Springer Handbook of Robotics

2154 Pages·2016·107.085 MB·english

by  Bruno Siciliano, Oussama Khatib (Eds.)

Springer
Handbook
of Robotics
2nd Edition
With1375Figuresand109Tables
Bruno Siciliano, Oussama Khatib (Eds.)
K
Editors
BrunoSiciliano
UniversityofNaplesFedericoII
DepartmentofElectricalEngineeringandInformationTechnology
Naples,Italy
[email protected]
OussamaKhatib
StanfordUniversity
DepartmentofComputerScience
ArtificialIntelligenceLaboratory
Stanford,USA
[email protected]
ISBN:978-3-319-32550-7 e-ISBN:978-3-319-32552-1
DOI10.1007/978-3-319-32552-1
LibraryofCongressControlNumber: 2016937424
©Springer-VerlagBerlinHeidelberg2016
ThisSpringerimprintispublishedbySpringerNature
TheregisteredcompanyisSpringer-VerlagGmbHBerlinHeidelberg
Foreword
Myfirstintroductiontoroboticscameviaaphonecall or tethered to a remote so-called
in 1964. The caller was Fred Terman, the author of minicomputer, or even a mainframe
the world-famous Radio Engineer’s Handbook, who computer.
was at the time Provost of Stanford University. Dr. Initially, some in the computer
Terman informed me that a computer science profes- sciencecommunityfeltthatcomput-
sor, John McCarthy, had just been awarded a large ers were powerful enough to con-
research grant, part of which required the develop- trolanymechanicaldeviceandmake
ment of computer-controlled manipulators. Someone it perform satisfactorily. Wequickly
had suggested to Terman that it would be prudent if learned that this was not to be
the mathematically oriented McCarthy had some con- the case. We started on a twofold
tact with mechanical designers. Since I was the only track. One was to develop particu-
oneontheStanfordfacultywhosespecialtywasmecha- lar devices for SAIL, so that hard-
nismdesign,Termandecidedtophoneme,eventhough ware demonstrations and proof-of- BernardRoth
we had never met and I was a youngassistantprofes- concept systems were available for Professorof
sor fresh out of graduate school with only 2 years at the fledgling robotics community to MechanicalEngineering
StanfordUniversity
Stanford. experiment with. The other track,
Dr.Terman’s phonecallled metoacloseassocia- whichwas moreorlessmoonlightedfromtheworkat
tionwithJohnMcCarthyandtheStanfordArtificialIn- SAIL,wasthedevelopmentofabasicmechanicalsci-
telligenceLaboratory(SAIL)thathefounded.Robotics enceofrobotics.Ihadastrongfeelingthatameaningful
becameoneofthepillarsofmyentireacademiccareer, sciencecouldbedeveloped,andthatitwouldbebestto
and I have maintained my interest in teaching and re- thinkin termsofgeneralconcepts ratherthan concen-
searchingthesubjectthroughtothepresentday. trateexclusivelyonparticulardevices.
The modern history of robotic manipulation dates Fortuitously, it turned out that the two tracks sup-
fromthelate1940swhenservoedarmsweredeveloped portedeachotherverynaturallyand,mostimportantly,
in connection with master–slave manipulator systems the right students were interested in doing their re-
usedtoprotecttechnicianshandlingnuclearmaterials. search in this area. Hardware developments proved to
Developmentsinthisareahavecontinuedtothepresent bespecificexamplesofmoregeneralconcepts,andthe
day. However, in the early 1960s there was very little studentswereabletodevelopboththehardwareandthe
academic or commercial activity in robotics. The first theory.
academicactivitywasthethesisofH.A.Ernst,in1961, Originally, we purchased an arm in order to get
atMIT. Heusedaslavearmequippedwithtouchsen- started quickly. A group at Rancho Los Amigos Hos-
sors,andranitundercomputercontrol.Theideainhis pital, in Los Angeles, was selling a tongue-switch-
studywastousetheinformationfromthetouchsensors controlled motor-driven exoskeleton arm to assist pa-
toguidethearm. tients withoutmuscularcontroloftheirarms. Wepur-
ThiswasfollowedbytheSAILprojectandasimi- chasedoneofthese,andconnectedittoatime-shared
larprojectstartedbyProfessorMarvinMinskyatMIT, PDP-6 computer. The device was named Butterfin-
which were the only sizeable academic ventures into gers;it was ourfirst experimentalrobot. Several films
roboticsatthattime.Therewereafewattemptsatcom- demonstrating visual feedback control, block stacking
mercialmanipulators,primarilyinconnectionwithpart tasks,andobstacleavoidanceweremadewithButterfin-
productionintheautomotiveindustry.IntheUSAthere gersasthestarperformer.
weretwodifferentmanipulatordesignsthatwerebeing Thefirstmanipulatorthatwedesignedonourown
experimentedwithintheautoindustry;onecamefrom was knownsimplyas theHydraulicArm. As its name
American Machine and Foundry(AMF) and theother implies, it was powered by hydraulics. The idea was
fromUnimation,Inc. to build a very fast arm. We designed special rotary
There were also a few mechanical devices devel- actuators, and thearm worked well. It becametheex-
opedashand,leg,andarmprosthetics,and,abitlater, perimental platform for testing the first ever dynamic
some exoskeletal devices to enhance human perfor- analysis and time-optimal control of a robotic arm.
mance. In those days there were no microprocessors. However, its use was limited since the design speeds
So,thesedeviceswereeitherwithoutcomputercontrol, were much faster than required due to the limitations
ofthecomputational,planning,andsensingcapabilities complete system integration became the norm; to this
thatwerecommonatthattime. day, this combination represents the hallmark of most
We made an attempt to develop a truly digital highlyregardedroboticdevices.This isthebasiccon-
arm.Thisledtoasnake-likestructurenamedtheOrm cept behind mechatronic, a word conied in Japan as
(theNorwegianwordforsnake.)TheOrmhadseveral aconcatenationofthewordsmechanicsandelectronics.
stages,eachwithanarrayofinflatablepneumaticactu- Mechatronicsthatreliesoncomputationistheessence
atorsthatwereeitherfullyextendedorfullycontracted. of the technology inherent in robotics as we know it
Thebasicideawasthat,eventhoughonlyafinitenum- today.
ber of positions in the workspace could be reached, As robotics developed around the world, a large
thesewouldbesufficientifthere were alargenumber number of people started working on various aspects,
of positions. A small prototypeproof-of-concept Orm and specificsubspecialtiesdeveloped.Thefirstbig di-
wasdeveloped.Itledtotherealizationthatthistypeof vision was between people working on manipulators
armwouldnotreallyservetheSAILcommunity. and those working on vision systems. Early on, vi-
The first truly functional arm from our group was sion systems seemed to hold more promise than any
designedbyVictorScheinman,whowasagraduatestu- othermethodforgivingrobotsinformationabouttheir
dent at the time. It was the very successful Stanford environment.
Arm, of which over ten copies were made as research The idea was to have a television camera capture
toolstobeusedinvariousuniversity,government,and pictures of objects in the environment, and then use
industrial laboratories. The arm had six independently algorithmsthatallowedthecomputerimagesofthepic-
driven joints; all driven by computer-controlled ser- turestobeanalyzed,soastoinferrequiredinformation
voed, DC electric motors. One joint was telescoping aboutlocation,orientation,andotherpropertiesofob-
(prismatic)andtheotherfivewererotary(revolute). jects. The initial successes with image systems were
WhereasthegeometryofButterfingersrequiredan in problems dealing with positioning blocks, solving
iterative solution of the inverse kinematics, the geo- object manipulation problems, and reading assembly
metric configuration of the Stanford Arm was chosen drawings. It was felt that vision held potential for use
so that the inverse kinematics could be programmed in roboticsystems in connectionwith factory automa-
in any easy-to-usetime-efficient closed form. Further- tion and space exploration. This led to research into
more, the mechanical design was specifically made to software thatwouldallowvisionsystemsto recognize
be compatible with the limitations inherent in time- machineparts(particularly partiallyoccludedparts,as
sharecomputercontrol.Variousend-effectorscouldbe occurred in the so-called bin-picking problems) and
attached to act as hands. On our version, the hand ragged-shapedrocks.
was in the form of a vise-grip jaw, with two slid- After the ability to see and move objects became
ing fingers drivenby aservoed actuator(hence, a true established,thenextlogicalneedhadtodowith plan-
seventh degree of freedom). It also had a specially ningasequenceofeventstoaccomplishacomplextask.
designed six-axis wrist force sensor. Victor Schein- This led to the developmentof planning as an impor-
man went on to develop other important robots: the tantbranchinrobotics.Makingfixedplansforaknown
first was a small humanoid arm with six revolute fixed environment is relatively straightforward. How-
joints. The original design was paid for by Marvin ever, in robotics, one of the challenges is to let the
Minsky at the MIT AI Lab. Scheinman founded Vi- robot discover its environment, and to modify its ac-
carm, a small company, and produced copies of this tionswhentheenvironmentchangesunexpectedlydue
arm and theStanfordArm for other labs. Vicarm later to errors or unplanned events. Some early landmark
became the West Coast Division of Unimation, Inc., studies in this area were carried out using a vehicle
where Scheinman designed the PUMA manipulator namedShakey,which,startingin1966,wasdeveloped
under General Motors sponsorship through Unima- byCharlieRosen’sgroupattheStanfordResearchIn-
tion. Later, for a company called Automatix, Schein- stitute (now called SRI). Shakey had a TV camera,
mandevelopedthenovelRobotWorld multirobotsys- a triangulating range finder, bump sensors, and was
tem. After Scheinman left Unimation, his colleagues connectedtoDECPDP-10andPDP-15computersvia
BrianCarlisleandBruceShimanoreorganizedUnima- radioandvideolinks.
tion’s West Coast Division into Adept, Inc., which to Shakey was the first mobile robot to reason about
this day is the largest US manufacturer of assembly its actions. It used programs that gave it the ability
robots. forindependentperception,worldmodeling,andaction
Quickly,themoderntrendofcarefullydetailedme- generation.Low-levelactionroutinestookcareofsim-
chanicalandelectronicdesign,optimizedsoftware,and plemoving,turning,androuteplanning.Intermediate-
levelactionscombinedthelow-levelonesinwaysthat RIAannualtradeshow,whichis nowcalledtheInter-
accomplished more complex tasks. The highest level nationalRobotsandVisionShowandConference.
programs could make and execute plans to achieve Thefirstregularseriesofconferencesemphasizing
high-levelgoalssuppliedbyauser. research,ratherthantheindustrial,aspectsofrobotics,
Vision is very useful for navigation, locating ob- was inaugurated in 1973. It was sponsored jointly
jects, and determining their relative positions and ori- by the International Center for Mechanical Sciences
entation. However, it is usually not sufficient for as- (CISM), based in Udine, Italy, and the International
semblingpartsorworkingwithrobotswherethereare Federation for the Theory of Mechanisms and Ma-
environmentalconstrainingforces.Thisledtotheneed chines(IFToMM).(AlthoughIFToMMisstillused,its
tomeasuretheforcesandtorquesgeneratedbytheen- meaning has been changed to the International Feder-
vironment,on a robot,and to usethese measurements ation for the Promotion of Mechanism and Machine
to control the robot’s actions. For many years, force- Science.)ItwasnamedtheSymposiumonTheoryand
controlledmanipulationbecameoneofthemaintopics Practice of Robots and Manipulators (RoManSy). Its
of study at SAIL, and several other labs around the trademarkwasanemphasisonthemechanicalsciences
world.Theuseofforcecontrolinindustrialpracticehas and the active participation of researchers from East-
always lagged the research developments in this area. ernandWesternEuropeaswellasNorthAmericaand
Thisseemstobeduetothefactthat,whileahighlevel Japan.Itisstillheldbiannually.Onapersonalnote,it
offorcecontrolisveryusefulforgeneralmanipulation isatRoManSywhereIfirstmeteachoftheeditorsof
issues,specificproblemsinveryrestrictedindustrialen- this Handbook: Dr. Khatib in 1978 and Dr. Siciliano
vironments can often be handled with limited, or no, in1984.Theywerebothstudents:BrunoSicilianohad
forcecontrol. beenworkingonhisPhDforaboutoneyear,andOus-
In the 1970s, specialized areas of study such as sama Khatib had just completed his PhD research. In
walking machines, hands, automated vehicles, sensor bothcases,itwasloveatfirstsight!
integration, and design for hostile environments be- RoManSy was quickly joined by a host of other
gantodeveloprapidly.Todaytherearealargenumber newconferencesandworkshops;todaytherearealarge
of different specialties studied under the heading of numberofresearchorientedroboticsmeetingsthattake
robotics. Some of these specialties are classical engi- place through the year in many countries. Currently,
neering subject areas within which results have been the largest conference is the International Conference
developed that have been particularized to the types on Roboticsand Automation(ICRA), which regularly
of machines called robots. Examples here are kine- drawswellover1000participants.
matics, dynamics,controls, machinedesign, topology, In the beginning of the 1980s, the first real text-
and trajectory planning. Each of these subjects has bookon roboticmanipulationin theUSA was written
a long history predating the study of robotics; yet by RichardLou Paul(Richard P. Paul, RobotManipu-
each has been an area of in-depth robotics research lators: Mathematics, Programming, and Control, The
in order to develop its special character in regard to MIT Press, Cambridge, MA, 1981). It used the idea
robotic-type systems and applications. In doing this oftakingclassicalsubjectsinmechanicsandapplying
specializeddevelopment,researchershaveenrichedthe them to robotics. In additionthere were several topics
classical subjects by increasing both their contentand developeddirectlyfromhisthesisresearchatSAIL.(In
scope. the book, many examples are based on Scheinman’s
At the same time that the theory was being devel- Stanford Arm.) Paul’s book was a landmark event in
oped,therewasaparallel,althoughsomewhatseparate, the USA; it created a pattern for several influential
growth of industrial robotics. Strong commercial de- future textbooks and also encouraged the creation of
velopment occurred in Japan and Europe, and there specialized robotics courses at a host of colleges and
was also continued growth in the USA. Industrial as- universities.
sociations were formed (the Japan Robot Association Ataboutthissametime,newjournalswerecreated
wasformedinMarch1971,andtheRoboticIndustries to deal primarily with research papers in the areas re-
Association (RIA) was founded in 1974 in the USA) latedtorobotics.TheInternationalJournalofRobotics
and trade shows, together with application-oriented Researchwasfoundedinthespringof1982,andthree
technicalsessions,were introducedandheldonareg- years later theIEEE JournalofRoboticsandAutoma-
ular basis. The most important were the International tion (now the IEEE Transactions on Robotics) was
SymposiumonIndustrialRobots,theConferenceonIn- founded.
dustrialRobotTechnology(nowcalledtheInternational As microprocessors became ubiquitous, the ques-
Conference on Industrial Robot Technology), and the tion of what is or is not a robot came more into
play. This issue has, in my mind, never been success- of possibilities, including those machines that have
fully resolved. I do not think a definition will ever be sensory environmental feedback and decision-making
universally agreed upon. There are of course the sci- capabilities.Inactualpractice,indevicesconsideredto
ence fiction creatures-from-outer-space varieties, and berobotic,theamountofsensoryanddecisionmaking
the robots of the theater, literature, and the movies. capabilitymayvaryfromagreatdealtonone.
Thereareexamplesofimaginaryrobot-likebeingsthat Inrecentdecadesthestudyofroboticshasexpanded
predate the industrial revolution, but how about more fromadisciplinecenteredonthestudyofmechatronic
down-to-Earthrobots?Inmyviewthedefinitionises- devices to a much broader interdisciplinary subject.
sentiallyamovingtargetthatchangesitscharacterwith An example of this is the area called human-centered
technologicalprogress.Forexample,when itwas first robotics.Hereonedealswith theinteractionsbetween
developed,aship’sgyroauto-compasswasconsidered humansandintelligentmachines.Thisisagrowingarea
arobot.Today,itisnotgenerallyincludedwhenwelist wherethestudyoftheinteractionsbetweenrobotsand
therobotsinourworld.Ithasbeendemotedandisnow humans has enlisted expertise from outside the clas-
consideredanautomaticcontroldevice. sical robotics domain. Concepts such as emotions in
Formany,theideaofa robotincludes theconcept bothrobotsandpeoplearebeingstudied,andolderar-
of multifunctionality, meaning the device is designed eas such as human physiology and biology are being
and built with the ability to be easily adapted or re- incorporated into the mainstream ofrobotics research.
programmed to do different tasks. In theory this idea Theseactivitiesenrichthefieldofrobotics,astheyin-
isvalid,butinpracticeitturnsoutthatmostroboticde- troduce new engineering and science dimensions into
vices are multifunctionalin only a very limited arena. theresearchdiscourse.
Inindustryitwasquicklydiscoveredthataspecialized Originally,thenascentroboticscommunitywasfo-
machine,ingeneral,performsmuchbetterthanagen- cused on getting things to work. Many early devices
eral purpose machine. Furthermore, when the volume were remarkable in that they worked at all, and little
ofproductionishighenough,aspecializedmachinecan notice was taken of their limited performance. Today,
costlesstomanufacturethanageneralizedone.So,spe- we have sophisticated, reliable devices as part of the
cialized robots were developed for painting, riveting, modern array of robotic systems. This progress is the
quasiplanarpartsassembly,pressloading,circuitboard result of the work of thousands of people throughout
stuffing,etc.Insomecasesrobotsareusedinsuchspe- the world. A lot of this work took place in universi-
cializedwaysthatitbecomesdifficulttodrawtheline ties,governmentresearchlaboratories,andcompanies.
between a so-called robot and an adjustable piece of It is a tribute to the worldwide engineeringand scien-
fixed automation. Much of this practical unfolding is tificcommunitythatithasbeenabletocreatethevast
contrary tothedream ofthepioneersin robotics,who amountofinformationthatiscontainedinthe64chap-
hadhopedforthedevelopmentofgeneralpurposema- tersofthisHandbook.Clearlytheseresultsdidnotarise
chinesthatwoulddoeverything,andhencesellingreat byanycentralplanningorbyanoverallorderlyscheme.
enoughvolumetoberelativelyinexpensive. Sotheeditorsofthishandbookwerefacedwiththedif-
Myviewisthatthenotionofarobothastodowith ficulttaskoforganizingthematerialintoalogicaland
which activities are, at a given time, associated with coherentwhole.
peopleandwhichareassociatedwithmachines.Ifama- The editors have accomplished this by organiz-
chine suddenlybecomes able to do what we normally ing the contributions into a three-layer structure. The
associatewithpeople,themachinecanbeupgradedin first layer deals with the foundations of the subject.
classification and classified as a robot. After a while, This layer consists of a single part of nine chapters
peoplegetusedtotheactivitybeingdonebymachines, in which the authors lay out the root subjects: kine-
andthedevicesgetdowngradedfromrobottomachine. matics, dynamics, control, mechanisms, architecture,
Machines that do not have fixed bases, and those that programming,reasoning,andsensing.Thesearetheba-
havearm-orleg-likeappendageshavetheadvantageof sictechnologicalbuildingblocksforroboticsstudyand
beingmorelikelycalledrobots,butitishardtothinkof development.
aconsistentsetofcriteriathatfitsallthecurrentnaming The second layer has four parts. The first of these
conventions. deals with robot structures; these are the arms, legs,
Inactualityanymachines,includingfamiliarhouse- hands,andotherpartsthatmostrobotsaremadeupof.
holdappliances,whichhavemicroprocessorsdirecting At first blush, the hardware of legs, arms, and hands
theiractionscanbeconsideredasrobots.Inadditionto maylookquitedifferentfromeachother,yettheyshare
vacuum cleaners, there are washing machines, refrig- a common set of attributes that allows them to all be
erators,anddishwashersthatcouldbeeasilymarketed treatedwiththesame,orcloselyrelated,aspectsofthe
as robotic devices. There are of course a wide range fundamentalsdescribedinthefirstlayer.
The second part of this layer deals with sensing extraordinaryaccomplishmentswouldprobablynotex-
andperception,which are basic abilities any truly au- istwithoutthepreviousdevelopmentsintroducedinthe
tonomous robotic system must have. As was pointed firsttwolayersofthisHandbook.
outearlier, in practice, manyso-called roboticdevices It is this intimate connection between theory and
have little of these abilities, but clearly the more ad- practice that has nurtured the growth of robotics and
vancedrobotscannotexistwithoutthem,andthetrend becomeahallmarkofmodernrobotics.Thesetwocom-
isverymuchtowardincorporatingsuchcapabilitiesinto plementaryaspectshavebeenasourceofgreatpersonal
robotic devices. The third part of this layer treats the satisfaction to those of us who have had the opportu-
subjectareasassociatedwiththetechnologyofmanip- nitytobothresearch anddeveloproboticdevices.The
ulationandtheinterfacingofdevices.Thefourthpartof contentsof this Handbookadmirablyreflect thiscom-
thislayerismadeupofeightchaptersthattreatmobile plementary aspect of the subject, and present a very
robotsandvariousformsofdistributedrobotics. useful bringing together of the vast accomplishments
Thethirdlayerconsistsoftwoseparateparts(ato- which have taken place in the last 50years. Certainly,
talof22chapters)thatdealwithadvancedapplications thecontentsofthisHandbookwillserveasavaluable
at the forefront of today’s research and development. toolandguidetothosewhowillproducetheevenmore
There are two parts to this layer; one deals with field capableanddiversenextgenerationsofroboticdevices.
and service robots, and the other deals with human- The editors and authors have my congratulations and
centeredandlifelikerobots.Totheuninitiatedobserver, admiration.
thesechaptersarewhatadvancedroboticsisallabout.
However, it is important to realize that many of these Stanford,August2007 BernardRoth
Foreword
To open this Handbook and unfold the richness of its algorithmic geometry as well as of
64chapters,wehereattemptabriefpersonaloverview astochasticframeworkapproachap-
tosketchtheevolutionofroboticsinitsmanyaspects, plied both to environmental mod-
concepts,trends,andcentralissues. eling and robot localization prob-
The modern story of Robotics began about half lems(SLAM,simultaneouslocaliza-
a century ago with developments in two different di- tionandmodeling),andfurtherfrom
rections. the development of decisional pro-
First, let us acknowledge the domain of me- ceduresviaBayesianestimationand
chanical arms, ranging from teleoperated tasks on decisionapproaches.
radiation-contaminated products to industrial arms, Forthelastdecadeofthemillen-
with the landmark machine UNIMATE – standing for nium,roboticslargelydealtwiththe
uni(versal)mate. The industrial development of prod- intelligent robot paradigm, blend-
ucts, mostly around the six-degree-of-freedom serial ing together robots and machine- GeorgesGiralt
links paradigm and active research and development, intelligence generic research within Emeritus Research
associatingmechanicalengineeringto thecontrolspe- themes covering advanced sensing Director
LAAS-CNRSToulouse
cialism,wasthemaindrivingforcehere.Ofparticular and perception, task reasoning and
(deceased)
note nowadays is the successfully pursued effort to planning,operationaland decisional
design novel application-optimized structures, using autonomy, functional integration architectures, intelli-
powerful sophisticated mathematical tools. In a simi- gent human–machine interfaces, safety, and depend-
lar way, an important issue concerns the design and ability.
theactualbuildingofarmsandhandsinthecontextof The second branch, for years referred to as non-
human-friendlyrobotsfortomorrow’scognitiverobot. manufacturing robotics, concerns a wide spectrum of
Second, and less well recognized, we should ac- research-driven real-world cases pertaining to field,
knowledgethestream of work concernedwith themes service, assistive, and, later, personal robotics. Here,
inartificialintelligence.Alandmarkprojectinthisarea machineintelligenceis,initsvariousthemes,thecen-
wasthemobilerobotShakeydevelopedatStanfordIn- tralresearchdirection,enablingtherobottoact:
ternational. This work, which aimed to bring together
1. Asahumansurrogate,inparticularforintervention
computer science, artificial intelligence, and applied
tasksinremoteand/orhostileenvironments
mathematicstodevelopintelligentmachines,remained
2. In close interaction with humans and operating
asecondaryareaforquitesometime.Duringthe1980s,
in human environments in all applications encom-
buildingstrengthfrommanystudycasesencompassing
passedbyhuman-friendlyrobotics,alsoreferredto
aspectacularspectrumrangingfromroversforextreme
ashuman-centeredrobotics
environments (planet exploration, Antarctica, etc.), to
3. In tight synergy with the user, expanding from
servicerobots(hospitals,museumguides,etc.),abroad
mechanical exoskeleton assistance, surgery, health
research domainarosein whichmachinescouldclaim
care,andrehabilitationintohumanaugmentation.
thestatusofintelligentrobots.
Hence robotics researches could bring together Consequently, at the turn of the millennium,
thesetwodifferentbranches,withintelligentrobotscat- robotics appears as a broad spectrum of research
egorized in a solely computational way as bounded themes both supporting market products for well-en-
rationality machines, expanding on the 1980s third- gineered industrial workplaces, and a large number
generationrobotdefinition: ofdomain-orientedapplicationcasesoperatinginhaz-
ardousand/orharshenvironments(underwaterrobotics,
(robot)... operatinginthethree-dimensionalworld
rough-terrainrovers,health/rehabilitationcarerobotics,
asamachineendowedwiththecapacitytointerpret
etc.) where robots exhibitmeaningfullevels of shared
andtoreasonaboutataskandaboutitsexecution,
autonomy.
byintelligentlyrelatingperceptiontoaction.
The evolution levels for robotics stress the role of
The field of autonomous robots, a widely rec- theoretical aspects, moving from application domains
ognized test-bed, has recently benefited from salient tothetechnicalandscientificarea.Theorganizationof
contributions in robot planning using the results of thisHandbookillustratesverywellthesedifferentlev-
els.Furthermore,itrightlyconsiders,besidesabodyof ligent interfaces, perception (scene analysis, category
softwaresystems,front-linemattersonphysicalappear- identification),open-endedlearning(understandingthe
ance and novel appendages, including legs, arms, and universeof action), skills acquisition, extensiverobot-
hands design in the context of human-friendly robots world data processing, decisional autonomy, and de-
fortomorrow’scognitiverobot. pendability(safety,reliability,communication,andop-
Forefront robotics in the first decade of the cur- eratingrobustness).
rentmillenniumis makingoutstandingprogress,com- There is an obvious synergistic effort between the
poundingthestrengthoftwogeneraldirections: two aforementioned approaches, in spite of theneces-
saryframeworktimedifferences.Thescientificlinknot
(cid:2) Short/mid-termapplication-orientedstudycases
onlybringstogethertheproblemsandobtainedresults
(cid:2) Mid/long-termgenericsituatedresearch.
butalsocreatesasynergisticexchangebetweenthetwo
For completeness, we should mention the large sidesandthebenefitsoftechnologicalprogress.
numberof peripheral, robotics-inspired subjects, quite Indeed, the corresponding research trends and ap-
often concerning entertainment, advertising, and so- plication developments are supported by an explosive
phisticatedtoys. evolution ofenabling technologies:computerprocess-
Thesalientfieldofhuman-friendlyroboticsencom- ing power, telecommunications, networking, sensing
passesseveralfront-lineapplicationdomainswherethe devices,knowledgeretrieval,newmaterials,micro-and
robots operate in a human environment and in close nanotechnologies.
interactionwithhumans(entertainmentandeducation, Today,lookingtothemid-andlong-termfuture,we
public-orientedservices, assistiveandpersonalrobots, arefacedwithverypositiveissuesandperspectivesbut
etc.), which introduces the critical issue of human– alsohavingtorespondtocriticalcommentsandloom-
robotinteraction. ing dangers for machines that are in physical contact
Rightatthecoreofthefield,emergestheforefront with the user and may also be capable of unwanted,
topicofpersonalrobotsforwhichthreegeneralcharac- unsafebehavior.Therefore, thereisaclear needtoin-
teristicsshouldbeemphasized: cludeattheresearchlevelsafetyissuesandthetopicof
multifaceddependabilityandthecorrespondingsystem
1. Theymaybeoperatedbyanonprofessionaluser;
constraints.
2. Theymaybedesignedtosharehigh-leveldecision
The Handbook of Robotics is an ambitious and
makingwiththehumanuser;
timelyendeavor.Itsummarizesalargenumberofprob-
3. Theymayincludealinktoenvironmentdevicesand
lems, questions,and facets considered by 164 authors
machine appendages, remote systems, and opera-
in 64 chapters. As such it not only provides an effi-
tors; the shared decisional autonomy concept (co-
cient display of basic topics and results obtained by
autonomy)implied hereunfoldsintoa largesetof
researchesaroundtheworld,butfurthermoregivesac-
cutting-edgeresearchissuesandethicalproblems.
cess to this variety of viewpoints and approaches to
The concept of the personal robot, expanding to everyone.Thisisindeedanimportanttoolforprogress
robotassistantanduniversalcompanion,isatrulygreat but,muchmore,isthecentralfactorthatwillestablish
challengeforroboticsasascientificandtechnicalfield, thetwofirstdecadesofthismillenniumasthedawnof
offering the mid/long-term perspective of achieving robotics, lifted to a scientific discipline at the core of
a paramount societal and economical impact. This in- machineintelligence.
troduces,andquestions,front-linetopicsencompassing
cognitive aspects: user-tunable human–machine intel- Toulouse,December2007 GeorgesGiralt
Foreword
The field of robotics was born in the middle of the robots to manipulate objects. Stud-
last century when emerging computers were altering ies on modeling, planning, knowl-
every field of science and engineering. Having gone edge, reasoning, and memorization
throughfastyetsteadygrowthviaaprocessionofstages expanded their intelligent proper-
frominfancy,childhood,andadolescencetoadulthood, ties. Robotics became defined as
roboticsis nowmatureandisexpectedtoenhancethe the study of intelligent connection
qualityofpeople’slivesinsocietyinthefuture. of sensing to actuation. This defini-
Initsinfancy,thecoreofroboticsconsistedofpat- tion covered all aspects of robotics:
ternrecognition,automaticcontrol,andartificialintel- three scientific cores and one syn-
ligence. Taking on these new challenge, scientists and thetic approach to integrate them.
engineers in these fields gathered to investigate novel Indeed, system integration became
robotic sensors and actuators, planning and program- a key aspect of robotic engineering
ming algorithms, and architectures to connect these as it allows the creation of lifelike HirochikaInoue
componentsintelligently.Insodoing,theycreatedarti- machines. The fun of creating such ProfessorEmeritus
factsthatcouldinteractwithhumansintherealworld. robotsattractedmanystudentstothe TheUniversityofTokyo
An integration of these early robotics studies yielded roboticsfield.
hand–eyesystems,thetest-bedofartificialintelligence In advancing robotics further, scientific interest
research. was directed at understanding humans. Comparative
The playground for childhood robotics was the studies of humans and robots led to new approaches
factory floor. Industrial robots were invented and in- in scientific modeling of human functions. Cognitive
troducedintothefactoryforautomatingspraying,spot robotics,lifelikebehavior,biologicallyinspiredrobots,
welding,grinding,materialshandling,andpartsassem- and a psychophysiological approach to robotic ma-
bly. Machines with sensors and memories made the chinesculminatedinexpandingthehorizonsofrobotic
factory floor smarter, and its operations more flexible, potential. Generally speaking, an immature field is
reliable,andprecise.Suchroboticautomationfreedhu- sparseinscientificunderstanding.Roboticsinthe1980s
mans from heavy and tedious labor. The automobile, and 1990s was in such a youthful stage, attracting
electricappliance,andsemiconductorindustriesrapidly agreatmanyinquisitiveresearcherstothisnewfrontier.
retooledtheirmanufacturinglinesintorobot-integrated Theircontinuousexplorationsintonewrealmsformthe
systems.Inthelate1970s,thewordmechatronics,orig- richscientificcontentsofthiscomprehensivevolume.
inally coined by the Japanese, defined a new concept Further challenges, along with expertise acquired
ofmachinery,oneinwhichelectronicswasfusedwith on the cutting edge of robotics, opened the way to
mechanical systems, making a wide range of indus- real-world applicationsformaturerobotics.Theearly-
trialproductssimpler,morefunctional,programmable, stageplaygroundgavewaytoaworkshopforindustrial
and intelligent. Robotics and mechatronics exerted an robotics. Medical robotics, robot surgery, and in vivo
evolutionary impact on the design and operation of imaging save patients from pain while providing doc-
manufacturing processes as well as on manufactured torswithpowerfultoolsforconductingoperations.New
products. robots in such areas as rehabilitation, health care, and
As robotics entered its adolescence, researchers welfareareexpectedtoimprovequalityoflifeinanag-
were ambitious to explore new horizons. Kinematics, ingsociety.Itisthedestinyofrobotstogoeverywhere,
dynamics,andcontrolsystemtheorywere refined and in the air, under water, and into space. They are ex-
appliedtorealcomplexrobotmechanisms.Toplanand pectedtoworkhandinhandwithhumansinsuchareas
carry out real tasks, robots had to be made cognizant as agriculture, forestry, mining,construction,and haz-
of their surroundings. Vision, the primary channel for ardousenvironmentsandrescueoperations,andtofind
externalsensing,wasexploitedasthemostgeneral,ef- utilitybothindomesticworkandinprovidingservices
fective, and efficient means for robots to understand in shops, stores, restaurants, and hospitals. In a myr-
theirexternalsituation.Advancedalgorithmsandpow- iad of ways, robotic devices are expected to support
erful devices were developed to improve the speed our daily lives. At this point, however, robot appli-
and robustness of robot vision systems. Tactile and cations are largely limited to structured environments,
forcesensingsystemsalsoneededtobedevelopedfor wheretheyareseparatedfromhumansforsafetysake.