Software reference architecture for smart environments: Perception

So wa e e e ence a chi ec u e o sma en i onmen s: Pe cep ion A. Fe nández-Mon es a, ⁎,J.A.O ega a , J.I. Sánchez-Venzalá a , L. González-Ab il b a ETS Ing. In o má ica, Uni e sidad de Se illa, Spain b EU Es udios Emp esa iales, Uni e sidad de Se illa, Spain abs ac a icle in o A icle his o y: Recei ed 27 May 2012 Recei ed in e ised o m 14 Janua y 2014 Accep ed 3 Feb ua y 2014 A ailable online 19 Feb ua y 2014 Keywo ds: Sma en i onmen So wa e a chi ec u e Ambien in elligence Pe cep ion Wi h he inc ease o in elligen de ices, ubiqui ous compu ing is sp eading o all scopes o people li e. Sma home (o indus ial) en i onmen s include au oma ion and con ol de ices o sa e ene gy, pe o m asks, assis and gi e com o in o de o sa is y specific p e e ences. This pape ocuses on he p oposal o So wa e Re e ence A chi ec u e o he de elopmen o sma applica ions and hei deploymen in sma en i onmen s. The mo i a ion o his Re e ence A chi ec u e and i s benefi s a e also explained. The p oposal conside s h ee main p ocesses in he so wa e a chi ec u e o hese applica ions: pe cep ion, easoning and ac ing. Thispape cen esa en ionon hedefini ion o he Pe cep ion p ocess and p o ides an example o i s implemen a ion and subsequen alida ion o he p oposal. The so wa e p esen ed implemen s he Pe cep ion p ocess o a sma en i onmen o a s anda d o fice, by e ie ing da a om he eal wo ld and s o ing i o u he easoning and ac ing p ocesses. The objec i es o his solu ion include he p o ision o com o o he use s and he sa ing o ene gy in ligh ing. Th ough his e ifica ion, i is also shown ha de elopmen s unde his p oposal p oduce majo benefi s wi hin he so wa e li e cycle. © 2014 Else ie B.V. All igh s ese ed. 1. In oduc ion A sma en i onmen (SE) can be defined as one ha is able o acqui e and apply knowledge abou he en i onmen and i s inhabi an s in o de o imp o e hei expe ience in ha en i onmen [1]. Sma home echnologies a e an impo an pa o ubiqui ous compu ing. Ma k Weise [2] ou lined he p inciples o Ubiqui ous Com- pu ing: he pu pose o a compu e is o help someone do some hing. Nowadays, due o he popula isa ion o compu a ional de ices and applica ions, ubiqui ous compu ing is ecognised as a e olu ion in he de elopmen o sma en i onmen s. Ne e heless, so wa e a e ac s ela ed o ubiqui ous compu ing, o- ge he wi h he wide spec um o compu a ional de ices (and he so - wa e needed o ulfil hei missions) a e oo he e ogeneous and hence di ficul o compa e o classi y. Each pieceo so wa ee ol esinanisola - ed way o only in ela ion o he ha dwa e o which i has been de el- oped. The p oblem add essed in his pape in ol es he o ches a ion o he a chi ec u e o a gene al so wa e model o he de elopmen o SEs. This So wa e Re e ence A chi ec u ewould a ou he de elopmen o a sma en i onmen solu ion by inc easing he euse o componen s, p omo ing in e ope abili y, and defining he compe ences o each pa o he so wa e. A good compa ison o his could be he Open Sys em In e connec- ion (OSI) model, which is a p esc ip ion o cha ac e izing and s anda dizing he unc ions o a communica ions sys em in e ms o abs ac ion laye s. The ambi ious goal o his a chi ec u e o ces i o emain e y gene al and o lea e specific aspec s un il he implemen a ion s age. The benefi s o he app oach include a be e unde s anding o he issues ha mus be aced when de eloping each componen o a sma en i onmen solu ion. The So wa e Re e ence A chi ec u e educes he cos s o he main cycles o so wa e (design, de elopmen , deploymen , and main enance) and a ou s he in e ope abili y be ween a ious solu ions. The main goal o his wo k is he p oposal o So wa e Re e ence A chi ec u e o he de elopmen o SEs (see Sec ion 3), whe e all he componen s can in e ac flawlessly and each au oma ism objec i es. To his end, he a chi ec u e p oposed seeks o imp o e he modula - i y, eusabili y and ex ensibili y o solu ions, he eby allowing a mo e coo dina ed e olu ion o SEs, which cu en ly emain unde indi idual and isola ed de elopmen . The a chi ec u e defines a middlewa e amewo k ha connec s he modules and es ablishes he esponsibili y o each module. The benefi s o de elope s using a defined amewo k o s anda d a chi ec u e o he domain ha e been ho oughly s udied by Fayad and Schmid [28], and include: a educ ion and ocus o he e o in ol ed, a so lea ning cu e, in eg abili y, main ainabili y, easie alida ion, e ficiency, and a highe le el o s anda diza ion. Compu e S anda ds & In e aces 36 (2014) 928–940 ⁎Co esponding au ho . E-mail add esses: a [email protected] (A. Fe nández-Mon es), jo [email protected] (J.A. O ega), [email protected] (J.I. Sánchez-Venzalá), [email protected] (L. González-Ab il). h p://dx.doi.o g/10.1016/j.csi.2014.02.004 0920-5489/© 2014 Else ie B.V. All igh s ese ed. Con en s lis s a ailable a ScienceDi ec Compu e S anda ds & In e aces jou nal homepage: www.else ie .com/loca e/csi As an example o he a chi ec u e usage, his pape p esen s he Pe cep ion p ocess and p o ides an example o implemen a ion by ollowing he So wa e Re e ence A chi ec u e p oposed. Typical componen s o a SE ha e been ho oughly s udied in he li e a u e, al hough he app oach o Cook and Das [20] dese es special men ion since i is cu en ly he mos widely accep ed app oach. Fig. 1 shows he gene al o ganiza ion o hese componen s. Componen s a e di ided in o ou laye s: a) physical; b) communica ion; c) in o ma ion; and d) decision. This app oach joins ha dwa e wi h so wa e agen s, and hence e y he e ogeneous elemen s, such as a decision make and senso s o ac ua o s, appea in he same componen model. All hese componen s mus collabo a e in o de o achie e he goals o au oma ism ha a SE equi es. Which asks belong o each compo- nen and how hey should collabo a e cons i u e he main mo i a ion o he So wa e Re e ence A chi ec u e p oposed. The So wa e Re e ence A chi ec u e p oposed is di ided in o h ee main pa s: Pe cep ion, Reasoning and Ac ing. This pape ocuses on he defini ion o Pe cep ion, as he fi s s ep in he gene al p ocess. Sec ion 2 analyses ela ed wo k in his a ea, and in Sec ion 3,Re e ence A chi ec u e is p esen ed and he Pe cep ion p ocess is explained. Finally, e ifica ion wi h a p o o ype o he Pe cep ion p ocess is shown in Sec ion 4, and conclusions a e d awn in Sec ion 5. 2. Rela ed wo k Ambien in elligence is a ending opic, and hence a wide a ie y o ela ed esea ch ini ia i es ha e appea ed. One o he mos common ap- plica ions in ambien in elligence is ha o SEs. Many esea che s a ound he wo ld a e de eloping p ojec s which in ol e SEs. In his sec ion, some o he mos popula p ojec s a e e iewed and compa ed. The sec ion has been di ided depending on whe e each p ojec is ocused: gene al sma en i onmen s, echnologies o a chi ec u es. 2.1. Sma en i onmen p ojec s Da Cos a [12] ocuses on he challenges and issues ha ubiqui ous compu ing applica ions ha e o deal wi h and summa izes hem: he e ogenei y, scalabili y, dependabili y and secu i y, p i acy and us , spon aneous in e ope a ion, mobili y, con ex awa eness, con ex Fig. 1. The componen s o a SE by Cook and Das [20]. 929A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 managemen , anspa en use in e ac ion and in isibili y. Thus, SE p ojec s, as subp oduc s o ubiqui ous compu ing, ha e o add ess hose challenges. Fo example, he objec i e o he Ma Home p ojec [3] is o c ea e a home ha ac s as an in elligen agen . I s a chi ec u e o ganizes he en i onmen as an agen ha can, in u n, be di ided in o se e al in el- ligen agen s which in e ac . The echnologies wi hin each agen a e sepa a ed in o ou coope a ing laye s: decision, in o ma ion, commu- nica ion, and physical. Pe cep ion is a bo om-up p ocess, while ac ing is op-down. DomoSEC p ojec [5], o e s a home au oma ion solu ion ha co e s necessi ies in indoo domo ics. DomoSEC is based on OSGi and composed o an embedded compu e ha cen alizes he home “in elli- gence”, and communica es wi h he con olled de ices. The in en ion o he Awa e Home p ojec [8] is o p oduce an en i- onmen capable o managing in o ma ion abou i sel , i s occupan s, and hei ac i i ies. I is composed o a se o subsys ems esponsible o he a ious echnologies deployed: human–compu e in e ac ion, machine lea ning, compu a ional pe cep ion, wea able compu ing, e hnog aphy, so wa e enginee ing, and senso s. Ene gy Awa e Sma Home [7] ocuses on he in eg a ion o he e o- geneous embedded de ices and powe me e ing plugs by middlewa e called Hyd a. I s pu pose is o e ie e ene gy-consump ion da a o he moni o ing and analysis o ene gy consumed, in o de o p og amme, con ol and use home appliances e ficien ly. The Place Lab p ojec [9] is cen ed on cap u ing inhabi an s' beha - iou h ough a g oup o cabine y componen s, which con ain a mic o- con olle , speake s, came as, and a se o senso s. These senso s eco d a comple e audio– isual log o ac i i y. Place Lab da a s eams a e employed o de elop new con ex -de ec ion algo i hms and con ex -awa e compu ing applica ions. Pa ch Panel [18] is a mechanism o he inc emen al addi ion o modifica ion o beha iou in exis ing ubiqui ous compu ing en i on- men s (such as iRoom), o example, by adding new inpu modali ies o cho eog aphing he beha iou o exis ing independen applica ions. I p o ides a gene al acili y o e a ge ing e en flow, and enables in e ac ions be ween ne wo ked ha dwa e and so wa e componen s o be c ea ed o modified in ubicomp en i onmen s. EasyLi ing [16] is a Mic oso p ojec which pu sues he easy agg e- ga ion o I/O de ices in o a single and cohe en en i onmen . The sys ems include middlewa e o acili a e dis ibu ed compu ing, wo ld modelling o p o ide loca ion-based con ex , pe cep ion o collec in o - ma ion abou he wo ld's s a e, and se ice desc ip ion o suppo decomposi ion o de ice con ol, in e nal logic and use in e ace. O he sys ems, such as he Sen ien Compu ing Sys em [19],can change hei beha iou based on a model o he en i onmen s hey cons uc by using senso da a. This sys em s i es o emo e obs acles by sha ing and configu ing de ices o exploi a ion in he wo ld model. CASAS [6] is an adap i e sma -home sys em ha u ilizes machine- lea ning echniques o disco e pa e ns in he daily ac i i ies o esiden s and o gene a e au oma ion policies. Da a om senso s is analysed in o de o de e mine ac i i y pa e ns o in e es o au oma- ion. Pa e ns a e modelled con inuously in a mul ile el s uc u e o build he con ex . Finally, a selec ion o he ac i i ies o be au oma ed is pe o med. 2.2. P ojec s wi h echnological imp o emen s O he p ojec s a e mo e ocused in echnology, and p o ide se e al kinds o echnological imp o emen s. Some examples o his a e desc ibed below. The ATRACO p ojec [4] aims o suppo e e yday ac i i ies in a meaning ul way. I uses a emo e OSGi pla o m connec ed wi h a esiden ial ga eway, which manages he de ices, senso s and ac ua o s ia UPnP p o ocol. These elemen s wo king oge he o e a con ex - awa e se ice o he en i onmen . A high-le el p og amming language, called Visual RDK, is p oposed by Weis [10], o p o o yping pe asi e applica ions. This language gen- e a es a debugging applica ion and a p o o ype applica ion om he same sou ce. The main ad an age o his p oposal is ha con ex is igh ly in eg a ed in o he language i sel , and hence de elope s can a ach unc ionali y o loca ions, people, o si ua ions ins ead o o he de ice. Howe e Bannach [11] ocuses on he p oblem om ano he poin o iew: p o o yping o Ac i i y Recogni ion applica ions. I ea u es mech- anisms o dis ibu ed p ocessing and suppo s o mobile and wea able de ices. The CRN Toolbox is a ool se specifically op imized o he implemen a ion o mul imodal, dis ibu ed ac i i y and con ex ecog- ni ion sys ems unning on Posix ope a ing sys ems. I also con ains a collec ion o eady- o-use algo i hms (signal p ocessing, pa e n classifica ion, and so on). I s implemen a ion is specially op imized o mobile de ices. 2.3. P ojec s wi h ele an a chi ec u es The majo con ibu ion o some p ojec s lie in he a chi ec u e p oposed by hem. The ollowing p ojec s a e examples o his case. The GAIA [13] me aope a ing sys em ex ends he each o ope a ing sys ems o manage ubiqui ous compu ing habi a s and li ing spaces as in eg a ed p og ammable en i onmen s. I p esen s a middlewa e in- as uc u e o ac i e spaces. This middlewa e, unlike he a chi ec u e p oposed in his pape , is localized a he ope a i e sys em le el. JCAF (Ja a Con ex -Awa eness F amewo k) [14] is a Ja a-based con ex -awa eness and se ice-o ien ed in as uc u e and an API o c ea ing con ex -awa e applica ions. I g oups he elemen s o he a chi ec u e in o ou ca ego ies: Con ex Se ices, En i ies and Con ex , Con ex Clien s, and Con ex E en s. These elemen s communica e be ween each o he . In e play [15] is middlewa e so wa e which in eg a es he e oge- neous home de ices o simpli y hei con ol o he use . I allows use s o use a pseudo-English in e ace o achie e home asks wi hou di ficul y. I s a chi ec u e is o ganized in o fi e laye s, which sends o - de s om he use in e ace o an unde lying middlewa e o de ice managemen . I s objec i e is o p o ide ad anced con ol o he de ices. I-Cen ic Se ices [32] om F aunho e FOKUS Be lin, es ablishes a axonomy o oles ha a e assumed by he di e en nodes wi hou dis inguishing be ween node ypes. Each node, usually a so wa e se ice, p o ides s anda d in e aces o a a ie y o asks. This plain a chi ec u e is e y ubiqui ous-compu ing-o ien ed, by gi ing all he componen s he conside a ion o nodes wi h di e en oles, bu wi hou mo e s uc u e. Mundo [33], om Da ms ad Uni e si y, es ablishes a s uc u ed node classifica ion cen ed on he scope o he node. This p ojec in o- duces majo aspec s, such as he communica ion and he associa ion o nodes, bu i emains a low-s uc u ed node classifica ion mo e han an a chi ec u e, om a componen -le el poin o iew. MundoCo e [27] communica ion middlewa e designed o he equi emen s o pe asi e compu ing is also de eloped in Da ms ad Uni e si y. I is low-le el so wa e o ien ed, as opposed o ou p oposal ha desc ibes a highe - le el so wa e a chi ec u e. Ga o Tech Sma House [31] om he Uni e si y o Flo ida, p oposes a laye ed Re e ence A chi ec u e, di ided in o ou laye s: applica ion, se ice, node, and physical. Each laye includes a se o sublaye s and componen s ha o m he whole a chi ec u e. I also in- cludes OSGi as a solu ion o he managemen o de ices in he se ice laye . This is an a chi ec u e e y simila o ou app oach, al hough we pu sue an a chi ec u e o a simple and mo e abs ac na u e. The main di e ence, be ween hei p oposal and ou s, is ha Ga o a chi ec- u e does no explici ly include an ac ing laye and does no define a cycle. Mo eo e Ga o a chi ec u e does no define asks o hei Senso laye , meanwhile ou p oposal defines fi e main asks o i . 930 A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 GAS-OS [17] is he closes app oxima ion o he so wa e Re e ence A chi ec u e p oposed in his pape . This so wa e implemen s he Gadge wa e A chi ec u al S yle (GAS) app oach, in which people configu e complex collec ions o in e ac ing eGadge s, in a simila way o ha o a sys em builde in designing a so wa e sys em and compo- nen s. The benefi s o he ubicomp a e he same as hose demons a ed by so wa e enginee ing: encapsula ion and composi ion. As can be concluded om a de ailed analysis o hese p ojec s, hei a chi ec u es emain he e ogeneous, al hough se e al common cha ac- e is ics can be iden ified such as: gene al Pe cep ion-Reasoning-Ac ing cycle (commonly men ioned and explained in he bibliog aphy, o example by Russel and No ig [25]), o a ce ain kind o in eg a ion and in e ac ion be ween de ices ( he OSGi amewo k is equen ), o he e maybe ea men o con ex - ela ed in o ma ion. The p ojec s usually ake a laye ed app oach, bu ail o ollow a common s uc u e, and hei laye s, componen s and/o de ices di e g ea ly and a e also e y a chi ec u e-dependen . The e olu ion o hese sys ems by in eg a ing new de ices, algo- i hms o me hodologies ha imp o e o pe o m new asks in ol es a complex ask, due o he low le el o encapsula ion and modula iza ion. So wa e design is a key ea u e in he o ganiza ion, in eg a ion, and scalabili y o hese nume ous componen s. This p oblem is one o he p ime easons why SEs ha e ye o e ol e, and why hey belong mo e o he wo ld o esea ch and academia han o indus ial solu ions. Sol ing his pushed us o p opose he ollowing So wa e Re e ence A chi ec u e ha o ganizes hose common asks which need o be accomplished in he design o a SE om a concep ual poin o iew, independen o he implemen a ion. This necessi y has al eady been indica ed by Muhlhause and Gu e ych [26]. 3. Re e ence A chi ec u e 3.1. In oduc ion This sec ion explains he p oposal o So wa e Re e ence A chi ec- u e o de elop he so wa e laye in SEs. The p oposal is based on he goals o ubiqui ous compu ing p oposed by Weise [2]. Taking his as a s a ing poin , au oma ion in SEs can be o ganized as a con inuous in e ac ion be ween h ee main p ocesses: a) pe cep ion, b) easoning and c) ac ing (see Fig. 2). As indica ed in he In oduc ion, hiswo kiscen edon hePe cep- ion p ocess o he Re e ence A chi ec u e, which is explained below. 3.2. Pe cep ion The Pe cep ion p ocess should be di ided in o de o spli complex pe cep ion ac i i y in o se e al a ainable asks. The esul o hese asks mus be an accu a e pe cep ion o he eal wo ld (see Fig. 3). Pe cep ion has o deal wi h low-le el de ails o e ie e da a om eal wo ld and o adap i o a knowledge base, which mus ag ee wi h he on ology o he SE. A simple in o ma ion model o he SE used in he p o o ype scena io is p oposed in Table 1. The p ocess has o clea his e ie ed da a o any e oneous, insignifican , and edundan alues in o de o build an accu a e ep esen a ion o he eal wo ld, as equi ed by he ollowing p ocess. Some o hese asks ha e common ea u es wi h o dina y p e- p ocessing o da a such as no maliza ion, adding a ibu es o eplacing missing alues. P e-p ocessing o da a o SE has been s udied by se e al au ho s. S anko ski and T nkoczy's [34] p oposal defines a able om he da a collec ed in o de o gene a e inpu s o ollowing p ocessing, bu does no co e e o de ec ion, epa a ion o he de ices which pe o m his p e-p ocessing. On he o he hand, Elnah awy [35] p oposes wo gene al p ocesses •cleaning da a (conside ing cleaning a senso le el o cleaning a da abase le el) by applying p obabilis ic unce ain y models and •que ying da a. Mo eo e Wu and Clemen s-C oome [36] men ion a Da a p epa a ion s ep whe e da a mine s c ea e ele an subse s bu do no lis scopes, ep- a a ion o e o de ec ion p oposals, like Zhang [37] which jus men ion ha a p e-p ocessing s ep is equi ed o in eg a ion o low-le el senso da a. The au ho 's app oach includes concep s p esen ed by p e ious p oposals and gene alizes hem. 3.2.1. Da a collec o This is he lowes -le el ask, and i s aim is o e ie e da a om phys- ical de ices wi hin he SE. The Da a Collec o usually has o deal wi h ga eway de ices o e e y ype o senso echnique deployed in he SE. Da a can be gene a ed by nume ous kinds o senso s such as empe - a u e, p essu e, op ical, acous ic, mechanical, mo ion, ib a ion, flow, posi ion, elec omagne ic, chemical, humidi y, and adia ion, and he e- o e a c ucial ques ion ha mus be add essed conce ning he ask o he Da a Collec o is ha o he unifica ion o da a ypes. In ac , only a small subse o he en i onmen p ope ies (Table 1)is necessa y o pe o m a pa icula applica ion o au oma ion p ocess (e.g. swi ching ligh s o when nobody is a home does no need condi- ioning in o ma ion om he en i onmen ). Simila ly, senso s mus be deployed in an o ganized manne in o de o p e en he p ocessing o useless in o ma ion. Rega ding he execu ion o he da a p ocessing, i is impo an o conside whe e he p ocessing should be pe o med since a numbe o de ices allow in e nal p og amming while o he s a e p e-p og ammed. Fig. 2. The cycle o he au oma ion p ocess in a sma en i onmen . Fig. 3. Tasks o he pe cep ion p ocess. 931A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 The p ocessing can be dis ibu ed, cen alized, o e en a combina ion o he wo i bo h kinds o de ices a e p esen in he en i onmen . Since he majo i y o de ices a e p e-p og ammed and canno ex end hei basic unc ionali y (e.g. X-10 mo ion senso s), he design o he SE has o be sui ably adap ed and de elope s ha e o adap o hem. On he o he hand, a g owing numbe o de ices ha e now he abili y o ex end o modi y hei pe o mance (e.g. Sen illa Tmo es). P og ammable de ices a e much mo e flexible, and he e o e pose a g ea e challenge o he designe , since hey can be adap ed o cu en needs o a specific applica ion o ci cums ance. In his case, he ask o he Da a Collec o mus in ol e he use o daemons de eloped o e ie e in o ma ion, and also small applica ions unning on de ices, and hence bo h elemen s ha e o ag ee on wha in o ma ion is sen , he pe iodici y o hese eques s, and so on. 3.2.2. Ve ifie The main pu pose o he Ve ifie ask, as i s name implies, is o e i y ha he da a is being ecei ed by he Da a Collec o co ec ly. Howe e he challenge he e is how he Ve ifie can de e mine whe he he da a is co ec o no . The e is no unique solu ion o all possible en i onmen s, so he e ifica ion has o be adap ed o each en i onmen by aking in o accoun he on ology used. In ou p oposal, Ve ifie main ains a ule engine whe e e ifica ion ules can be deployed, modified and checked in o de o de e mine whe e da a is igh o w ong o he cu en en i onmen . The ule engine has o o e p og amme s a flexible way o add, modi y and dele e ules, and he e o e hese should be human eadable. The se o ules o he Ve ifie should e ol e o e ime, as en i onmen changes a e p oduced. This ask has o wo k side by side wi h he Repai e when inco ec da a is ecei ed in o de o fix in alid da a. The mechanism o commu- nica ion be ween he Ve ifie and he Repai e mus be de e mined. Typical implemen a ions o his mechanism include he publica ion o a epai se ice by he Repai e ha is in oked by he Ve ifie . I can be concluded ha he Ve ifie can be seen as a fil e applied o e all he da a ecei ed, and i can be used o ejec da a o any eason (inco ec , edundan …) by using he a o emen ioned ule engine and based on he on ology model p oposed in Sec ion 3.2.5. 3.2.3. Repai e The ask o he Repai e is o fix inco ec da a de ec ed by he Ve ifie . The epai applied o he da a mus always conside he defined on ology and can be pe o med in a wide a ie y o ways, such as: •Igno e da a. The fi s op ion is o igno e he da a, which means se ing i wi h an unknown alue. The On ologize sa es his alueas equi edby he s o age sys em (e.g. he Weka-a ‘?’cha ac e , o he SQL null alue). •Adjus da a. I a alue is inco ec , bu i s dis ance om a co ec alue is less han a p e iously specified h eshold amoun , hen he alue could be adjus ed o he nea es co ec alue. •Replace da a. Ano he op ion in ol es eplacing da a wi h p e iously co ec da a (e.g. i empe a u e senso e u ns 100 °C, and p e ious da a is 25, hen he cu en alue can be eplaced wi h 25). •Rejec da a. I cu en alues a e no sui able o epai , hen he Repai e will ejec hem. Once a se o da a has been ecei ed, e ified and epai ed, i has o be sen o he On ologize o be o ganized and s o ed. 3.2.4. Fil e Some imes, no all he da a ecei ed om he en i onmen is neces- sa y o he easoning p ocess, and hence he on ology dismisses his in- o ma ion. This is he objec i e o he fil e : o p e en his supe fluous in o ma ion om being sen o he On ologize . The implemen a ion o his fil e , i p esen , should be based on he on ology defined o he ep esen a ion o he en i onmen . 3.2.5. On ologize An on ology ep esen s he knowledge abou he wo ld (o en i on- men ) as a se o classes, p ope ies and ela ionships, wi hin a domain. The easoning is pe o med using he en i ies ep esen ed, and hence da a e ie ed om he en i onmen should be o ganized be o e apply- ing a ificial in elligence echniques in o de o ha e a solid knowledge base wi h which o wo k. The main goal o he On ologize is o o ganize, homogenize, synch onize and agg ega e da a o o m a model o he eal wo ld suppo ed by he on ology defined o he SE. I is necessa ily an impo an e o o building a model o a SE which p o ides da a in e ope abili y and makes possible o ealize in- e ence, as sugges ed by Nucci [29] whe e an on ology amewo k is used o desc ibe all ele an in o ma ion o he en i onmen : de ices, se ices and con ex . I also ackled he u he di ficul y o de ice man- u ac u e s because o he lack o s anda diza ion in seman ic echnolo- gies wi hin hese scena ios. Ene gy is ano he key aspec o ake in o accoun o he on ology as p oposed by Kofle [30], whe e on ology includes in o ma ion no only ela ed o he en i onmen , bu also abou ene gy supply and p o ide . Some o he s udies like Cook [3],Das[21] and Li [22] ha e helped in he composi ion o he abs ac model p oposed, which has been a anged in o ou main ca ego ies as explained in Table 1.Thismodel can help de elope s asce ain he main en i ies ha need o be moni- o ed in he en i onmen . De ice ela ed —This ca ego y is he mos ob ious, and i is ela ed wi h he main elemen s in a SE. Ambien in elligence algo i hms should be awa e o he ollowing main fields: •S a us. Algo i hms mus know he cu en s a es o de ices ins alled in he SE. Ob iously his is essen ial o hese algo i hms, and one o he p ime ac o s o building o u u e p edic ions. Ene gy awa e algo i hms may also need in o ma ion abou ene gy needed by hese de ices o ope a e in o de o apply any ene gy sa ing policy. •Loca ion. De ices usually emain a a loca ion o a long ime, and hence his in o ma ion can be used by ambien in elligence algo i hms. The model mus also be able o handle mobile de ices, such as mo o ized cleane obo s. Table 1 P oposal o modelling sma en i onmen s. Ca ego ies Fields Desc ip ion Example De ice- ela ed S a us Cu en s a e o he de ices Senso empe a u e measu es 25 °C. Loca ion Whe e he de ices a e Cleane obo is in he li ing oom. Inhabi an - ela ed Pe sonal da a Name, age, sex Diane is 45 y.o. Loca ion Whe e he inhabi an s a e Ma k is in he bed oom. Physical s a e Illness, inju ies, and condi ion o inhabi an s Roy has a cold. Men al s a e Psychological s a e o he inhabi an s Da id is dep essed. En i onmen - ela ed Da e, ime Tempo a y in o ma ion Cu en ime is 13:36. En i onmen al condi ions The phenomena ha a e cu en ly occu ing in he a mosphe e I is ainy. Home backg ound Ine en i y loca ion Whe e hese en i ies a e The so a is in he li ing oom. Home limi s and p ope ies The p ope ies o he home s uc u e and i s limi s Bed oom window opaci y is 70%. 932 A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 Inhabi an - ela ed —SE algo i hms mus be awa e o he inhabi an s' s a us o o e app op ia e p edic ions o any use o o he whole g oup o inhabi an s. Along his line, se e al o he necessa y fields o in e inhabi an -awa e p edic ions a e discussed: •Pe sonal da a. This field includes all he da a conce ning a pa icula pe son, such as name, age, and gende . •Loca ion. Inhabi an s can mo e be ween di e en spaces, so SE sys ems should be able o iden i y and loca e each inhabi an . •Physical s a e. This field is ela ed wi h he illnesses and inju ies ha an inhabi an can su e . SE echnologies mus adap o hese si ua- ions and o e app op ia e esponses. •Men al s a e. The s a e o mind o a pe son can be defined as he em- po a y psychological s a e. The beha iou o a dep essed inhabi an usually di e s om ha o a eupho ic inhabi an , and hence SEs mus be consis en wi h hese ci cums ances. En i onmen - ela ed —This ca ego y is p obably he mos di use since i co e s he e ogeneous and di ficul - o-limi fields, as discussed in he ollowing lis : •Da e, ime, season. Ob iously SE beha iou di e s unde each empo- al condi ion. Fo example, he ai condi ioning policy is al e ed be ween summe and win e . •En i onmen al condi ions. This field is comp ised o cu en en i on- men al condi ions (sunny, cloudy, ainy, among o he s). A SE should also eques a wea he o ecas , which could be significan in he assessmen o u u e decisions. Home backg ound —This ca ego y mus con ain all he ele an i ems ega ding ine en i ies and hei p ope ies and quali ies. This ca ego y is he leas ele an discussed, bu could emain significan in ce ain specific applica ions. Two ela ed fields a e p oposed in he ollowing lis ing: •Fu ni u e loca ion and posi ion. Fu ni u e occupies space a home and can be mo ed. Loca ion ( oom whe e he u ni u e is loca ed) and posi ion (place wi hin he oom) should be egis e ed by he sma home sys ems since i could be use ul in specific applica ions, such as obo mo emen - ela ed algo i hms, and p esence de ec ion- ela ed algo i hms. •Home limi s and p ope ies. The ex u e o a floo , he colou o a wall, and he opaci y o he windows could be significan in specificcases, such as empe a u e-adjus men applica ions. The e a e ye wo mo e issues conce ning he On ologize :Synch o- niza ion and Agg ega ion. Synch oniza ion —The On ologize has o synch onize da a om a wo ld ull o asynch onous de ices, and e en s. Response ime cons i u es a majo ac o when easoning abou e en s. Au oma ion applica ions usually need se s o da a composed o alues om mul iple de ices, cap u ed a a ious momen s. Da a om a a ie y o de ices mus be syn- ch onized o i s la e agg ega ion, and hence his ask has o define he logic o synch onize alues om mul iple and e y he e ogeneous sou ces. Implemen a ions o he synch oniza ion p ocess a y depending on he goals o each specific sma applica ion and on i s ype o da a. Howe e all implemen a ions sha e ce ain common elemen s such as: •Da a bu e , which s o es ecei ed da a ha is wai ing o be pai ed wi h o he da a. •Ga bage collec o , which supe ises he size o he bu e , and pe iod- ically cleans he bu e o da a ha canno be pai ed. Agg ega ion —Once da a is synch onized, i is agg ega ed in a se o da a o a specific sma applica ion ha con o ms o he SE model. When da a is agg ega ed, i is eady o be s o ed in he knowledge base ha eeds he easoning asks and lea ning p ocess. The knowledge base o ma can ake he o m o any o he de ac o s anda ds, such as a (Weka so wa e o ma file o inpu da a), o ha o a ela ional da abase. 3.3. De ice Abs ac ion The e is a gap ha needs o be co e ed be ween he modules defined by he Re e ence A chi ec u e and he physical de ices. One in e es ing ini ia i e which sol es his p oblem is De ice Abs ac ion.Thisis he esul o p e ious wo k by au ho s wi hin he OSAmI p ojec [23]. In De ice Abs ac ion a de ice abs ac ion laye is p o ided which desc ibes a se o s anda ds and con en ions o con olling, configu ing and accessing he da a gene a ed om all kinds o de ices ela ed o Ambien In elligence. The in eg a ion o senso s and ac ua o s is p o id- ed by ollowing he De ice Abs ac ion model in hei con ol so wa e. This hie a chical model, shown in Figs. 4 and 5,unifies c i e ia in o de o acili a e he access o he de ices, hei unc ionali y, and hei gene a ed da a. In his way, he me hods used a e independen o he unde lying p o ocols, and allow easie de ice swi ching. The use o De ice Abs ac ion is a s ep owa ds s anda diza ion o SEs. The p ima y ole o De ice Abs ac ion is he classifica ion o de ices in senso s and ac ua o s. Senso s a e ca ego ized as ei he me e s o de- ec o s, while ac ua o s depend on hei unc ionali y as pulse, swi ch, dimme and mo emen . Each de ice ca ego y p o ides a se o specific me hods o con ol o he de ice ac ions. The e is s ill ano he so wa e a e ac be ween De ice Abs ac ion and he de ices: he API which ansla es high-le el me hods in oked in p o ocol-dependen eques s. These APIs a e usually p o ided by de ice manu ac u e s. 3.4. Communica ions be ween p ocesses Reasoningp ocessesinSEscanbesepa a ed in o se e al asks which in e ac o achie e h ee main goals: a) o lea n, b) o eason, and c) o p edic . Finally, in o de o close he ci cle, SEs mus ac au oma ically o achie e a specific sma applica ion. This is he main pu pose o he Ac ing p ocess. The decisions and specific asks o de ed by he Reasoning p o- cess, ha e o pass h ough h ee main askmas e s: a) policy manage , b) ask schedule , and c) ask unne . In o de o eed bo h Reasoning and Ac ing p ocesses, an e en - d i en a chi ec u e pa adigm (EDA) is p oposed, in he o m o a publishing-subsc ibe message sys em. Reasoning asks a e subsc ibe s o da a gene a ed by Pe cep ion p ocesses and Reasoning asks a e publishe s o in e ed knowledge. In his way, Ac ing asks become subsc ibe s o he knowledge gene a ed by he Reasoning p ocess. 4. Ve ifica ion wi h p o o ype An example o he Pe cep ion p ocess implemen a ion, which ollows he Re e ence A chi ec u e p oposed in his pape , is p esen ed in his sec ion as p oo o i s use ulness. 4.1. P o o ype o e iew The main objec i e o he applica ion is he pe cep ion o a sma o - fice. I e ie es da a on he localiza ion o wo ke s, and on luminosi y, empe a u e, and humidi y (see Fig. 6). The pu pose o he e ie al o his in o ma ion is o acqui e knowledge abou inhabi an s/wo ke s e- ga ding hei habi s wi h espec o ligh ing condi ions, empe a u e, e c. in o de o make sma use o a ificial ligh ing o ene gy sa ing pu poses. The ha dwa e o he de elopmen en i onmen is composed o : •a compu e , which ac s as he da a ecei e ga eway, • h ee Sen illa Tmo es ( wo in he o fice as shown in Fig. 6 and a hi d in ano he o fice), as senso de ices o he collec ion o da a on he quan i y o ligh , empe a u e, and humidi y, •an X10 mo ion senso in o de o de e mine whe he he o fice is occupied, 933A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 Fig. 4. De ice Abs ac ion class model o sensing. 934 A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 Fig. 5. Ac ua ion class model o De ice Abs ac ion. 935A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940 •IEEE 802.15.4 (ZigBee) p o ocol b idge connec ed o he compu e in o de o communica e wi h he Sen illa Tmo es, •X10 anscei e o communica e wi h ac ua o s and he mo ion senso . The so wa e o he de elopmen en i onmen is composed o : Sen illa Wo k, which is an Eclipse-based IDE o he c ea ion, deploy- men , and debugging o ubiqui ous applica ions; Weka, a ool sui e which acili a es he use o machine lea ning echniques; and he de el- oped so wa e based on he p oposed Re e ence A chi ec u e and on he De ice Abs ac ion Model as he so wa e pa adigm o he de ice managemen so wa e. 4.2. P o o ype implemen a ion The so wa e de eloped by he au ho s ollows he Re e ence A chi- ec u e ou lined in Sec ion 3 and is ocused on he Pe cep ion p ocess de ailed in Sec ion 3.2. This so wa e he e o e ollows he asks p oposed in he Pe cep ion p ocess and is esponsible o he low- le el in e ac ion wi h de ices (Da a Collec o ask), in he o m o e i- fica ion, epai and fil e ing o he da a, and o he s o age o he da a by agg ega ing se e al da a sou ces and by ollowing he De ice Abs ac- ion Model. The so wa e i sel is dis ibu ed be ween a ious de ices, and hence a numbe o hese asks a e pe o med by he cen al compu e while o he s a e pe o med by he Sen illa Tmo es. 4.2.1. Sen illa Tmo es so wa e Sen illa Tmo es implemen a Ja a Vi ual Machine (JVM) called Sen illa Poin so ha i can un Ja a applica ions. In o de o access he ha dwa e capabili ies o he de ice, Sen illa o e s a Ja a lib a y which p o ides access o he da a ga he ed by he senso s and o he elemen s such as ex ension po s and leads. This low-le el so wa e cons i u es he de ice API which is used by he so wa e o access de ices. The Ja a applica ion un by Tmo es accesses senso da a and ansmi i ia he Zigbee in e ace. The senso s o m a mesh ne wo k whe e mo es ac as epea e s. This ype o ne wo k makes i possible o co e wide a eas e en hough Zigbee p o ocol has a adio scope o a me e 10 m. When de eloping his applica ion, i was obse ed ha he quan i y o luminosi y (measu ed in luxes) cap u ed by i s pho osyn he ically ac i e adia ion (PAR) senso fluc ua ed i he fluo escen ligh o he o fice was le swi ched on. The eason o his beha iou is ha fluo es- cen ligh is cons an ly swi ching o and on bu i is no pe cep ible by he human eye due o i s high equency. Howe e , his beha iou posed a p oblem wi h he so wa e, so i had o be ackled and included as a pa o he Repai e ask (see Sec ion 3.2.3)a hispoin .The solu ion is qui e simple: ins ead o e ie ing jus one alue, n alues a e e ie ed and hei a e age is compu ed and hen sen o he cen al ga eway. 4.2.2. Mo e Dashboa d In o de o ecei e all he in o ma ion om he Sen illa mo es, an appli- ca ion has been de eloped, which implemen s he app op ia e in e aces in each case: Mo ionDe ec o , Tempe a u eSenso , B igh nessSenso , o Humidi ySenso . Mo eo e , he au ho s ha e de eloped so wa e o he cen al s a- ion called Mo e Dashboa d. This so wa e shows he in o ma ion ha is being ecei ed om he mo es in eal ime. This applica ion ca ies ou all he main asks o he Pe cep ion p ocess in Sec ion 3.2.Asumma- y o he classes and in e aces o he implemen a ion is shown in Fig. 7: 1. Da a collec o . This p ocess is esponsible o managing he ecep ion o da a om he mo es, and o using he mo e ga eway supplied wi h he de elopmen ki , and an X10 con olle de eloped by he au ho s. 2. Ve ifie . Th ee simple e ifie s ha e been de eloped and pe o m sim- ple es s, simila o p econdi ions, o e da a ecei ed om TSR-PAR luminosi y, humidi y, and empe a u e senso s (e.g. luminosi y ≥0). Fig. 6. Room se up. 936 A. Fe nández-Mon es e al. / Compu e S anda ds & In e aces 36 (2014) 928–940