IntrSCity: Addrssing Futur Intrnt Rsarch Challngs for Smart Citis Danil Macêdo Batista, Alfrdo Goldman, Robrto Hirata Jr., Fabio Kon Dpartmnt of Computr Scinc Univrsity of São Paulo Email: {batista,gold,hirata,kon}@im.usp.br Fabio M. Costa Instituto d Informática Univrsidad Fdral d Goiás Email: fmc@inf.ufg.br Markus Endlr Dpartmnt of Informatics PUC-Rio Email: ndlr@inf.puc-rio.br Abstract Th Futur Intrnt will intgrat larg-scal systms constructd from th composition of thousands of distributd srvics, whil intracting dirctly with th physical world via snsors and actuators, which compos th Intrnt of Things. This Futur Intrnt will nabl th ralization of th Smart Citis vision, in which th urban infrastructur will b usd to its fullst xtnt to offr a bttr quality of lif for its citizns. Ky to th fficint and ffctiv ralization of Smart Citis is th scintific and tchnological rsarch covring th multipl layrs that mak up th Intrnt. This papr discusss th rsarch challngs and initiativs rlatd to Futur Intrnt and Smart Citis in th scop of th IntrSCity projct. Th challngs and initiativs ar organizd in thr fronts: (1) Ntworking and High-Prformanc Distributd Computing; (2) Softwar Enginring for th Futur Intrnt; and (3) Analysis and Mathmatical Modling for th Futur Intrnt and Smart Citis. IntrSCity aims at dvloping an intgratd opn-sourc platform containing all th major building blocks for th dvlopmnt of robust, intgratd, sophisticatd applications for th smart citis of th futur. Indx Trms Smart Citis, Intrnt of Things, Futur Intrnt, Big Data, Machin Larning, Mathmatical Modling. I. INTRODUCTION Th trm Smart City has gaind traction in rcnt yars; it has bn usd by rsarchrs from multipl filds to rfr to th us of modrn approachs to us city rsourcs in a mor intllignt way. Som xtrapolat th softwar contxt, focusing on social or businss aspcts. Rgarding softwar systms, many authors dfin a Smart City as th intgration of social, physical, and IT infrastructur to improv th quality of city srvics [1], [2]. Othr authors focus on a st of Information and Communication Tchnology (ICT) tools usd to crat an intgratd nvironmnt [2], [3], [4]. Whn valuating th various dfinitions rlatd to IT, intgration is a frquntly citd aspct and w considr th tchnologis associatd with th Futur Intrnt an ssntial aspct to guarant such a intgration. W ar alignd with th vision that th city infrastructur must provid an intgratd nvironmnt, facilitating th dvlopmnt, dploymnt, and opration of introprabl smart city applications from multipl domains. Basd on that, in our viw [5]: a Smart City is a city in which its social, businss, and tchnological aspcts ar supportd by Information and 978 1 5090 4671 3/16/$31.00 c 2016 IEEE Communication Tchnologis to improv th xprinc of th citizn within th city. To achiv that, th city must provid public and privat srvics that oprat in an intgratd, affordabl, and sustainabl way. Dspit th xistnc of multipl smart city initiativs in diffrnt countris around th world [1], [6], [7], [8], [9], ths dploymnts ar oftn basd on custom systms that ar nithr introprabl, nor portabl across citis, xtnsibl, or cost-ffctiv [10]. To rsolv ths limitations, th rsarch community must addrss ky challngs in th aras of Ntworking and High-Prformanc Distributd Computing, Softwar Enginring, Machin Larning, Data Analysis, and Mathmatical Modling, among othrs. Ths challngs involv highly-spcializd knowldg, rquiring a dp undrstanding of diffrnt aras, but thy also rquir clos coopration among spcialists from multipl filds working in a collaborativ way. IntrSCity 1 is a multidisciplinary projct aiming at addrssing th ky rsarch challngs around smart citis from a softwar infrastructur point of viw. Its goal is to dvlop an intgratd opn-sourc platform containing all major building blocks for th dvlopmnt of robust, intgratd, sophisticatd applications for th smart citis of th futur. This papr aims at motivating rsarchrs and graduat studnts to ngag in innovativ rsarch on Smart Citis, hlping to advanc th stat-of-th-art. Th papr is organizd as follows. Sction II discusss in mor dtail th objctivs of th IntrScity projct. Sction III prsnts th rsarch challngs that will nd to b addrssd to mt thos objctivs. Sction IV prsnts our rfrnc architctur for Smart City platforms, and Sction V concluds th papr and considrs th rsults that ar xpctd from th IntrSCity projct. II. REQUIREMENTS AND OBJECTIVES To bttr dfin IntrSCity s objctivs, w conductd an xtnsiv litratur rviw on Smart City Platforms [5]. Aftr analyzing 47 projcts, w drivd th rquirmnts for Smart City Softwar Platforms, which ar dividd into functional and non-functional. Th most rlvant functional rquirmnts ar 1 http://intrscity.org
[big] data managmnt, application run-tim support, wirlss snsor ntwork managmnt, data procssing and HPC, xtrnal/opn data accss, srvic managmnt, advancd softwar nginring mthods and tools, and city modls. Th most rlvant non-functional rquirmnts ar introprability, scalability, scurity, privacy, contxt awarnss, runtim adaptation, xtnsibility, configurability, and nrgy awarnss. Mor dtails about ach rquirmnt can b found in [5]. To mt ths rquirmnts and work towards our Smart City Softwar platform vision, th main objctivs of th IntrSCity projct hav bn dfind as follows. 1) Th dvlopmnt of analytic mathmatical modls for th rprsntation, study and analysis of larg structurd collctions of objcts in futur Smart City nvironmnts. Ths modls will b instrumntal for th fficint managmnt and planning of rsourc usag and dploymnt in smart citis, with th goal of optimizing prformanc, maximizing sustainability of systms and promoting savings of nrgy and othr rsourcs. 2) Dsign, implmntation and valuation of nw softwar architcturs for distributd, larg-scal, slfconfigurabl and slf-adapting systms basd on intllignt algorithms and machin larning tchniqus. 3) Dsign, implmntation and valuation of nw communication protocols and mchanisms to procss larg amounts of multimdia data and strams (txt, audio, imag, vido, tc.) with guarantd quality of srvic (QoS). 4) Schduling algorithms for larg scal nvironmnts involving larg numbrs of tasks and rsourcs in multipl clouds, mobil dvics and ntworks of snsors and actuators. 5) Dvlopmnt and xprimntation with nw mthods, tchniqus and softwar nginring tools and IDEs to support th dvlopmnt, tsting and dploymnt of high-quality, complx, larg-scal distributd systms for Smart Citis. 6) Application of th tchnologis dvlopd within th projct in ral-world scnarios involving ral smart city srvics, aiming at improving quality of lif for citizns and providing solutions for mor fficint managmnt of larg and mga citis. This will b achivd both via partnrships with city govrnmnts and through tchnology transfr to th public and privat sctors, spcially considring startups. III. RESEARCH CHALLENGES To fac th scintific and tchnological challngs and provid innovativ solutions to th problms around th Futur Intrnt and Smart Citis, th IntrSCity projct will dvlop scintific and tchnological rsarch covring th diffrnt layrs that compos ntworkd srvics and applications. Th projct will contribut to th dvlopmnt of th Futur Intrnt and to th dvlopmnt of rusabl tchnologis and mthods for Smart Citis, with spcial focus on problms such as urban mobility in citis of th dvloping world. Mor spcifically, as illustratd in Figur 1, rsarch is bing dvlopd at thr lvls that ovrlap and intract with ach othr: Figur 1. Tchnical-Scintific Structur of th IntrSCity Projct Infrastructur includs basic support for computr ntworks, protocols, cloud datacntrs, softwar dfind ntworks, and snsor ntworks. Middlwar includs softwar systms for th support and managmnt of virtual machins, srvic composition, communication, Big Data, Intrnt of Things, procsss schduling, workflows, data stram managmnt, and nrgy managmnt. Srvics and applications includ social ntworks, collaborativ ntworks, advancd Big Data mchanisms to procss larg volums of multimdia strams, and spcific support for Smart Citis. Notably, infrastructur and middlwar srvics will provid a basis for th dvlopmnt of high lvl intgratd srvics and applications in a numbr of intrrlatd domains. This will ultimatly nabl stratgic dcision making and th planning of public policis, and will subsidiz th dvlopmnt and opration of tools to manag th diffrnt vryday aspcts of smart citis.
Th objcts of rsarch shown in Figur 1 will b tackld from thr diffrnt prspctivs, as shown at th bottom of th figur. Th rspctiv rsarch challngs and initiativs ar discussd nxt. A. Ntworking and High-Prformanc Distributd Computing With fast growing dploymnt of snsors and smart dvics in all sctors of conomy and our daily lif, th Intrnt of Things (IoT) is bcoming a rality [11]. This can b dfind as a global ntwork of htrognous dvics and objcts, addrssabl in a uniform way and intracting through standard communication protocols, thrby xtnding th ubiquitous connctivity alrady prsnt in th currnt wirlss Intrnt (3G/4G and WiFi) to all sorts of dvics, including snsors and appliancs with procssing capacity and wirlss intrfacs (th so-calld Smart Objcts). IoT also ncompasss th dirct objct-to-objct or machin-to-machin (M2M) communication, without th intrvntion of popl, and at global scal. According to som rliabl projctions, until 2020, th numbr of nods/objcts in th IoT will xcd 50 billion, ncompassing applications in practically all sctors of th conomy, such as manufacturing, commrc, halthcar, financial systms, transport, scurity, homs, and smart buildings [12]. To support larg-scal M2M communication and th procssing of larg volums of snsor data from Smart Objcts, IoT systms count on cloud-basd infrastructur srvics for mobility (known as Mobil Cloud). Thus, th majority of th proposd architcturs for IoT consists of th following four gnral ntwork layrs/typs: opportunistic/ad hoc ntworks that oprat at low powr and short rangs (Edg Ntworks); long-rang ntworks btwn hubs and cloud srvrs (Dvicto-Cloud); ntworks btwn srvrs of a data-cntr (Intra- Cloud); and ntworks btwn cntrs (Intr-Cloud), as shown in Figur 2. Figur 2. Ntwork layrs for IoT Du to th incrasing amount of Smart Objcts and thir gographical disprsion and htrognity, as wll as th divrsity of th ntworks involvd, th IoT prsnts many challngs, mainly rlatd to introprability, discovry, slction, and accss to snsors and actuators in Smart Objcts, fficint communication at all lvls, and th dvlopmnt and maintnanc of applications and systms. In particular, IoT poss major challngs rlatd to th capacity to procss continuous flows of snsor data gnratd by a larg numbr of Smart Objcts. This charactrizs th Data Stram Procssing paradigm, whr data is volatil and is only rlvant immdiatly aftr its gnration, whil th quris on this data ar prmannt, as thy may nd to b procssd rpatdly ovr long priods of tim. Th fficint procssing and th managmnt of ths flows of data ar transvrsal concrns of th IoT that nd to b invstigatd urgntly, as thy will affct th data modls of futur IoT systms. Solutions for th majority of ths challngs rquir th dvlopmnt of Middlwar infrastructurs for IoT [13], in addition to daling with htrognity, introprability, compnsation for diffrnt capacitis and volatility of nods and ntworks, and th fficint procssing of data flows. Ths middlwar systms should also provid a st of abstractions (concpts and APIs) that assist in th projct, xcution, and maintnanc of IoT applications. Thr ar also diffrnt modls bing succssfully applid in distinct aras, such as wathr forcast, traffic control, smart buildings, and nrgy and wast managmnt. Many of thos modls ar tratd sparatly. Howvr, in ordr to optimiz th dsign, volution, and opration of citis, it is ssntial to quantify, undrstand and formaliz thir intraction. Nowadays, xampls of such sophisticatd (composit) modls alrady us powrful suprcomputrs, pointing at th nd for furthr rsarch on xascal computing. Som tchniqus, tools and mchanisms bing usd to fac ths challngs ar th SDDL (Scalabl Data Distribution Layr) middlwar [14], nsrfidsim [15], th MapRduc programming modl, workflow managmnt systms, Complx Evnt Procssing (CEP), mrging hardwar pattrns for Fog Computing, Softwar Dfind Ntworking (SDN), Ntwork Function Virtualization (NFV), and th FITS tstbd (Futur Intrnt Tstbd with Scurity) [16]. B. Softwar Enginring for th Futur Intrnt Building complx larg-scal systms such as th ons ndd to comply with th rquirmnts prsntd in Sction II will invitably dmand a disciplind approach basd on Softwar Enginring. Howvr, th currnt tchnologis and mthodologis providd by xisting practics in Softwar Enginring do not compltly addrss th challngs involvd in th dvlopmnt of softwar systms for th Futur Intrnt and Smart Citis. This nw scnario involvs thousands of popl dvloping systms that must b intgratd in a complx mannr, rsulting in vn mor complx systms of systms [17]. This trnd will lad to an xplosiv growth in th dgr of distribution and htrognity of ths systms, as wll as in thir scal, both in trms of numbr of usrs and numbr of machins. In this contxt, it will bcom incrasingly common to crat systms composd of humans, srvics and things, which intract among thmslvs to prform distributd activitis. Th complxity of larg-scal systms, combind with issus such as introprability, htrognity,
mobility, adaptability, scurity, and privacy constitut som of th challngs imposd by futur Smart Citis. Dtrmining th tchniqus and mthods that ar suitabl in this contxt is still an opn issu. Th mrgnc of a nw dvlopmnt scnario such as dscribd abov causs nw workflows to b cratd in an ad hoc way, many of thm suscptibl to machin and human rrors, as wll as to rwork and intgration failurs. To avoid ths undsirabl ffcts, diffrnt aras of Softwar Enginring must b rvisitd and xtndd to adapt to th nw challngs. This will rquir, in many cass, a complt rconcption of th mthods and th dvlopmnt of nw tchniqus, approachs and tools [18]. To that nd, th following lmnts ar ndd: (1) architctural pattrns, softwar architcturs and approachs to dal with th complxity of Futur Intrnt systms, whil considring th nd for opnnss, flxibility, and xtnsibility; (2) nw tools, mthods and mtrics to valuat th intrnal and xtrnal quality of softwar for th Futur Intrnt, making th rus of cod incrasingly likly, rducing maintnanc costs, nhancing usability, and nabling th volution of softwar to incorporat nw functionality and to adapt to changing contxts and nds; (3) nw protocols, tchniqus and procsss nabling th construction of robust, rsilint systms with strong support for fault tolranc; and (4) nw tools, IDEs, procsss and guidlins to support collaborativ dvlopmnt of larg-scal systms and applications for futur Smart Citis, with mphasis on crating and fostring distributd opn sourc softwar communitis and on th widsprad us of this kind of softwar. Som tchniqus, tools, and mchanisms currntly bing usd to fac ths challngs ar agil softwar dvlopmnt, xprimntal softwar nginring, mining of softwar rpositoris, discovry and managmnt of intra- and intr-systms dpndncis btwn artifacts and moduls, analysis of sociotchnical ntworks, automatd classification of softwar quality and thrats, srvic-orintd architcturs, and Wb srvics composition. C. Analysis and Mathmatical Modling for th Futur Intrnt and Smart Citis Th ra of scarcity of ral data, whn rsarchrs and practitionrs would, most of th tim, dpnd on data gnratd by simulation, sms to b coming to an nd, xcpt for a fw spcific aras. Currntly, th big challng is having larg amounts of data to dal with. Particularly for smart citis, Big Data plays a fundamntal rol [19]. Th following factors hav causd this data xplosion: fast dvlopmnt and widsprad adoption of snsor tchnology (such as position and motion snsors, camras, and a varity of lctromagntic and chmical snsors); production and availability of larg amounts of contnt on th Intrnt; and th larg numbr of computrs and othr kinds of smart dvics connctd to th Intrnt. This abundanc of data will hlp to undrstand svral phnomna that affct th daily lif, such as thos involving knowldg of physical phnomna (in th nvironmnt, tlcommunications, transport, tc.), knowldg of biological phnomna (such as nw mdical tchniqus), and knowldg of social phnomna (politics, conomy, ducation, city managmnt, tc.), with th potntial to promot significant advancs in th quality of lif. Howvr, du to th natur of snsors, nois and missing data ar a fact of lif, dmanding nw mathmatical mthods and idas to avoid making big mistaks and big bad dcisions with Big Data. This raiss th nd for th continud volution of storag and information rtrival tchniqus, as wll as for advancs in data sampling, missing data analysis and inputation, anomaly dtction, statistical analysis, and high prformanc computing. Th analysis of mdium to larg datasts may still b fasibl nowadays. Howvr, it may not b practical in big data scnarios without statistical sampling tchniqus on vry larg datasts [20], [21] (du to tim and spac complxity). Rcnt advancs on sampling tchniqus, combind with databas tchnology and mthods, mainly for storag and rtrival, will probably bcom promising rsarch aras in databass. On th othr hand, th classical hypothsis tst, typically usd for small and non-rprsntativ sampls, may caus bad gnralizations on big datasts. Thrfor, sampling rprsntativ data from big datasts will also b an important problm to b solvd in th coming yars to achiv good accuracy and robustnss of th stimats. Th typ of data analysis rquird tnds to gnrat vn mor data, as it is basd on mathmatical transformations, whos intrmdiat rsults ar gnrally stord during th analysis procss. This will allow for furthr mta-analyss, namly th combination of rsults from svral prvious analyss. In th contxt of Smart Citis, an xampl of a particular kind of data that will nd intnsiv procssing is vido. It is xpctd that larg amounts of vidos (ssntially, tmporally ordrd sts of imags) of traffic and urban mobility will b mad availabl in ral tim for analysis. Efficint tchniqus nd to b dvlopd in this ara for th managmnt and manipulation of raw and procssd data. Th Futur Intrnt rquirs prforming pattrn rcognition on all sorts of mdia and data, in nvironmnts that ar full of nois, imprcision and uncrtaintis, which in turn can b modld by Fuzzy Sts (FS) and imprcis probability modls. A rcnt tchniqu that has provn to b vry ffctiv to lowr th classification rror mploys FS to gnrat tailord krnls for SVM (Support Vctor Machin) classifirs as a way to modl this kind of uncrtain data [22]. In addition, mthods wr cratd for discovring outlirs and anomalous substs of labld or unlabld data [23]. An application of ths tchniqus is th study of bus tickting data (.g., data about usrs and bus lins, dat, and tim of day) to discovr bhavior pattrns for diffrnt kinds of usrs, such as studnts, workrs and occasional usrs. For xampl, th analysis of th data by gndr and ag (modld as intrvals) can rval spcific bhavioral pattrns about snior usrs. Thr ar othr tchniqus, tools and mchanisms that
Strams HPC Big Data Managmnt Analytics Visualization Machin Larning Data Claning App Rpository Modl Rpository Data Rpository Srvic Middlwar Citizn Applications Managr IoT Middlwar Cloud & Ntworking Dvlopr Social Ntwork Gatway Usr Managmnt D v l o p m n t T o o l k i t S c u r i t y a n d P r i v a c y Figur 3. Rfrnc Architctur for Smart City Platforms hav bn usd to dal with big datasts and that will opn nw aras of rsarch, or improv xisting ons. A fw of ths aras includ: visualization tchniqus and algorithms, clustring algorithms, automatic dsign of imag procssing oprators, computr-aidd imag sgmntation, computational larning tchniqus, graph databass, and scalabl algorithms for tractabl fragmnts of th problm of probabilistic satisfiability (PSAT). IV. REFERENCE ARCHITECTURE To guid our work, as part of a comprhnsiv litratur rviw on Smart City Platforms that analyzd 23 diffrnt platforms [5], w drivd a rfrnc architctur for Futur Smart City Softwar Platforms. Figur 3 prsnts th lmnts rquird for th dvlopmnt of a highly ffctiv softwar platform to facilitat th construction of highly scalabl, intgratd Smart City applications. Th lowst lvl componnt of th rfrnc architctur is Cloud and Ntworking, which is rsponsibl for th managmnt and communication of city ntwork nods in a scalabl and xtnsibl way. This componnt is in charg of idntifying all th dvics connctd to th platform, including srvrs, snsors, actuators, and usr dvics. On top of th Cloud and Ntworking infrastructur, th rfrnc architctur includs th IoT Middlwar and th Srvic Middlwar componnts. Th formr is in charg of managing th city IoT ntwork and nabling ffctiv communication of th platform with usr dvics, city snsors, and actuators. Th lattr is in charg of managing th srvics that th platform provids to applications, prforming oprations such as publishing, nacting, monitoring, composing, and chorographing thos srvics. To provid bttr srvics to citizns, it is important for th platform to hav accss to som usr data and prfrncs, which is th rol of th Usr Managmnt componnt. Howvr, to nsur usr privacy, this data must b proprly protctd, and prmission to stor it must b acquird from th usr. Morovr, as th city platform will hav many applications, it can b hlpful to offr a singl sign-on mchanism. Social ntworks will play a major rol in Smart Citis. Thy can b usd to discovr data about city conditions, and can b an fficint communication channl btwn th platform and its usrs. Thrfor, it is important to allow th intgration of th Smart City platform with xisting social ntworks. This is th rol of th Social Ntwork Gatway. Th Big Data Managmnt modul manags all th data in th platform, including data collctd from th city and data gnratd by th platform. To this xtnt, it has thr rpositoris: (1) an App Rpository to stor applications, including thir sourc/binary cod, imags, and associatd documnts; (2) a Modl Rpository to stor city modls, such as traffic modls, snsor ntwork modls, data modls, city maps, and nrgy distribution modls; and (3) a Data Rpository to stor data collctd from snsors, citizns, and applications. Bsids data storag, th Big Data Managmnt modul is also rsponsibl for th procssing of city data. Thr ar two typs of data procssing that ar suitabl for diffrnt situations: Stram procssing and Analytics, to prform raltim analytics and data-flow procssing; and Batch procssing (HPC), to analyz larg datasts. Morovr, this modul must b capabl of prforming usful pr-procssing tasks, such as
data filtring, normalization, and transformation. Th Big Data modul also has a Machin Larning componnt, which facilitats undrstanding of th city by automatically building bhavior modls of city procsss and making prdictions of city phnomna. Sinc a Smart City will produc normous amounts of data, information Visualization is an important componnt, along with a Data Claning componnt for dlting data that is no longr ndd and archiving old data on slowr, high capacity stors. Th platform must also provid an application dvlopmnt toolkit, including tools such as an Intgratd Dvlopmnt Environmnt (IDE), libraris, and framworks, as wll as a Smart City Simulator for dbugging and xprimnting with applications bfor actual dploymnt. Finally, a numbr of non-functional proprtis, notably scurity and privacy, must b supportd across all layrs. W ar currntly working on th incrmntal implmntation of this architctur, following an Agil Softwar Dvlopmnt mthod, and rusing opn sourc tools, libraris, and framworks as much as possibl. Our initial prototyp, which is basd on Ruby on Rails, is availabl as opn sourc softwar at https://gitlab.com/groups/smart-city-softwar-platform and mor information about th projct can b found at http://intrscity.org. V. CONCLUSIONS AND EXPECTED RESULTS Th ralization of ffctiv Smart City applications dpnds on rsarch findings in th aras of Ntworking and High- Prformanc Distributd Computing, Softwar Enginring, and Analysis and Mathmatical Modling. Th solutions bing dvlopd in ths thr aras in th IntrSCity projct shall gnrat rusabl opn sourc tools, libraris, framworks, systms, and applications. This softwar will b availabl to th socity and will srv as a basis for th laboration of furthr projcts focusing on th Futur Intrnt, targting th acadmic community, govrnmnts, stablishd companis, and startups. In addition, w will sk to collct and crat opn datasts from diffrnt aras of a smart city to srv as a basis for xprimntation within th ara. W xpct that th tchnological and scintific rsarch undr dvlopmnt within th contxt of th IntrSCity projct will gnrat concrt rsults in th broadr ara of Smart Citis, as wll as in spcific applications rlatd to aras such as urban mobility and intllignt transportation systms, public safty, pollution control and air quality, ntrtainmnt, and watr, nrgy and wast managmnt. Spcifically, w shall dvlop a gnric opn sourc platform for data managmnt, communication and procssing of activitis in Smart Citis, which can b usd by corporations and govrnmnts for th nhancmnt of svral aspcts of urban lif. Acknowldgmnts This papr rcivd contributions from th profssors, graduat studnts, and industrial and govrnmntal partnrs that compos th IntrSCity rsarch tam, to whom w ar sincrly gratful. Th projct will b fundd by CNPq, proc. 465446/2014-0, and FAPESP. REFERENCES [1] A. Caragliu, C. Dl Bo, and P. Nijkamp, Smart citis in Europ, Journal of Urban Tchnology, vol. 18, no. 2, pp. 65 82, 2011. [2] R. G. Hollands, Will th ral smart city plas stand up? intllignt, progrssiv or ntrprnurial? City, vol. 12, no. 3, pp. 303 320, 2008. [3] D. Washburn, U. Sindhu, S. Balaouras, R. A. Dins, N. M. Hays, and L. E. Nlson, Hlping CIOs undrstand smart city initiativs, Growth, vol. 17, 2009. [4] R. E. Hall, B. Bowrman, J. Bravrman, J. Taylor, H. Todosow, and U. Von Wimmrsprg, Th vision of a smart city, in 2nd Intrnational Lif Extnsion Tchnology Workshop, Paris, 2000. [5] E. F. Z. Santana, A. P. Chavs, M. A. Grosa, F. Kon, and D. S. Milojicic, Softwar platforms for smart citis: Concpts, rquirmnts, challngs, and a unifid rfrnc architctur, Submittd for publication (availabl at http://arxiv.org/abs/1609.08089), 2016. [6] C. Manvill, G. Cochran, J. Cav, J. Millard, J. K. Pdrson, R. K. Thaarup, A. Lib, M. Wissnr, R. Massink, and B. Kottrink, Mapping smart citis in th EU, RAND Corporation, Tch. Rp. EP-50486, 2014. [7] P. Liu and Z. Png, China s smart city pilots: A progrss rport, Computr, vol. 47, no. 10, 2014. [8] M. Forts, V. Frrira, G. Sotlo, A. Cabral, W. Corria, and O. Pachco, Dploymnt of smart mtring in th Búzios City, in Transmission & Distribution Confrnc and Exposition-Latin Amrica (PES T&D-LA), 2014 IEEE PES. IEEE, 2014, pp. 1 6. [9] I. Janajrh, L. Su, and F. Alan, Wind nrgy assssmnt: Masdar City cas study, Rnwabl Enrgy, vol. 52, pp. 8 15, 2013. [10] Intrnational Tchnical Working Group on IoT-Enabld Smart City Framwork, NIST, 2016, https://pags.nist.gov/smartcitisarchitctur/. Accssd Spt. 27, 2016. [11] H. Schaffrs, N. Komninos, M. Pallot, B. Trouss, M. Nilsson, and A. Olivira, Smart citis and th futur intrnt: Towards coopration framworks for opn innovation, in Th Futur Intrnt, sr. LNCS. Springr Brlin Hidlbrg, 2011, vol. 6656, pp. 431 446. [12] L. Atzori, A. Ira, and G. Morabito, Th intrnt of things: A survy, Computr ntworks, vol. 54, no. 15, pp. 2787 2805, 2010. [13] C. Prra, A. Zaslavsky, P. Christn, and D. Gorgakopoulos, Contxt awar computing for th intrnt of things: A survy, Communications Survys & Tutorials, IEEE, vol. 16, no. 1, pp. 414 454, 2014. [14] L. David, R. Vasconclos, L. Alvs, R. André, and M. Endlr, A DDSbasd middlwar for scalabl tracking, communication and collaboration of mobil nods, Journal of Intrnt Srvics and Applications, vol. 4, no. 1, pp. 1 15, 2013. [15] R. P. B. Mota and D. M. Batista, An NS-2 modul for simulating passiv RFID systms, in Procdings of th 11th IEEE/IFIP Intrnational Confrnc on Embddd and Ubiquitous Computing (EUC). IEEE, 2013, pp. 2263 2270. [16] I. M. Moras, D. M. Mattos, L. H. G. Frraz, M. E. M. Campista, M. G. Rubinstin, L. H. M. Costa, M. D. d Amorim, P. B. Vlloso, O. C. M. Duart, and G. Pujoll, FITS: A flxibl virtual ntwork tstbd architctur, Computr Ntworks, vol. 63, no. 0, pp. 221 237, 2014, Spcial issu on Futur Intrnt Tstbds Part II. [17] P. H. Filr, K. Sullivan, K. C. Wallnau, R. P. Gabril, J. B. Goodnough, R. C. Lingr, T. A. Longstaff, R. Kazman, M. H. Klin, L. M. Northrop, and D. Schmidt, Ultra-Larg-Scal Systms: Th Softwar Challng of th Futur, 1st d. Softwar Enginring Institut, 2006. [18] I. Gorton, A. B. Bnr, and A. Mockus, Softwar nginring for big data systms, IEEE Softwar, vol. 33, no. 2, pp. 32 35, 2016. [19] E. A. Nuaimi, H. A. Nyadi, N. Mohamd, and J. Al-Jaroodi, Applications of big data to smart citis, Journal of Intrnt Srvics and Applications, vol. 6, no. 25, 2015. [20] S. Kumar, M. Mohri, and A. Talwalkar, Sampling mthods for th nyström mthod, Journal of Machin Larning Rsarch, vol. 13, no. Apr, pp. 981 1006, 2012. [21] A. Klinr, A. Talwalkar, P. Sarkar, and M. I. Jordan, A scalabl bootstrap for massiv data, Journal of th Royal Statistical Socity: Sris B (Statistical Mthodology), vol. 76, no. 4, pp. 795 816, 2014. [Onlin]. Availabl: http://dx.doi.org/10.1111/rssb.12050 [22] J. G. Díaz, R. H. Jr., and S. Canu, Positiv dfinit krnl functions on fuzzy sts, in IEEE Intrnational Confrnc on Fuzzy Systms, 2014. [23] J. G. Díaz, S. Canu, and R. H. Jr., Support masur data dscription for group anomaly dtction, in ODDx3 Workshop on Outlir Dfinition, Dtction, and Dscription, at 21st ACM SIGKDD Intl. Conf. on Knowldg Discovry and Data Mining, Sydny, Australia, 2015.