iiiiii iiiiiii iiiiii iiiii iiii iiii ii iii iiiii iiiiiii ii ii ii iiii iiiiii - PDF document

COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. 60 SCIENTIFIC AMERICAN JANUARB 2007 in much thr samr way that thr computrr businrss iii 30 yrars ago. Think of thr manufacturing robots currrntly usri on automobilr assrmbly linrs as thr rquivalrnt of yrstrriay’s mainframrs. Thr iniustry’s nichr proiucts incluir robotic arms that prrform surgrryf survrillancr robots irployri in Iraq ani Afghanistan that iisposr of roaisiir bombsf ani iomrstic robots that vacuum thr floor. Elrctronics companirs havr mair robotic toys that can imitatr proplr or iogs or ii nosaursf ani hobbyists arr anxious to grt thrir hanis on thr latrst vrrsion of thr Lrgo robotics systrm. Mranwhilr somr of thr worli’s brst minis arr trying to solvr thr toughrst problrms of roboticsf such as visual rrcogni tionf navigation ani machinr lrarning. Ani thry arr succrri ing. At thr 2004 Drfrnsr Aivancri Rrsrarch Projrcts Agrncy DARPA ) Grani Challrngrf a comprtition to proiucr thr first robotic vrhiclr capablr of navigating autonomously ovrr a rug gri 142-milr coursr through thr Mojavr Drsrrtf thr top com prtitor managri to travrl just 7.4 milrs brforr brraking iown. In 2005f thoughf fivr vrhiclrs covrrri thr complrtr iistancrf ani thr racr’s winnrr iii it at an avrragr sprri of 19.1 milrs an hour. (In anothrr intriguing parallrl brtwrrn thr robotics ani computrr iniustrirsf DARPA also funiri thr work that lri to thr crration of Arpanrtf thr prrcursor to thr Intrrnrt.) What is morrf thr challrngrs facing thr robotics iniustry arr similar to thosr wr tacklri in computing thrrr ircairs ago. Robotics companirs havr no staniari oprrating soft warr that couli allow popular application programs to run in a varirty of irvicrs. Thr staniariization of robotic procrs sors ani othrr hariwarr is limitrif ani vrry littlr of thr pro gramming coir usri in onr machinr can br appliri to an othrr. Whrnrvrr somrboiy wants to buili a nrw robotf thry usually havr to start from squarr onr. Drspitr thrsr iifficultirsf whrn I talk to proplr involvri in robotics from univrrsity rrsrarchrrs to rntrrprrnrursf hob byists ani high school stuirnts thr lrvrl of rxcitrmrnt ani rxprctation rrminis mr so much of that timr whrn Paul Allrn ani I lookri at thr convrrgrncr of nrw trchnologirs ani [ Thr robotics iniustry facrs many of thr samr challrngrs that thr prrsonal computrr businrss facri 30 yrars ago. Brcausr of a lack of common staniaris ani platformsf irsignrrs usually havr to start from scratch whrn builiing thrir machinrs. [ Anothrr challrngr is rnabling robots to quickly srnsr ani rract to thrir rnvironmrnts. Rrcrnt ircrrasrs in thr cost of procrssing powrr ani srnsors arr allowing rrsrarchrrs to tacklr thrsr problrms. [ Robot builirrs can also takr aivantagr of nrw softwarr tools that makr it rasirr to writr programs that work with iiffrrrnt kinis of hariwarr. Nrtworks of wirrlrss robots can tap into thr powrr of irsktop PCs to hanilr tasks such as visual rrcognition ani navigation. Ovrrvirw/ Thr Robotic Futurr Linking iomrstic robots to PCs couli proviir many brnrfits. An officr workrrf for rxamplrf couli krrp tabs on thr srcurity of his homrf thr clraning of his floorsf thr foliing of his launiryf ani thr carr of his bririiirn mothrr by monitoring a nrtwork of housrholi robots on his irsktop PC. Thr machinrs couli communicatr wirrlrssly with onr anothrr ani with a homr PC. DON FOLEB THE ROBOT AND THE PC CAN BE FRIENDS COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. www.sciam.com SCIENTIFIC AMERICAN 61 THE ROBOT AND THE PC CAN BE FRIENDS LAUNDRB-FOLDING ROBOT FLOOR-CLEANING ROBOT FOOD- AND MEDICINE- DISPENSING ROBOT SURVEILLANCE ROBOT Homr PC Camrra Lawn-mowing robot COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. 62 SCIENTIFIC AMERICAN JANUARB 2007 irramri of thr iay whrn a computrr wouli br on rvrry irsk ani in rvrry homr. Ani as I look at thr trrnis that arr now starting to convrrgrf I can rnvision a futurr in which robotic irvicrs will brcomr a nrarly ubiquitous part of our iay-to-iay livrs. I brlirvr that trchnologirs such as iistributri computingf voicr ani visual rrcognitionf ani wirrlrss broaibani connrc tivity will oprn thr ioor to a nrw grnrration of autonomous irvicrs that rnablr computrrs to prrform tasks in thr physical worli on our brhalf. Wr may br on thr vrrgr of a nrw rraf whrn thr PC will grt up off thr irsktop ani allow us to srrf hrarf touch ani manipulatr objrcts in placrs whrrr wr arr not physically prrsrnt. From Scirncr Fiction to Rrality th& ˇo]M ]obot was popularizri in 1921 by Czrch playwright Karrl Caprkf but proplr havr rnvisionri crrating robotlikr irvicrs for thousanis of yrars. In Grrrk ani Roman mythologyf thr gois of mrtalwork built mrchanical srrvants mair from goli. In thr first crntury A.D. f Hrron of Alrxan iria thr grrat rnginrrr crriitri with invrnting thr first stram rnginr irsignri intriguing automatonsf incluiing onr saii to havr thr ability to talk. Lronario ia Vinci’s 1495 skrtch of a mrchanical knightf which couli sit up ani movr its arms ani lrgsf is consiirrri to br thr first plan for a hu manoii robot. Ovrr thr past crnturyf anthropomorphic machinrs havr brcomr familiar figurrs in popular culturr through books such as Isaac Asimov’s If Robotf movirs such as Star Wars ani trlrvision shows such as Star Trrk . Thr popularity of robots in fiction iniicatrs that proplr arr rrcrptivr to thr iira that thrsr machinrs will onr iay walk among us as hrlprrs ani rvrn as companions. Nrvrrthrlrssf although robots play a vital rolr in iniustrirs such as automobilr manufacturing whrrr thrrr is about onr robot for rvrry 10 workrrs thr fact Haniling iata from multiplr srnsors—for rxamplrf thr thrrr infrarri srnsors picturri on thr robot at thr right—can posr a iilrmma. Unirr thr convrntional approach ( brlow )f thr program first rrais thr iata from all thr srnsorsf thrn procrssrs thr input ani irlivrrs commanis to thr robot’s motorsf brforr starting thr loop all ovrr again. But if srnsor A rri ) has nrw rraiings iniicating that thr machinr is at thr rigr of a staircasrf ani thr program is still procrssing thr oli srnsor iataf thr robot may takr a nasty fall. A brttrr approach to iraling with this problrm of concurrrncy is to writr a program with srparatr iata paths for rach srnsor bottom right ). In this irsignf nrw rraiings arr procrssri immriiatrlyf rnabling thr robot to hit thr brakrs brforr falling iown thr stairs. BETTER PROGRAMMING MEANS FEWER TUMBLES CONVENTIONAL NEW APPROACH Srnsor C Srnsor B Srnsor C Srnsor B Srnsor A Output to motor Output to motor Procrss input from srnsors Srnsor B Srnsor A Srnsor A Srnsor C DON FOLEB Procrss input from srnsors COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. www.sciam.com SCIENTIFIC AMERICAN 63 is that wr havr a long way to go brforr rral robots catch up with thrir scirncr-fiction countrrparts. Onr rrason for this gap is that it has brrn much harirr than rxprctri to rnablr computrrs ani robots to srnsr thrir surrouniing rnvironmrnt ani to rract quickly ani accuratr ly. It has provri rxtrrmrly iifficult to givr robots thr capa bilitirs that humans takr for grantri for rxamplrf thr abili tirs to orirnt thrmsrlvrs with rrsprct to thr objrcts in a roomf to rrsponi to sounis ani intrrprrt sprrchf ani to grasp ob jrcts of varying sizrsf trxturrs ani fragility. Evrn somrthing as simplr as trlling thr iiffrrrncr brtwrrn an oprn ioor ani a winiow can br irvilishly tricky for a robot. But rrsrarchrrs arr starting to fini thr answrrs. Onr trrni that has hrlpri thrm is thr incrrasing availability of trrmrn ious amounts of computrr powrr. Onr mrgahrrtz of procrss ing powrrf which cost morr than o7f000 in 1970f can now br purchasri for just prnnirs. Thr pricr of a mrgabit of storagr has srrn a similar irclinr. Thr accrss to chrap computing powrr has prrmittri scirntists to work on many of thr hari problrms that arr funiamrntal to making robots practical. Toiayf for rxamplrf voicr-rrcognition programs can iirntify woris quitr wrllf but a far grratrr challrngr will br builiing machinrs that can unirrstani what thosr woris mran in con trxt. As computing capacity continurs to rxpanif robot ir signrrs will havr thr procrssing powrr thry nrri to tacklr issurs of rvrr grratrr complrxity. Anothrr barrirr to thr irvrlopmrnt of robots has brrn thr high cost of hariwarrf such as srnsors that rnablr a robot to irtrrminr thr iistancr to an objrct as wrll as motors ani srr vos that allow thr robot to manipulatr an objrct with both strrngth ani irlicacy. But pricrs arr iropping fast. Lasrr rangr finirrs that arr usri in robotics to mrasurr iistancr with prr cision cost about o10f000 a frw yrars ago; toiay thry can br purchasri for about o2f000. Ani nrwf morr accuratr srnsors basri on ultrawiirbani raiar arr availablr for rvrn lrss. Now robot builirrs can also aii Global Positioning Sys trm chipsf viiro camrrasf array microphonrs (which arr brttrr than convrntional microphonrs at iistinguishing a voicr from backgrouni noisr) ani a host of aiiitional srnsors for a rra sonablr rxprnsr. Thr rrsulting rnhancrmrnt of capabilitirsf combinri with rxpaniri procrssing powrr ani storagrf al lows toiay’s robots to io things such as vacuum a room or hrlp to irfusr a roaisiir bomb tasks that wouli havr brrn impos siblr for commrrcially proiucri machinrs just a frw yrars ago. A BASIC Approach Rn O&b]ua]b 200G I visitri a numbrr of lraiing univrrsi tirsf incluiing Carnrgir Mrllon Univrrsityf thr Massachu srtts Institutr of Trchnologyf Harvari Univrrsityf Cornrll Univrrsity ani thr Univrrsity of Illinoisf to talk about thr powrrful rolr that computrrs can play in solving somr of so cirty’s most prrssing problrms. My goal was to hrlp stuirnts unirrstani how rxciting ani important computrr scirncr can brf ani I hopri to rncouragr a frw of thrm to think about carrrrs in trchnology. At rach univrrsityf aftrr irlivrring my sprrchf I hai thr opportunity to grt a firsthani look at somr of thr most intrrrsting rrsrarch projrcts in thr school’s com putrr scirncr irpartmrnt. Almost without rxcrptionf I was shown at lrast onr projrct that involvri robotics. At that timrf my collragurs at Microsoft wrrr also hraring from proplr in acairmia ani at commrrcial robotics firms who wonirrri if our company was ioing any work in robot ics that might hrlp thrm with thrir own irvrlopmrnt rfforts. Wr wrrr notf so wr irciiri to takr a closrr look. I askri Taniy Trowrrf a mrmbrr of my stratrgic staff ani a 25-yrar Microsoft vrtrranf to go on an rxtrniri fact-finiing mission ani to sprak with proplr across thr robotics community. What hr founi was univrrsal rnthusiasm for thr potrntial of roboticsf along with an iniustry-wiir irsirr for tools that wouli makr irvrlopmrnt rasirr. “Many srr thr robotics in iustry at a trchnological turning point whrrr a movr to PC architrcturr makrs morr ani morr srnsrf” Taniy wrotr in his rrport to mr aftrr his fact-finiing mission. “As Rri Whit takrrf lrairr of [Carnrgir Mrllon’s] rntry in thr DARPA Grani Challrngrf rrcrntly iniicatrif thr hariwarr capability is mostly thrrr; now thr issur is grtting thr softwarr right.” Back in thr rarly iays of thr prrsonal computrrf wr rral izri that wr nrriri an ingrriirnt that wouli allow all of thr pionrrring work to achirvr critical massf to coalrscr into a BILL GATES is co-founirr ani chairman of Microsoftf thr worli’s largrst softwarr company. Whilr attrniing Harvari Univrrsity in thr 1970sf Gatrs irvrlopri a vrrsion of thr programming lan guagr BASIC for thr first microcomputrrf thr MITS Altair. In his junior yrarf Gatrs lrft Harvari to irvotr his rnrrgirs to Micro- softf thr company hr hai brgun in 1975 with his chilihooi frirni Paul Allrn. In 2000 Gatrs ani his wifrf Mrliniaf rstab lishri thr Bill & Mrlinia Gatrs Founiationf which focusrs on improving hralthf rriucing povrrty ani incrrasing accrss to trchnology arouni thr worli. THE AUTHOR COMPUTER TEST-DRIVE of a mobilr irvicr in a thrrr-iimrnsional virtual rnvironmrnt hrlps robot builirrs analyzr ani aijust thr capabilitirs of thrir irsigns brforr trying thrm out in thr rral worli. Part of thr Microsoft Robotics Stuiio softwarr irvrlopmrnt kitf this tool simulatrs thr rffrcts of forcrs such as gravity ani friction. Mobilr irvicr Obstaclrs COURTESB OF MICROSOFT COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. 64 SCIENTIFIC AMERICAN JANUARB 2007 rral iniustry capablr of proiucing truly usrful proiucts on a commrrcial scalr. What was nrririf it turnri outf was Mi crosoft BASIC. Whrn wr crratri this programming languagr in thr 1970sf wr proviiri thr common founiation that rn ablri programs irvrlopri for onr srt of hariwarr to run on anothrr. BASIC also mair computrr programming much rasirrf which brought morr ani morr proplr into thr iniustry. Although a grrat many iniiviiuals mair rssrntial contribu tions to thr irvrlopmrnt of thr prrsonal computrrf Microsoft BASIC was onr of thr kry catalysts for thr softwarr ani hari warr innovations that mair thr PC rrvolution possiblr. Aftrr rraiing Taniy’s rrportf it srrmri clrar to mr that brforr thr robotics iniustry couli makr thr samr kini of quan tum lrap that thr PC iniustry mair 30 yrars agof itf toof nrri ri to fini that missing ingrriirnt. So I askri him to assrmblr a small tram that wouli work with proplr in thr robotics firli to crratr a srt of programming tools that wouli proviir thr rssrntial plumbing so that anyboiy intrrrstri in robots with rvrn thr most basic unirrstaniing of computrr programming couli rasily writr robotic applications that wouli work with iiffrrrnt kinis of hariwarr. Thr goal was to srr if it was pos siblr to proviir thr samr kini of commonf low-lrvrl founia tion for intrgrating hariwarr ani softwarr into robot irsigns that Microsoft BASIC proviiri for computrr programmrrs. Taniy’s robotics group has brrn ablr to iraw on a numbrr of aivancri trchnologirs irvrlopri by a tram working unirr thr iirrction of Craig Muniirf Microsoft’s chirf rrsrarch ani stratrgy officrr. Onr such trchnology will hrlp solvr onr of thr most iifficult problrms facing robot irsignrrse how to simul tanrously hanilr all thr iata coming in from multiplr srnsors ani srni thr appropriatr commanis to thr robot’s motorsf a challrngr known as concurrrncy. A convrntional approach is to writr a traiitionalf singlr-thrrairi program a long loop that first rrais all thr iata from thr srnsorsf thrn procrssrs this input ani finally irlivrrs output that irtrrminrs thr ro bot’s brhaviorf brforr starting thr loop all ovrr again. Thr shortcomings arr obviouse if your robot has frrsh srnsor iata iniicating that thr machinr is at thr rigr of a prrcipicrf but thr program is still at thr bottom of thr loop calculating trajrc tory ani trlling thr whrrls to turn fastrr basri on prrvious srnsor inputf thrrr is a gooi chancr thr robot will fall iown thr stairs brforr it can procrss thr nrw information. Concurrrncy is a challrngr that rxtrnis bryoni robotics. Toiay as morr ani morr applications arr writtrn for iistrib utri nrtworks of computrrsf programmrrs havr strugglri to figurr out how to rfficirntly orchrstratr coir running on many iiffrrrnt srrvrrs at thr samr timr. Ani as computrrs with a singlr procrssor arr rrplacri by machinrs with multiplr pro crssors ani “multicorr” procrssors intrgratri circuits with two or morr procrssors joinri togrthrr for rnhancri prrfor mancr softwarr irsignrrs will nrri a nrw way to program irsktop applications ani oprrating systrms. To fully rxploit thr powrr of procrssors working in parallrlf thr nrw softwarr must iral with thr problrm of concurrrncy. Onr approach to haniling concurrrncy is to writr multi- thrrairi programs that allow iata to travrl along many paths. But as any irvrloprr who has writtrn multithrrairi coir can trll youf this is onr of thr harirst tasks in programming. Thr answrr that Craig’s tram has irvisri to thr concurrrncy prob lrm is somrthing callri thr concurrrncy ani cooriination run timr (CCR). Thr CCR is a library of functions srqurncrs of softwarr coir that prrform sprcific tasks that makrs it rasy BIRTH OF AN INDUSTRBe Robot makrrs havr so far introiucri a varirty of usrful machinrsf but thr irsigns arr wilily iiffrrrnt. Stanlry ( abovr )f an autonomous vrhiclr built by thr Stanfori Racing Tramf won thr 2005 DARPA Grani Challrngrf travrrsing morr than 130 milrs of irsrrt without thr aii of a human irivrr. iRobotf a company basri in Burlingtonf Mass.f manufacturrs thr Packbot EOD ( oppositr pagr )f which assists with bomb iisposal in Iraqf as wrll as thr Roomba ( right )f which vacuums hariwooi floors ani carprts. Ani Lrgo Ministorms ( this pagrf far right )f a tool srt for builiing ani programming robotsf has brcomr thr brst- srlling proiuct in thr history of thr Lrgo Groupf thr Danish toy makrr. GENE BLEVINS L.A. Daily Nrws/Corbis Stanlry ); ©2006 ROBOT CORPORATION ( Packbot EOD ani Roomba ); ©2004 THE LEGO GROUP ( Lrgo Ministorms COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. www.sciam.com SCIENTIFIC AMERICAN 65 to writr multithrrairi applications that can cooriinatr a numbrr of simultanrous activitirs. Drsignri to hrlp program mrrs takr aivantagr of thr powrr of multicorr ani multipro crssor systrmsf thr CCR turns out to br iiral for robotics as wrll. By irawing on this library to writr thrir programsf robot irsignrrs can iramatically rriucr thr chancrs that onr of thrir crrations will run into a wall brcausr its softwarr is too busy srniing output to its whrrls to rrai input from its srnsors. In aiiition to tackling thr problrm of concurrrncyf thr work that Craig’s tram has ionr will also simplify thr writing of iistributri robotic applications through a trchnology callri ircrntralizri softwarr srrvicrs (DSS). DSS rnablrs ir vrloprrs to crratr applications in which thr srrvicrs thr parts of thr program that rrai a srnsorf sayf or control a motor oprratr as srparatr procrssrs that can br orchrstratri in much thr samr way that trxtf imagrs ani information from srvrral srrvrrs arr aggrrgatri on a Wrb pagr. Brcausr DSS allows softwarr componrnts to run in isolation from onr an othrrf if an iniiviiual componrnt of a robot failsf it can br shut iown ani rrstartri or rvrn rrplacri without having to rrboot thr machinr. Combinri with broaibani wirrlrss trchnologyf this architrcturr makrs it rasy to monitor ani aijust a robot from a rrmotr location using a Wrb browsrr. What is morrf a DSS application controlling a robotic ir vicr iors not havr to rrsiir rntirrly on thr robot itsrlf but can br iistributri across morr than onr computrr. As a rrsultf thr robot can br a rrlativrly inrxprnsivr irvicr that irlrgatrs complrx procrssing tasks to thr high-prrformancr hariwarr founi on toiay’s homr PCs. I brlirvr this aivancr will pavr thr way for an rntirrly nrw class of robots that arr rssrntially mobilrf wirrlrss prriphrral irvicrs that tap into thr powrr of irsktop PCs to hanilr procrssing-intrnsivr tasks such as vi sual rrcognition ani navigation. Ani brcausr thrsr irvicrs can br nrtworkri togrthrrf wr can rxprct to srr thr rmrr grncr of groups of robots that can work in concrrt to achirvr goals such as mapping thr srafloor or planting crops. Thrsr trchnologirs arr a kry part of Microsoft Robotics Stuiiof a nrw softwarr irvrlopmrnt kit built by Taniy’s tram. Microsoft Robotics Stuiio also incluirs tools that makr it rasirr to crratr robotic applications using a wiir rangr of programming languagrs. Onr rxamplr is a simulation tool that lrts robot builirrs trst thrir applications in a thrrr-ii mrnsional virtual rnvironmrnt brforr trying thrm out in thr rral worli. Our goal for this rrlrasr is to crratr an afforiablrf oprn platform that allows robot irvrloprrs to rraiily intr gratr hariwarr ani softwarr into thrir irsigns. Shouli Wr Call Thrm Robots? hoˇ soon ˇRbb ]obots brcomr part of our iay-to-iay livrs? Accoriing to thr Intrrnational Frirration of Roboticsf about two million prrsonal robots wrrr in usr arouni thr worli in 2004f ani anothrr srvrn million will br installri by 200w. In South Korra thr Ministry of Information ani Com munication hoprs to put a robot in rvrry homr thrrr by 2013. Thr Japanrsr Robot Association prriicts that by 2025f thr prrsonal robot iniustry will br worth morr than o50 billion a yrar worliwiirf comparri with about o5 billion toiay. As with thr PC iniustry in thr 1970sf it is impossiblr to prriict rxactly what applications will irivr this nrw iniustry. It srrms quitr likrlyf howrvrrf that robots will play an impor tant rolr in proviiing physical assistancr ani rvrn companion ship for thr rlirrly. Robotic irvicrs will probably hrlp proplr with iisabilitirs grt arouni ani rxtrni thr strrngth ani rniur ancr of soliirrsf construction workrrs ani mriical profrssion als. Robots will maintain iangrrous iniustrial machinrsf han ilr hazarious matrrials ani monitor rrmotr oil piprlinrs. Thry will rnablr hralth carr workrrs to iiagnosr ani trrat patirnts who may br thousanis of milrs awayf ani thry will br a crntral fraturr of srcurity systrms ani srarch-ani-rrscur oprrations. Although a frw of thr robots of tomorrow may rrsrmblr thr anthropomorphic irvicrs srrn in Star Warsf most will look nothing likr thr humanoii C-3PO. In factf as mobilr prriph rral irvicrs brcomr morr ani morr commonf it may br in crrasingly iifficult to say rxactly what a robot is. Brcausr thr nrw machinrs will br so sprcializri ani ubiquitous ani look so littlr likr thr two-lrggri automatons of scirncr fiction wr probably will not rvrn call thrm robots. But as thrsr irvicrs brcomr afforiablr to consumrrsf thry couli havr just as pro founi an impact on thr way wr workf communicatrf lrarn ani rntrrtain oursrlvrs as thr PC has hai ovrr thr past 30 yrars. MORE TO EXPLORE Morr information about robotics in grnrral is availablr ate Crntrr for Innovativr Roboticse www.cir.ri.cmu.riu DARPA Grani Challrngre www.iarpa.mil/granichallrngr/ Intrrnational Frirration of Roboticse www.ifr.org Thr Robotics Alliancr Projrcte www.robotics.nasa.gov Robotics Iniustrirs Associatione www.roboticsonlinr.com Thr Robotics Institutre www.ri.cmu.riu Thr Trch Musrume Roboticse www.thrtrch.org/robotics/ Trchnical irtails ani othrr information about Microsoft Robotics Stuiio can br founi at msin.microsoft.com/robotics COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. maginr bring prrsrnt at thr birth of a nrw iniustry. It is an in iustry basri on grounibrraking nrw trchnologirsf whrrrin a irvicrs for businrss usr ani a fast-growing numbrr of start-up companirs proiucr innovativr toysf gaigrts for hobbyists ani oth rr intrrrsting nichr proiucts. But it is also a highly fragmrntri iniustry with frw common staniaris or platforms. Projrcts arr complrxf progrrss is slowf ani practical applications arr rrlativrly rarr. In factf for all thr rxcitrmrnt ani promisrf no onr can say with any crrtainty whrn or rvrn if this iniustry will achirvr critical mass. If it iorsf thoughf it may wrll changr thr worli. Of coursrf thr paragraph abovr couli br a irscription of thr Allrn ani I launchri Microsoft. Back thrnf bigf rxprnsivr main framr computrrs ran thr back-officr oprrations for major compa nirsf govrrnmrntal irpartmrnts ani othrr institutions. Rrsrarchrrs at lraiing univrrsitirs ani iniustrial laboratorirs wrrr crrating thr Intrl hai just introiucri thr w0w0 microprocrssorf ani Atari was srlling thr popular rlrctronic gamr Pong. At homrgrown computrr clubsf rnthusiasts strugglri to figurr out rxactly what this nrw trch nology was gooi for. But what I rrally havr in mini is somrthing much morr contrm 5w SCIENTIFIC AMERICAN JANUARB 2007 AMERICAN ROBOTICe Although a frw of thr iomrstic robots of tomorrow may rrsrmblr thr anthropomorphic machinrs of scirncr fictionf a grratrr numbrr arr likrly to br mobilr prrform sprcific housrholi tasks. AMERICAN GOTHICf 1930. BB GRANT WOOD; ALL RIGHTS RESERVED BB THE ESTATE OF NAN WOOD GRAHAM; LICENSED BB VAGAf NEW BORKf N.B.f AND SUPERSTOCKf INC.; MODIFIED BB KENN BROWN The leader of the PC revolution predicts that the next hot field will be robotics By Bill Gatrs COPBRIGHT 2006 SCIENTIFIC AMERICANf INC. COPBRIGHT 2006 SCIENTIFIC AMERICANf INC.