Propulsion Systems
variation in produced lift will be 33%, enough to shake
the aircraft a bit, but not unusually bad. As for the drag- Several propulsion systems exist for aerial robots, in-
onfly, the same gusts will create airspeed variations of cluding: jet, internal combustion, rocket, and electric.
well over 100%, and the produced lift will vary from Older but recurrent options also include pulse engines
zero to five or six times the nominal lift. A rough ride such as those used on the German V1.
naturally follows, and indeed, the flight of smaller ve- Owing to established aircraft and helicopter propul-
hicles often looks much less smooth than that of large sion technologies, internal combustion engines and jet
ones. engines form the bulk of the propulsion means for
medium to large-sized operational vehicles (50 kg or
Reduce the Angle of Attack more), allowing many of them to fly reliably over
The latter option, reducing the angle of attack, rests periods of several hours to several tens of hours.
upon the fact that, roughly, the lift created by a wing When considering operational robots, the kind of
(or a rotor), is a linear function of the angle of at- fuel used matters: preference is given to fuels al-

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1018 Part F Field and Service Robotics

a) 44.5.3 Flight Vehicle Types

and Flight Regimes
Several vehicle types form the bulk of aerial robots, in-
cluding fixed-wing machines, helicopters, flapping wing
systems, and combinations thereof. The boundaries be-
tween these vehicle types are, however, mostly inherited
from historical developments and intellectual stove-
piping, rather than any fundamental guidelines dictated
by the laws of mechanics and thermodynamics. For that
b) reason, it is easier and more logical to introduce dif-
ferent flight regimes than flight vehicles, although to
every regime there naturally corresponds one particular
vehicle.
Part F 44.5

There are essentially two flight regimes. In the first

regime, called hover, the speed of the vehicle relative to
the surrounding air is small, such that few or no forces
act on the vehicle except those resulting from the propul-
sion system itself. In the second regime, which we may
call cruising flight, there is a significant relative speed
between the vehicle and its surrounding environment,
and significant aerodynamic forces act on the vehicle;
these aerodynamic forces then largely dominate those
generated by the power system.

Hover
Hover is the condition when the vehicle body does not
Fig. 44.8 (a) Helios high-altitude long-endurance aircraft, move significantly with respect to the air mass sur-
(b) Mars aircraft [44.20] (Source: NASA) rounding it. Under these conditions, only propulsion
systems are available to keep the vehicle up in the
ready used in other devices, and preference goes to air (for heavier-than-air systems). Helicopters epitomize
heavy fuels, which are less prone to sudden and dan- these situations, and they are especially designed to sus-
gerous combustion or explosion, for example after tain such hover conditions over long periods of time.
a crash. Helicopters come in all sizes and shapes. Robotic he-
Electric propulsion systems, once unthinkable, have licopters are best represented by Yamaha’s R-50, and
become a reality for several small-sized aerial robots, now RMAX models (see, for example, Fig. 44.4), and
thanks to the development of affordable brushless both have been a staple of airborne robotics research
electric engines and lightweight batteries. Initially de- for years at several academic institutions because of
veloped for computer and communication applications, their reliability and available payload, which allows
these batteries have been very quickly adopted by small- them to carry many instruments of interest to robotics
sized (a few kg) aerial robots such as Aerovironment’s research (including navigation sensors such as video
Pointer and Raven aircraft, which are able to fly over cameras, laser range finders, and radars). With the evo-
periods exceeding one hour. The National Aeronautics lution of the economic and political context, one is
and Space Administration’s (NASA) Pathfinder un- bound to see other such machines abound in the fu-
manned aircraft combines lightweight electric engines ture. Indeed, the ability to hover is extremely useful for
with wing-mounted solar panels to yield the aircraft delivery/pickup of materials, rescue missions, and, in
shown in Fig. 44.8. general, any operations that require close proximity to
A notable departure from these propulsion systems rugged terrain.
is the Mars airplane’s propulsion system [44.20]: with Helicopters are not the only vehicles capable of
an inert, low-density atmosphere on Mars, such a vehicle hover, see for example the hover-capable fixed-wing air-
relies on a rocket engine for propulsion. craft in Fig. 44.9. Hovering aircraft, such as tailsitters,
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.5 Basic Aerial Robot Flight Concepts 1019

have been tested successfully since the 1950s at the very a) b)

least, and it is a classic trick for experienced remote con-
trol pilots to hover airplanes. Transitions from hover to
forward flight and back have been automated [44.21].
As radio-control (R/C) equipment shrinks in size and
mass, new generations of hovering vehicles will become
available. Some of these vehicles include micro air and
flapping wing vehicles.
Hovering flight typically not very fuel inefficient: c)
the fuel consumption of a hovering vehicle can exceed
that of a fixed-wing vehicle by an order of magnitude
of more. This kind of consideration has led manufac-
turers to seek some of the mixed configurations shown
in Fig. 44.9.

Part F 44.5
Cruising Flight Fig. 44.9a–c Non-helicopter, hover-capable vehicles: (a) Joint
During cruising flight the aerial robot mostly uses its Strike Fighter, (b) 1950s tailsitter aircraft and (c) Aurora Flight
available surfaces and its speed relative to the sur- Sciences’ Golden Eye 100
rounding atmosphere to generate lift and maintain
altitude. Unlike hovering flight, cruising flight usu- and the lift produced by the wing decreases as the
ally results in the aerial robot constantly meeting fresh angle of attack keeps increasing. This reduced aero-
air, which makes the range of adverse events to flight dynamic lift must then be compensated by increased
quite narrower. This, of course, is not true in the throttle, resulting in a situation where the aircraft pro-
case when aerial robots fly in formation, in which peller not only acts as a means to move the aircraft
case turbulence created by one robot may affect its forward, but also directly participates in maintaining
neighbor(s), sometime adversely, and sometimes pos- aircraft altitude.
itively [44.22]. This flight condition, whether experienced on a heli-
Robotic airplanes such as those shown in Fig. 44.8 copter or airplane, often results in important changes of
epitomize fuel-efficient cruising flight, with large and the effect of control mechanisms, for example, a stalled
highly optimized wings. Both aircraft are part of current Piper Tomahawk trainer aircraft at low throttle setting
NASA programs. While optimized for flight, the wings will experience ineffective ailerons, while its rudder
of the Mars aircraft must also be optimized for tight efficiency will shift from yaw axis to roll axis con-
packaging and deployment constraints at the end of its trol [44.23].
long trip from Earth to Mars.
While many fixed-wing systems are optimized for 44.5.4 Lighter-Than-Air Systems
cruising flight, any system in forward flight operates
according to the same principles; for example, a heli- One way to deflect some of the concerns associated
copter in forward flight operates like an airplane whose with high fuel consumption of heavier-than-air aircraft
wing is a flat disc spanning the area covered by its rotor. is to rely on lighter-than-air vehicles. While these ve-
Constant-velocity cruising flight that generates lift from hicles are often associated with spectacular accidents
the available aerodynamic surfaces is more fuel efficient and slow motion, they also offer an unmatched capa-
than hovering. bility to fly for long periods of time (more than 48 h)
and to do so silently [44.24, 25]. Establishing control
Stalled and High-Angle-of-Attack Flight over lighter-than-air vehicles can be, however, some-
This flight condition can be seen as a transitional flight what challenging. In particular these vehicles are quite
condition, where the characteristics of both forward sensitive to winds and often tend to go where the wind
flight and hover are present. Typically, an aircraft stalls takes them. Smaller platforms used for research must
when its tries to maintain altitude at low speeds: flying therefore evolve in closed environments [44.26]. The
level at lower speeds forces the aircraft’s angle-of-attack use of such vehicles for outdoor research can be quite
to increase for the wings to produce more lift. How- daunting, because their large size requires considerable
ever, past a critical angle of attack, the trend reverses infrastructure to store them.

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1020 Part F Field and Service Robotics

44.6 The Entry Level for Aerial Robotics: Inner-Loop Control

Aerial robots exhibit the complex flight dynamics asso- and three orthogonally mounted gyroscopes (more may
ciated with flight vehicles. As a consequence, precise be used for the purpose of achieving redundancy). The
motion control rapidly becomes a necessity for any accelerometer suite measures, up to sensor error, the ac-
aerial robotics activity to occur successfully. This means celerations experienced by the vehicle at the location
that effort must go into reliable basic flight control be- of the inertial sensor minus gravity. Gyroscopes mea-
fore more advanced, intelligent mission management sure vehicle angular velocities. Modern inertial sensors
can be attempted. Inner-loop control is achieved by the are usually rigidly linked to the vehicle to form strap-
right sensing equipment, and by adequate control algo- down inertial measurements systems. Such systems have
rithms. Efficient hovering vehicles tend to be unstable, become very cheap and very popular. Unlike many
which makes their stabilization more difficult than that nonflying applications, unaided inertial measurement
of purely cruising vehicles, for which many commercial packages are not sufficient for estimating the attitude of
control packages are now commonly available. an airborne vehicle. Indeed, consider Fig. 44.10, show-
Part F 44.6

ing two helicopters equipped with inertial measurement

44.6.1 Sensing and Estimation systems. One, straight up, is hovering. The second, up-
side down, races towards the ground with an acceleration
Airborne robots come with a variety of sensing options, of 2g. The accelerations and rotation rates recorded by
which include the onboard inertial measurement unit will be strictly
the same in both cases. While attitude can be estimated
• inertial navigation systems (gyroscopes, accelerom-
by integrating angular rates over time, the approach will
eters)
eventually fail without correcting for accumulated error
• global navigation satellite systems (GLONASS,
from another source. It remains that inertial measure-
GPS, Galileo)
ment units are extremely useful to measure variations in
• terrestrial radio navigation systems (VHF om-
acceleration and angular velocities, and constitute a sta-
nidirectional range (VOR), distance measuring
ple of inner-loop control systems. Small radio-controlled
equipment (DME), instrument landing system (ILS))
helicopters now come with built-in gyroscopic yaw
• air data probes and altimeters
dampers that make their manual operation much more
• radar and passive vision sensors
manageable. A key progress was made when Analog
• magnetic compasses
Devices introduced a low-cost micromechanical gyro-
• distance measuring (altitude radars, ultrasonic sen-
scope [44.27]. Since then, this technology appears to
sors, and laser range finders)
have made its way into most commercially available
The choice of sensors is critical to obtaining a properly inertial measurement units (IMUs), thereby greatly re-
flying robot. Usually, the same suite of sensors may not ducing their cost and weight. Practical inertial navigation
apply to all phases of flight. We will concentrate our systems on aircraft typically receive at least position
discussion on the first four sensor types. updates from other sensors (discussed below), called
inertial aiding.
Inertial Navigation Systems
Inertial measurement systems consist of a combination Global Navigation Satellite Systems
of usually three orthogonally mounted accelerometers The Global Positioning System ( GPS) and its Rus-
sian equivalent GLONASS and future European Galileo
space-based systems offer real-time absolute position
information, using a constellation of satellites circum-
navigating the earth. Ever since the beginning of their
operation, global navigation systems (GNS) have been
a=0 a = 2g g the object of a debate concerning their use in aerial
robotics, with many researchers recommending against
using such a large manmade navigation infrastructure
to achieve true autonomy. Their arguments tend to be-
Fig. 44.10 Two inertially equivalent helicopter configura- come justified by the occurrence of recent needs in such
tions applications as Mars exploration and low-altitude flight
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.6 The Entry Level for Aerial Robotics: Inner-Loop Control 1021

in obstacle-laden environments (such as cities) where

satellite-based navigation is often unavailable. Wher-
ever they are available, however, satellite navigation
systems are a convenient and cheap means for a ve-
hicle to locate itself. This modest investment has often
been the enabler of automatic flight for many researchers
and is currently used by virtually all existing indus-
trial systems. Pushed to their limits, satellite navigation
systems have been shown to achieve the entire range
of desired navigation and sensing functions, which in-
clude vehicle position and attitude: GPS-only flight for
a small helicopter robot was achieved in 1995 at Stanford
University [44.28]. Fig. 44.11 Air data probe (Source: NASA Dryden Flight
Research Center). The vanes are used to measure angle of

Part F 44.6
Altimeter and Air Data Probes attack and sideslip angle
Pressure-measuring devices are immensely useful sen-
sors in aerial robotics. With ingenious arrangements of ing at tracking relatively invariant features such as the
pressure sensors (such as pitot tubes) it is possible to horizon [44.29].
measure (1) the atmospheric pressure at the location of
the robot and (2) the so-called dynamic pressure, ρv2 /2, 44.6.2 Estimator Design
along all vehicle axes. These data can themselves be
transformed into precious information about the aerial The individual inputs collected from each sensor are
robot’s altitude and direction of motion relative to the usually not sufficient to estimate the state of the vehicle.
air it is flying in. Depending on the vehicle used, air data Different sensors may be efficient over different flight
probes may be challenging to build and constitute an in- regimes. The proper way to leverage individual informa-
teresting field of investigation. Indeed, pressure probes tion provided by each sensor is through an appropriate
are very sensitive to flow perturbations generated by filtering process that can yield rather comprehensive in-
fuselage, wings, and most importantly rotors and pro- formation about the entire system’s state. Unlike ground
pellers. Air data probes are therefore positioned as far robots, the necessity for good robot state estimates arises
away from the main elements of the vehicle as possi- early in the robot development process since closed-
ble (for example, along a boom extending forward of loop flight would be impossible otherwise. However,
the vehicle fuselage). Figure 44.11 shows one such air the structure of the filters is usually a great deal simpler
data probe configuration. Mounted together with iner- than their ground-based equivalent, since the difficulty
tial measurement systems, air data probes allow aircraft of flight is compensated by a rather simple and uniform
to maintain stable flight at a prescribed altitude. With environment structure. As a consequence, simple filters
the current state of technology, they remain somewhat such as (extended) Kalman filters are usually enough
insufficient to achieve, alone, stable hovering flight for for a large number of applications and quickly enable
helicopters. flight [44.30].

Passive Vision 44.6.3 Inner-Loop Control

Passive vision has become a very popular sensor for
inner-loop control. Even unsophisticated light sensors Inner-loop control of aerial vehicles naturally builds
able to differentiate between the intensity of infrared upon the previously discussed state estimator, and is
activity from the ground versus that emitted by the relatively easy for routine flight operations. By this we
sky has made its way into small commercial prod- mean that the process by which a good vehicle controller
ucts, mostly aimed at assisting remote-controlled vehicle is obtained only requires following standard textbook
flight. As will be discussed later, passive vision devices techniques such as proportional, integral, and derivative
have also found applications for vehicle–obstacle and control or linear quadratic control [44.31–33], appro-
vehicle–vehicle proximity management, and for land- priately scheduled against essential parameters such as
ing applications. Recent research aimed at using vision vehicle speed and altitude. As a result, several research
for inner-loop control applications includes work aim- groups, and now several companies, have built basic

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1022 Part F Field and Service Robotics

guidance and control packages suitable for aerial robots, experimental application of adaptive and learning con-
both fixed-wing and helicopter. trol techniques [44.21, 34, 35], which offer stable
Among the notable recent advances for the inner- controlled helicopter flight from very coarse initial ve-
loop control of aerial robots, we find the successful hicle dynamics knowledge.

44.7 Active Research Areas

This section presents some of the active research in originally designed for manned systems, to aerial robots.
aerial robotics. Such research efforts aim at answering Such systems are based on cooperative position informa-
the challenges outlined in Sect. 44.4. tion sharing between aircraft extracted from radar-based
navigation and surveillance systems. The reason for em-
44.7.1 Interfacing phasizing such systems over other, newer technologies
with the Human Infrastructure: is that they have already undergone extensive, and ex-
Part F 44.7

Meeting the Regulations pensive, development, validation and testing. As such,

several unmanned vehicles, such as the Global Hawk
While manned flight operations indeed have an excellent unmanned aircraft, are now fitted with ACAS sys-
safety record, the price paid for this safety is a strong tems [44.37]. However, such systems are not automated,
specialization of those regulations to human-operated meaning that the remote human pilot ultimately decides
systems, and a slow evolution of these regulations to- whether to execute the maneuvers recommended by the
wards accepting aerial robots of all sizes. Currently none system. The possibility of completely automating such
of the existing aerial robots is able to meet these regu- collision avoidance systems is the object of recent stud-
lations in the absence of a human pilot. This includes ies [44.38], with the clear intent of fitting them on aerial
the ability to see (or sense) and avoid other aircraft, robots. However, the weight of ACAS system hard-
comply with air-traffic rules, and operate harmoniously ware, as well as the power required to operate them,
with the current ground-based, manned air-traffic con- makes such systems suitable only for large vehicles. In
trol system [44.4, 5, 36]. In the case of the US however, the context of rapidly evolving technology (such as the
the FAA has acknowledged the economic potential generalization of positioning information using global
of aerial robots by opening an office specifically de- navigation satellite systems), current ACAS technology
voted to such systems (the Unmanned Aircraft System may also become rapidly obsolete.
Group), and by delivering permits to fly over certain ar-
eas (especially disaster areas) within a couple of hours Sense and Avoid
of a request. However, the FAA currently emphasizes The idea, encouraged by institutional service providers,
access for remotely piloted machines (as opposed to is to ensure that aerial robots are able to detect the pres-
fully autonomous machines), where a ground-based pi- ence of other traffic and avoid it as necessary and at
lot must have the means to communicate by voice with least as well as a human pilot. Several candidate sens-
the FAA control center in charge of the geographical ing technologies are currently in development, including
area where the vehicle is operated. passive vision systems [44.39–42], in an effort aimed at
Efforts to help aerial robots improve their interface making aerial robots as able as humans to avoid other
with other vehicles include the adaptation of existing traffic when the sky is clear (visual flight rules). While
systems to prevent mid-air collisions between aerial much can be done in the visible spectrum, concerns over
robots and other traffic, the development of see-and- vehicle flight in clouds raises the necessity to consider
avoid procedures, and means to interact with an aerial other frequency bands, such as the near infrared, if aerial
robot as one would interact with a human pilot (natural robots will need to be able to detect other traffic better
language interfaces). We now detail these efforts. than a human pilot [44.43].

Collision Avoidance Human Interfacing

for Remotely Piloted Vehicles Another active research venue is to facilitate the in-
The most immediate efforts aimed at inserting aerial terfacing of robots with humans (e.g., an aerial robot
robots in the civilian airspace consists of adapting ex- interacting with a human air-traffic controller). Re-
isting airborne collision avoidance systems (ACAS), cent research in human–aerial robot interaction has
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.7 Active Research Areas 1023

shown that aerial robots can interact productively with The corresponding levels of achievable agility have
humans, by combining natural language processing in- evolved correspondingly. By the early 2000s, basic
terfaces with advanced vehicle path and task planning aerobatic maneuvers became feasible [44.47], and by
capabilities [44.44–46]. A natural interface might cap- 2007 fully fledged aerobatics had been reported [44.48].
ture standard, unambiguous phraseology such as North Other efforts involving unusual flight attitudes and fault
Atlantic Treaty Organization (NATO) phraseology or recovery include those of Chiba University (Japan),
air-traffic control phraseology. The impact of such tech- who demonstrated autorotation landings for autonomous
nology on aerial robots would be profound since they helicopters [44.49].
would then be able to enter airspace with little or no Parallel to rotorcraft agile flight, several efforts have
visibility and be able to interact with the predomi- also successfully enabled aerial agility for fixed-wing
nantly ground-based, human-intensive air-traffic control robots [44.21].
structure.
44.7.3 Take-Off, Landing,
44.7.2 High-Agility Flight and Interaction with Other Vehicles

Part F 44.7
One of the important characteristics of aerial robots One of the richest current areas of investigation for aerial
is the ability to operate at the limit of its structural robots involves vehicle operation next to other vehicles
strength, unimpeded by the presence and physiolog- or infrastructures. These operations include take-off,
ical limitations of a human pilot. This allows aerial landing, docking, and separation.
robots, especially small ones, to operate very aggres-
sively. As a result, several research groups have explored Take-Off and Landing
the possibility of achieving aggressive flight with either Take-off and landing experimentation and research is
fixed-wing or rotary-wing vehicles. The key factors that proving particularly interesting for small-sized aerial
have enabled the onset of highly aggressive flight has robots. Indeed, the dynamics of smaller vehicles enable
been the emergence of lightweight computing environ- strong departures from conventional, manned-vehicle
ments and sensors, notably GPS and inertial systems. take-off and landing operations, for example, most fixed-
Indeed, aggressive flight (where aggressive flight refers wing unmanned aerial vehicles under 5 kg are better off
to any abrupt change in vehicle attitude) is closely re- simply flying into the ground than attempting to land in
lated to available vehicle mass and size, as discussed a smooth fashion. One of the best illustrations of how
earlier. vehicle landing procedures may dramatically change for
Figure 44.12 shows the evolution of three helicopter smaller-sized aerial robots is Insitu’s and Hood Technol-
configurations over time. While the vehicle platform ogy’s Skyhook concept: small, fixed-wing aerial robots
has evolved little or not at all, the onboard avion- are recovered by allowing them to catch a vertical cable
ics has progressively shrunk. In the mid-1990s, the with the tip of one of their wings [44.12]. The cable itself
onboard avionics typically would weigh the same or is held by means of a crane, itself mounted on a surface
more than the helicopter mass. By the early 2000s, vehicle (e.g., truck or ship). At take-off, similar scaling
the onboard avionics would be about half of the vehi- considerations apply, with many fixed-wing vehicles be-
cle mass, while by the mid-2000s the onboard avionics ing launched by hand or by means of a catapult. One of
represent only a small fraction of the helicopter mass. the consequences of the increased tolerance of small ve-

a) b) c)

Fig. 44.12a–c Avionics versus vehicle. (a) Stanford Helicopter (c. 1995). (b) MIT helicopter (c. 2001). (c) Stanford
Helicopter (c. 2006)

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1024 Part F Field and Service Robotics

challenging situations (e.g., sloped terrain or moving

platforms), traditionally handled by humans in large
platforms, remain difficult for aerial robots. For this rea-
son, the helicopter landing problem has attracted the
attention of many research teams. On the one hand,
there have been many efforts combining advanced sens-
ing environments [44.51–55] with advanced control
algorithms to enable affordable landing in structured
environments which are not simply horizontal landing
pads. On the other hand, identifying suitable landing
places in unprepared environments by means of remote
sensing and signal processing is also an area of active
research [44.56–58].
Part F 44.7

Operations in the Vicinity of Other Vehicles:

Docking and Undocking
Docking operations for unmanned aerial vehicles are
necessary to improve their range and autonomy. Indeed,
it is conceivable that some optimal aerial robot configu-
ration consist of a parent–child system, whereby a larger
machine provides a primary deployment and retrieval
mechanism for several smaller vehicles. Such a con-
Fig. 44.13 Skyhook system in action (courtesy Insitu, Inc.) cept has existed for a long time, with airships acting as
carriers for smaller aircraft [44.59]. More recently how-
hicles to crash landings is also their reduced need for ever, it is in-flight aerial refueling that has motivated
high-resolution navigation information, for example, it recent research on docking aerial robots. Indeed, the
has been shown possible to land small-sized vehicles on possibility for such vehicles to refuel considerably in-
a designated target with monocular vision only [44.50]. creases their operational range [44.60–62]. The NASA
Helicopter robots have, so far, not benefitted from Dryden Flight Research Center has recently reported
the same kind of developments, and much of their take- the completion of the first vision-aided fully automated
off and landing procedures are similar to their larger aerial refueling operation, using computer vision for
counterparts. The main reason may be attributed to the the purpose of recognizing and tracking the fuel hose
presence of a fragile rotor that spins at high speed,
and that must avoid contact with other vehicles or the
ground. Many of the current robotic helicopter landing
procedures simply consist of hovering above the land-
ing area, then commanding a limited descent rate until
the vehicle records it has touched the ground. More

a) b)

Fig. 44.14a,b Automatic airborne refueling. (a) Typical refueling Fig. 44.15 Georgia Tech’s unmanned helicopter in
configuration. (b) Camera view of refueling basket (Source: NASA a parent–child configuration. The child is a hover-capable
Dryden Flight Research Center) ducted fan
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics 44.7 Active Research Areas 1025

that must be captured by the aerial robot, as shown problem, including [44.65–67] and many others. An-
in Fig. 44.14. other key issue in aerial robot trajectory planning arises
Undocking operations are comparatively easier to when there is a discrepancy between the complexity
perform. They remain, however, spectacular since the of the environment and the maneuvering space needed
dynamics of the aerial robot dramatically change as for the vehicle. When planned for finite time or geo-
it is dropped from its mother ship. An extreme exam- graphical horizons, it becomes important that a planner
ple of such a situation is illustrated by Georgia Tech’s constantly keep a feasible loitering solution within the
successful dropping of a small ducted fan aerial robot known environment [44.46].
from a larger autonomous helicopter. The small ducted
fan then successfully stabilized itself. Pictures of this Multirobot Path Planning and Coordination
experiment are shown in Fig. 44.15. There has recently been a surge in research activities
for multivehicle path planning and coordination. Such
44.7.4 Reactive Flight research activities have been motivated by problems as
in Cluttered Environments diverse as the generation of noncolliding paths, the gen-

Part F 44.7
and Obstacle Avoidance eration of swarming behaviors for applications such as
phased-array, robot-borne antenna systems, collabora-
Flight in cluttered environments includes any phase of tive target detection and prosecution, and collaborative
the flight where vehicles are in close proximity to ob- search for thermal currents.
stacles. This flight mode is particularly important for This rich literature, of which only a few references
low-altitude applications. Several achievements have have been cited, stems from the conjunction of several
been reported in this area in the recent past, using constraints in the problem under study, including
a variety of sensing techniques.
• highly constrained dynamical systems (with re-
Among the first significant works relying on pas-
stricted radius of curvatures and minimum speed
sive vision techniques, Beard and McLain’s certainly
requirements, for example)
stands out as one of the most entertaining and spec-
• a variety of information management possibilities
tacular [44.63], using fixed-wing vehicles performing
(including centralized, decentralized, distributed in-
autonomous flight within a canyon using low-cost, opti-
formation)
cal flow computation techniques.
• catastrophic consequences in case of failures
Other institutions involved with active as well as
passive sensing techniques for vehicle navigation in Initial work aimed at studying aerial-robot coor-
cluttered environments and obstacle avoidance include dination from the perspective of mission execution
Carnegie-Mellon University [44.64], where the authors include [44.68–71]. Swarming behaviors, or the ability
report fast vehicle flight in highly cluttered environ- for a vehicle group to generate a coherent, consensual
ments, including obstacles as difficult to deal with as behavior using only local information, has become the
suspended cables. The NASA Ames research center focus of much attention in the research community since
also recently reported successes along similar lines as the recent paper [44.72].
part of their work on adaptive landing in unprepared Collision avoidance has also formed the motivation
environments [44.56–58]. for much research in multirobot coordination, see for
example [44.73–75].
44.7.5 Path Planning and Higher-Level
Planning Capabilities 44.7.6 Integrated Aerial
Robotic Operations:
Single-Robot Path Planning Aerial Robotics Contests
Path planning for aerial robots resembles path planning
for any robot, with the following distinctive character- Research on the ability of aerial robots to perform com-
istics: aerial robots are able to fly very fast (or may pletely autonomous missions, especially at low altitudes,
have to fly fast). Thus there is the distinct possibility is clearly well represented in contests such as the Inter-
of significant discrepancies between intended and ac- national Aerial Robotics Competition, initiated in 1991
tual trajectories. The vehicle dynamics must be fully by Michelson [44.1]. In this contest, universities, pos-
accounted for when designing trajectories. Several path sibly supported by industry and government, compete
planning concepts have been proposed to handle this against each other by demonstrating how their vehicles,

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1026 Part F Field and Service Robotics

or vehicle systems, meet the requirements of the com- ners include Carnegie-Mellon University, the Georgia
petition. A basic tenet of the competition is that the Institute of Technology, MIT/Draper Laboratory, and
small aerial robotic systems entrants must be capable of Stanford University.
complete autonomy (no human interaction) during the Other aerial robotic competitions have since been
mission. established. For example, the French governmental orga-
The rules of the competition have evolved from nization Delegation Generale pour L’Armement (DGA),
the inception of this effort to reflect advances in the together with the Supaero and Ecole Nationale Su-
capabilities of the proposed systems. One of the key perieure Des Constructions Aeronautiques (ENSICA)
characteristics of the competition is that it has always engineering schools have proposed a contest involving
emphasized the simultaneous demonstration of several very small-sized aerial robots in 2004, with a focus on
robotic functionalities, including basic mission execu- their flight mechanics at various flight regimes. The gov-
tion, object reconnaissance and detection, and object ernment of Queensland, Australia together with the local
manipulation. During the early days of the competition, research organizations Commonwealth Scientific and
the task asked for an aerial robot to recognize and pick Industrial Research Organization (CSIRO) and Queens-
Part F 44.8

up an object in a designated area, and carry it to an- land University of Technology launched a new contest
other designated area. As universities were able to meet focusing on search-and-rescue missions in 2007.
the initial challenges posed by the competition rules, The distribution of vehicle types involved in these
the rules have evolved to a higher level of sophistica- contests is very different from the distribution of op-
tion. As of today, the competition rules require complete erational aerial robots. While operational aerial robotic
autonomous operation of the vehicles over longer dis- systems are overwhelmingly of fixed-wing type, the ma-
tances. The vehicles must now find and reach a village chines used by universities during these contests offer
from a distance of three kilometers. They must also a much more balanced distribution of aircraft and ro-
evolve towards higher reasoning capabilities about the torcraft. Several reasons contribute to these differences
objects and events being encountered. Moreover, em- and they have been outlined earlier. The operation of
phasis has been placed on multimodal robotics, since fixed-wing aircraft at relatively high altitude, for re-
the robotic system must be able to enter a building and connaissance and surveillance missions offers a large
explore it, a task currently best performed by ground and technologically easy market to reach, although it
robots. faces significant regulatory constraints. In comparison,
Recognizing the growing gap between experienced the operation of small vehicles in cluttered environ-
participants and new entrants, several different competi- ments definitely favors hovering-like vehicles. However,
tion levels have been established. While US participation these vehicles, like other robots, face significantly more
in the competition is predominant, several non-US par- constraints in terms of environment sensing, obstacle
ticipants are also present, including Germany, England, avoidance, and task planning and execution complexity.
Switzerland, Canada, and India. In 2000, the Technische As such, they are closer to the realm of basic research
Universitaet Berlin won the contest. Other contest win- typical of universities.

44.8 Conclusions and Further Reading

Aerial robots represent a very interesting and excit- to insert them into airspace occupied by other traffic
ing area of robotics, involving very dynamic platforms such as manned systems. The resulting technical re-
whose size ranges from a few centimeters to several search challenges include the development of a proper
tens of meters. It seems highly probably they will con- and affordable sense-and-avoid technology, and the abil-
tinue to see new applications, beginning with those that ity for aerial robots to be conversant with other traffic
happen in relatively unpopulated areas and relatively and the ground control infrastructure.
high altitudes. The current applications of aerial robots Lower-altitude aerial robotics, often operating in
are focused primarily on military operations. However, cluttered environments, offers the opportunity to ex-
an ambitious civilian market led by Japan is currently plore many generic robotics topics, including vision,
burgeoning. path planning, mapping, and other algorithms in a pro-
Aerial robots currently pose a challenge to all reg- gressive manner, while offering potential benefit of still
ulatory agencies, which must find modalities and rules more important applications.
Springer Handbook of Robotics
1 Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics References 1027

A very dynamic research and development field, of aerial robotics and its applications clearly indicates
aerial robotics can be seen from a historical perspective that the field is still very young, that operational expe-
by reading [44.3]. A snapshot of current UAV technol- rience is slowly building up, and that many challenges,
ogy can be obtained, for example, from [44.76]. The most notably regulatory and safety challenges, must still
lack of a known comprehensive, book-like presentation be overcome.

References

44.1 R.C. Michelson: International aerial robotics com- actions Simultaneously from Three Stacked Au-
petition - The world’s smallest intelligent flying tonomous Unmanned Aerial Vehicles (AUAVs), Tech.
machines, 13th RPVs/UAVs International Confer- Rep., Center for Atmospheric Sciences and the
ence, UK (1998) pp. 31.1–31.9 Center for Clouds, Chemistry and Climate Scripps In-
44.2 J.D. Anderson: Samuel Pierpont Langley - Amer- stitution of Oceanography, University of California,
ica’s first aeronautical engineer, 38th AIAA San Diego (2006)

Part F 44
Aerospace Sciences Meeting and Exhibit, Reno 44.14 M. Patterson, A. Mulligan, J. Douglas, J. Robinson,
(2000) paper 0164 J. Pallister: Volcano Surveillance by ACR Silver Fox,
44.3 L.R. Newcome: Unmanned Aviation, a Brief History AIAA Infotechs., Aerospace, Arlington (2005) paper
of Unmanned Aerial Vehicles (American Institute of 6954
Aeronautics and Astronautics, Reston 2004) 44.15 North Atlantic Treaty Organization: Standard In-
44.4 R.E. Weibel, R.J. Hansman: Safety considerations terfaces of UAV Control System (UCS) for NATO UAV
for operation of different classes of UAVs in the Interoperability, NATO STANAG, 4586 (2004)
national airspace system, AIAA 3rd Unmanned Un- 44.16 T. Jordan, J. Foster, R. Bailey, C.M. Belcastro:
limited Technical Conference, Chicago (2004) paper AirSTAR: A UAV platform for flight dynamics and
6244 control system testing, 25th AIAA Aerodynamic
44.5 R.E. Weibel, R.J. Hansman: An Integrated Approach Measurement Technology and Ground Testing Con-
to Evaluating Risk Mitigation Measures for UAV Op- ference, San Francisco (2006) paper 3307
erational Concepts in the National Airspace System, 44.17 H. Tennekes: The Simple Science of Flight: From
AIAA Infotechs., Aerospace, Arlington (2005) paper Insects to Jumbo Jets (MIT Press, Cambridge 1992–
6957 97)
44.6 D. Hughes: UAVs face hurdles in gaining access to 44.18 J.D. Anderson: Introduction to flight (McGraw-Hill,
civil airspace, Aviation week (2007) New York 1989)
44.7 K.C. Wong, C. Bil, G. Gordon, P.W. Gibbens: Study of 44.19 R.C. Nelson: Flight Stability and Automatic Control
the Unmanned Aerial Vehicle (UAV) Market in Aus- (McGraw Hill, New York 1998)
tralia, Aerospace Technology Forum Report (1997) 44.20 R.D. Braun, H.S. Wright, M.A. Croom, J.S. Levine,
44.8 R. Sugiura, N. Noguchi, K. Ishii, H. Terao: De- D.A. Spencer: The Mars airplane: A credible science
velopment of remote sensing system using an platform, IEEE Aerospace Conference (Big Sky 2004)
unmanned helicopter, J. JSAM 65(1) (2003) p. 408
44.9 R. Sugiura, T. Fukagawa, N. Noguchi, K. Ishii, Y. 44.21 A.D. Wu, M.A. Turbe, S. Kannan, J.C. Neidhoefer,
Shibata, K. Toriyama: Field information system us- E.N. Johnson: Flight Results of Autonomous Air-
ing an agricultural helicopter towards precision plane Transitions to and from Vertical Hover, AIAA
farming, IEEE/ASME International Conference on Guidance, Navigation and Control Conference, Key-
Advanced Intelligent Mechatronics (Kobe 2003) pp. stone (2006) paper 6775
1073–1078 44.22 W.R. Williamson, M.F. Abdel-Hafez, I. Rhee, J.D.
44.10 K. Schutte, H. Sahli, D. Schrottmayer, F.J. Varas: Wolfe, J.L. Speyer: A Formation Flight Experiment
ARC: A camcopter based mine field detection sys- Using Differential Carrier Phase for Precise Relative
tem, Fifth International Airborne Remote Sensing Navigation, Institute of Navigation - GPS meeting
Conference, San Francisco (2001) (2002)
44.11 S.E. Wright: UAVs in Community Police Work, AIAA 44.23 J. Sanderson: Private Pilot Manual (Jeppesen
Infotechs., Aerospace, Arlington (2005) paper 6955 Sanderson, Englewood 1984)
44.12 T. McGeer, S. Sliwa, D. Sliwa: Commercial ap- 44.24 I. Schafer, R. Kuke, P. Lindstrand: Airships as Un-
plications for UAVs within the fishing industries, manned Platforms: Challenge and Chance, AIAA’s
AIAA’s 1st Technical Conference and Workshop on 1st Technical Conference and Workshop on Un-
Unmanned Aerospace Vehicles, Portsmouth (2002) manned Aerospace Vehicles, Portsmouth (2002)
paper 3401 paper 3423
44.13 V. Ramanathan: Maldives AUAV Campaign (MAC): 44.25 N. Mayer: Lighter-than-air systems, Aerosp. Am.
Observing Aerosol-Cloud-Radiation-Climate Inter- 12, 30–31 (2005)

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1
1028 Part F Field and Service Robotics

44.26 E.A. Olsen, C.-W. Park, J.P. How: 3D Formation erations Conference and Workshop, Portsmouth
Flight using Differential Carrier-phase GPS Sen- (2002)
sors, Institute of Navigation GPS Meeting, Nashville 44.42 J. Utt, J. McCalmont, M. Deschenes: Test and Inte-
(1998) gration of a Detect and Avoid System, AIAA Third
44.27 J.A. Geen, S.J. Sherman, J.F. Chang, S.R. Lewis: Unmanned Unlimited Technical Conference, Work-
Single-chip surface micromachined integrated gy- shop and Exhibit, Portsmouth (2004) paper 6424
roscope with 50/spl deg/h Allan deviation, IEEE J. 44.43 J.F. McCalmont, J. Utt, M. Deschenes, M.J. Taylor:
Solid-State Circ. 37(12), 1860–1866 (2002) Sense and avoid technology for Global Hawk and
44.28 A. Conway: Autonomous Control of an unstable Predator UAVs, Infrared Technology and Applica-
model helicopter using carrier-phase GPS only. tions, XXXI SPIE Conference (2005) pp. 684–692
Ph.D. Thesis (Stanford University, Stanford 1995) 44.44 P. Doherty, G. Granlund, K. Kuchcinski, E. Sande-
44.29 T.P. Webb, J. Prazenica, A.J. Kurdila, R. Lind: wall, K. Nordberg, E. Skarman, J. Wiklund: The
Vision-Based State Estimation for Autonomous Mi- WITAS Unmanned Aerial Vehicle Project, 14th Eu-
cro Air Vehicles, AIAA Guidance, Navigation and ropean Conference on Artificial Intelligence (Berlin
Control Conference (San Francisco 2005) paper 5869 2000) pp. 747–755
Part F 44

44.30 B. Anderson, J.B. Moore: Optimal Filtering 44.45 E. Sandewall, P. Doherty, O. Lemon, S. Peters:
(Prentice-Hall, Upper Saddle River 1979) Words at the right time: Real-time dialogues with
44.31 A.E. Bryson, Y.C. Ho: Applied Optimal Control the WITAS unmanned aerial vehicle. In: KI 2003:
(Hemisphere, Washington 1975) Advances in Artificial Intelligence, ed. by A. Gunter,
44.32 A.E. Bryson: Control of Spacecraft and Aircraft R. Kruse, B. Neumann (Springer Lectur Notes in
(Princeton Univ. Press, Princeton 1994) Artificial Intelligence, Berlin 2003) pp. 52–78
44.33 G.F. Franklin, J.D. Powell, A. Emami-Naeni: Feed- 44.46 T. Schouwenaars, M. Valenti, E. Feron, J. How,
back Control of Dynamic Systems (Addison-Wesley, E. Roche: Linear programming and language
Reading 1986) processing for human/unmanned-aerial-vehicle
44.34 E.N. Johnson, A.J. Calise: Limited Authority Adap- team missions, AIAA J. Guid. Control Dyn. 29(2),
tive Flight Control for Reusable Launch Vehicles, 303–313 (2006)
AIAA J. Guidance Control Dyn. 26(6), 906–913 (2003) 44.47 V. Gavrilets, B. Mettler, E. Feron: Human-inspired
44.35 A.Y. Ng, A. Coates, M. Diel, V. Ganapathi, J. Schulte, control logic for automated maneuvering of a
B. Tse, E. Berger, E. Liang: Inverted autonomous miniature helicopter, AIAA J. Guid. Control Dyn.
helicopter flight via reinforcement learning. In: 27(5), 752–759 (2004)
Experimental Robotics, ed. by M. Ang, O. Khatib 44.48 P. Abbeel, A. Coates, M. Quigley, A.Y. Ng: An ap-
(Springer, Berlin, Heidelberg 2006) plication of reinforcement learning to aerobatic
44.36 R. Clothier, R. Walker: Determination and Evalu- helicopter flight., Neural Information Processing
ation of UAV Safety Objectives, 21st International Systems Conference, Vancouver (2006)
Unmanned Air Vehicle Systems Conference, Bristol 44.49 K. Nonami: Personal communication (2003)
(2006) pp. 18.1–18.16 44.50 A. Proctor, E. Johnson: Vision-Only Approach and
44.37 J. Kuchar, J. Andrews, A. Drumm, T. Hall, V. Landing, AIAA Guidance, Navigation, and Control
Heinz, S. Thompson, J. Welch: A Safety Analysis Conference and Exhibit, San Francisco (2005) paper
Process for the Traffic Alert and Collision Avoid- 5871
ance System (TCAS) and See-and-Avoid Systems 44.51 P. Corke: An inertial and visual sensing system for
on Remotely Piloted Vehicles, AIAA 3rd Unmanned a small autonomous helicopter, J. Robot. Syst. 21,
Unlimited Technical Conference, Workshop and Ex- 43–51 (2004)
hibit, Chicago (2004) paper 6423 44.52 S. Saripalli, G.S. Sukhatme: Landing on a moving
44.38 J.L. Less, J. Chen: Flight testing the automatic air target using an autonomous helicopter, Interna-
collision avoidance system (AUTO ACAS), Adv. Astro- tional Conference on Field and Service Robotics, Mt
naut. Sci. 118, 133–144 (2004) Fuji (2003)
44.39 R.J. Carnie, R.A. Walker, P.I. Corke: Computer 44.53 S.S. Sastry, C.S. Sharp, O. Shakernia: A vision system
Vision Based Collision Avoidance for UAVs, 11th for landing an unmanned aerial vehicle, IEEE Inter-
Australian International Aerospace Congress, Mel- national Conference on Robotics and Automation
bourne (2005) (Seoul 2001) pp. 1720–1727
44.40 R. Carnie, R. Walker, P. Corke: Image processing 44.54 G. Tournier, M. Valenti, J. How, E. Feron: Estimation
algorithms for UAV sense and avoid, IEEE Inter- and Control of a quadrotor vehicle using monoc-
national Conference on Robotics and Automation ular vision and moiré patterns, AIAA Guidance,
Orlando, Melbourne (2006) pp. 2848–2853 Navigation and Control, Keystone (2006) paper 6711
44.41 J. McCalmont, J. Utt, M. Deschenes: Detect and 44.55 T. Merz, S. Duranti, G. Conte: Autonomous land-
avoid technology demonstration, 1st Unmanned ing of an unmanned helicopter based on vision
Aerospace Vehicles, Systems, Technologies and Op- and inertial sensing. In: Experimental Robotics, ed.

1 Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008
 Aerial Robotics References 1029

by M. Ang, O. Khatib (Springer, Berlin, Heidelberg 44.65 E. Frazzoli, M. Dahleh, E. Feron: Maneuver-based
2006) motion planning for nonlinear systems with sym-
44.56 C. Theodore, S. Sheldon, D. Rowley, T. McLain, W. metries, IEEE Trans. Robot. 21(6), 1077–1091 (2005)
Dai, M. Takahashi: Full mission simulation of a ro- 44.66 R. Murray, Z. Li, S. Sastry: A Mathematical Intro-
torcraft unmanned aerial vehicle for landing in duction to Robotic Manipulation (CRC, Boca Raton
a non-cooperative environment, 61st Annual Fo- 1993)
rum of the American Helicopter Society, Grapevine 44.67 A. Richards, J. How, T. Schouwenaars, E. Feron:
(2005) Spacecraft trajectory planning with collision and
44.57 C. Theodore, D. Rowley, D. Hubbard, A. Ansar, L. plume avoidance, AIAA J. Guid. Control Dyn. 25(4),
Matthies, S. Goldberg, M. Whalley: Flight trials 755–764 (2002)
of a rotorcraft unmanned aerial vehicle landing 44.68 P.R. Chandler, M. Pachter, S. Rasmussen: UAV co-
autonomously at unprepared sites, American Heli- operative control, American Control Conference,
copter Society 62nd Annual Forum, Phoenix (2006) Arlington (2001) pp. 50–55
44.58 C. Theodore, M. Tischler: Precision autonomous 44.69 M. Alighanbari, Y. Kuwata, J.P. How: Coordination
landing adaptive control experiment (PALACE), 25th and Control of Multiple UAVs with Timing Con-

Part F 44
Army Science Conference, Orlando (2006) straints and Loitering, American Control Conference
44.59 D. Topping: When Giants Roamed the Sky: Karl (2003) pp. 5311–5316
Arnstein and the Rise of Airships from Zeppelin to 44.70 T. Lemaire, R. Alami, S. Lacroix: A distributed tasks
Goodyear (Univ. of Akron Press, Akron 2001) allocation scheme in multi-UAV context, IEEE Inter-
44.60 L. Pollini, G. Campa, F. Giulietti, M. Inno- national Conference on Robotics and Automation,
centi: Virtual Simulation Set-Up for UAVs Aerial New Orleans (2004) pp. 3622–3627
Refuelling, AIAA Modeling and Simulation Tech- 44.71 Y. Jin, A.A. Minai, M.M. Polycarpou: Cooperative
nologies Conference and Exhibit, Austin (2003) real-time search and task allocation in UAV teams,
paper 5682 IEEE Conference on Decision and Control, Maui
44.61 J. Valasek, J. Kimmett, D. Hughes, K. Gunnam, (2003) pp. 7–12
J. Junkins: Vision based sensor and navigation 44.72 A. Jadbabaie, J. Lin, A.S. Morse: Coordination of
system for autonomous aerial refueling, AIAA’s groups of mobile autonomous agents using nearest
1st Technical Conference and Workshop on Un- neighbor rules, IEEE Trans. Automat. Control 1(4),
manned Aerospace Vehicles, Portsmouth (2002) 221–231 (2003)
paper 3441 44.73 L. Pallottino, A. Bicchi, E. Frazzoli: Probabilistic
44.62 J. Hansen, J. Murray, N. Campos: The NASA Dryden verification of decentralized multi-agent control
AAR Project: A Flight Test Approach to an Aerial Re- strategies: a Case Study in Conflict Avoidance,
fueling System, AIAA Atmospheric Flight Mechanics American Control Conference, New-York (2007) pp.
Conference and Exhibit, Providence (2004) paper 170–175
4939 44.74 L. Pallottino, A. Bicchi: On optimal cooperative
44.63 J.B. Saunders, B. Call, A. Curtis, R.W. Beard, T.W. conflict resolution for air traffic management sys-
McLain: Static and dynamic obstacle avoidance in tems, IEEE Trans. Intell. Transport. Syst. 1(4), 221–231
miniature air vehicles, AIAA Infotechs., Aerospace, (2000)
Arlington (2005) paper 6950 44.75 C.J. Tomlin, I. Mitchell, A. Bayen, M. Oishi: Compu-
44.64 S. Scherer, S. Singh, L. Chamberlain, S. Saripalli: tational techniques for the verification of hybrid
Flying fast and low among obstacles, International systems, Proc. IEEE 91(7), 986–1001 (2003)
Conference on Robotics and Automation, Roma 44.76 J.R. Wilson: UAV Worldwide Roundup 2007, Aerosp.
(2007) pp. 2023–2029 Am. 5, 30–37 (2007)

Springer Handbook of Robotics

Siciliano, Khatib (Eds.) · ©Springer 2008 1

View publication stats