Int. J. Sensor Networks, Vol. 2, Nos. 5/6,

Transcription

1 In. J. Sensor Neworks, Vol. 2, Nos. 5/6, Design and applicaion of enhanced communicaion proocols for wireless sensor neworks operaing in environmenal monioring F. Chii, M. Ciabai, G. Collodi, D. Di Palma and Romano Fanacci* Diparimeno di Eleronica e elecomunicazioni, Universiá degli Sudi di Firenze, Ialy chii@lens.de.unifi.i michele.ciabai@unifi.i collodi@ing.unifi.i davide.dipalma@unifi.i romano.fanacci@unifi.i *Corresponding auhor A. Manes Diparimeno di Energeica, Universiá degli Sudi di Firenze, Ialy anonio.manes@unifi.i Absrac: he effecive applicaion of Wireless Sensor Neworks (WSNs) o wide area environmenal monioring currenly represens a challenge for his echnology. Since devices are usually energy consrained, smar power saving procedures are needed o increase he nework lifeime. A common approach is o swich off he node s radio; his definiely requires he adopion of synchronisaion procedures ha properly schedule packes ransmission o avoid boh overhearing effecs and collisions. o his end, a novel class of MAC layer proocols, named SAR MAC, is presened. his is furher improved by inroducing SAR+ scheme and properly inegraing wihin a rouing soluion, according o he cross-layer design. I allows he opimisaion of energeic consumpion, or equivalenly, nework life-ime, ogeher wih minimising he signalling overhead. he proposed soluion is hen applied o a realisic user-defined scenario, highlighing remarkable advanages boh in erms of robusness, self reconfigurabiliy, cos, complexiy and QoS, hus validaing he proposed soluion. Keywords: wireless sensor neworks; WSNs; MAC proocols; energy efficiency; synchronisaion. Reference o his paper should be made as follows: Chii, F., Ciabai, M., Collodi, G., Di Palma, D., Fanacci, R. and Manes, A. (27) Design and applicaion of enhanced communicaion proocols for wireless sensor neworks operaing in environmenal monioring, In. J. Sensor Neworks, Vol. 2, Nos. 5/6, pp Biographical noes: Francesco Chii (M ) received he elecommunicaions Engineering and he PhD in Informaics and elecommunicaions Engineering degrees from he Universiy of Florence in 2 and 24. His curren research opics are devoed o MAC, LLC and nework layers proocols for boh public and privae wireless communicaions sysems and ad hoc and sensor neworks as well. He ook par in several European research projecs as IP GoodFood, he NoEs Nexway, SaNex, Newcom and CRUISE, he GJU WIS, he ESI SF179 and he COS 289 acion. Michele Ciabai received he Laurea degree in elecommunicaion Engineering from he Universiy of Florence Ialy, in 24, wih a hesis on performance and error analysys of waer seam reconsrucion algorihms hrough acive measures in LIMB mode. Since January 25, he is a PhD suden on Wireless Sensor Nework Communicaion proocols. Giovanni Collodi received he Laurea and he PhD. degrees in Elecronic Engineering from he Universiy of Florence, Florence, Ialy, in 1996 and 2, respecively. Since 1996, he joined he Deparmen of Elecronics Engineering, Universiy of Florence, where he became an Assisan Professor in 24. During his academic career, he has been Lecurer in courses of Applied Elecronics, and now he eaches Inegraed Circui for Wirelss Applicaion. His research Copyrigh 27 Inderscience Enerprises Ld.

2 3 F. Chii e al. aciviies have covered he sudy of acive and passive compac srucures for MMICs, he design of mulifuncion MMICs for low-power wireless applicaions, CAD and numerical modelling. Currenly, he is working on applied researches on wireless sensor nework and heir applicaions on environmenal monioring. Davide Di Palma received a Laurea degree in Biomedical Elecronic Engineering from he Universiy of Florence, Florence, Ialy in July 24. In 25, he joined he MicLab: currenly, he is a PhD suden and his main research aciviy is relaed o he sudy and he realisaion of wireless sensors neworks, using new proocols and differen boards. Romano Fanacci, received he Dr Ing degree in Elecronics Engineering from he Universiy of Florence, Florence, Ialy in 1982 and a PhD in elecommunicaion Engineering in In 1982, he joined in he Deparmen of Elecronics and elecommunicaions in he Universiy of Florence, firs as Researcher, hen as an Associae Professor, and currenly as Full-ime Professor. His curren research ineress are digial communicaions, compuer communicaions, queuing heory, saellie communicaion sysems, wireless broadband communicaion neworks, ad hoc and sensor neworks. He has been involved in several naional, European and naionwide research projecs. He is IEEE Fellow Member and akes par in several IEEE COMSOC Commiees ogeher wih being Gues Edior of IEEE COMSOC journal special issues and Member of he Ediorial Board of several inernaional journals. Anonio Manes afer receiving his degree in elecommunicaion Engineering from he Universiy of Florence in 22, he sared a PhD course on Energeic and Indusrial echnologies, compleed in 25 wih a hesis on Wireless Sensor Neworks for Indusrial Monioring and Conrol. Sill working a he Deparmen of Energeics, he is involved in EU founded research projecs of he 6h FP; acually his research is oriened on developmen of low power wireless devices and proocols for indusrial applicaions. 1 Inroducion In recen years, he availabiliy of on-field monioring became a key issue for he assessmen of environmenal processes and parameers. In his conex, he applicaion of Wireless Sensor Neworks (WSNs) echnology represens a significan advance over radiional invasive mehods of monioring as pioneering envisioned in Akyildiz e al. (22) and applied o several fields in Hać (23). As a maer of fac, insrumening naural spaces wih neworked microsensors migh enable long-erm daa collecion wih enhanced accuraeness. Furher, he communicaion capabiliies allow nodes o cooperae in performing more complex asks, like saisical sampling or daa aggregaion, no feasible wih a poin-o-poin elemery. Finally, he compuing and neworking capabiliies permis boh he reprogramming or even he self-reconfiguring of he whole sysem. he increasing need for producs conrolling a differen criical seps of he agrofood chain open a novel applicaion field for WSN. Among he relaed iniiaives founded by public insiuions, he GoodFood Inegraed Projec (hp:// presened wihin he Informaion Sociey echnology (IS) hemaic area of European Communiy Sixh Frame Projec (EU VI FP), aims a developing he new generaion of analyical mehods based on Micro and Nanoechnology (M&N) soluions for he safey and qualiy assurance along he food chain in he agrofood indusry. GoodFood approach will comply, hrough he developmen of innovaive M&N soluions, wih he needs of ubiquiy, low cos and low power, fas response, simple use and fully inerconnecion o he decisional bodies. he projec s WorkPackage 7, aimed a implemening inelligen and innovaive communicaion soluions for food safey and qualiy raceabiliy, is developing a Pilo Sie, oriened o he wine producion phase monioring; his Pilo Sie is being deployed a he Monepaldi Farm (in he Chiani region, Ialy) and is expeced o be fully operaive by he end of 25. he requiremens ha WSN adopion is expeced o saisfy in performing an effecive agrofood monioring are mainly concerned wih he operaing period. In paricular, a large number of baery operaed nodes, deployed in oudoor environmens, is expeced o operae a leas for a farming season (usually 3 6 monhs), wihou mainenance. Besides, since node are usually unaended, hey migh cope wih node fails or nework congesion by means of adaping or even self-reconfiguring heir funcionaliies and proocols. Finally, correc daa delivering o remoe user via gaeway is o be guaraneed o allow a proper monioring. he aforemenioned requiremens compels he WSN communicaions proocols o be carefully designed and opimised. In his paper, a cross-layer soluion is presened, joining he Medium Access Conrol (MAC) and nework layers operaions and applying i o a pracical case sudy, allowing o develop self configuring, long life-ime and effecive nework infrasrucure and o saisfy user-oriened specificaions. he MAC layer proocol feaures are described in Secion 2 ogeher wih he performance analysis as o he nework synchronisaion capabiliies. he proocol has been analyically characerised and generalised o he case of compleely independen nodes in erms of frame period and duy cycle values. Furher, i is

3 Design and applicaion of enhanced communicaion proocols 31 modified in order o cope wih densely deployed neworks, giving rise o he Enhanced Synchronous ransmission Asynchronous Recepion (SAR+) MAC and he provided gain is invesigaed wihin he Appendix. In Secion 2.4 he cross-layer rouing proocol is highlighed and he overall communicaions performance is presened for he pracical es case offered by he GoodFood projec. Finally, some conclusions are drawn explaining he fuure direcions of he presen research aciviy. 2 MAC layer proocol 2.1 Sysem requiremens he mos relevan sysem requiremens leading he design of an efficien MAC proocol for an environmenal monioring WSN, as o GoodFood projec, are mainly concerned wih he power consumpion issues and he capabiliy of quickly seup an end-o-end communicaion infrasrucure ha suppors boh synchronous and asynchronous queries. In paricular, he mos relevan challenge is o make a sysem able o run unaended for a long period, as nodes are expeced o be deployed in zones hardly manageable for mainenance. his calls for an opimal energy managemen, since i is anyhow a limied resource and he failure of a nodes may compromise WSN conneciviy, as he nework ges pariioned. herefore, he MAC layer is o be opimised ensuring ha he energy i spends is direcly relaed o he amoun of raffic ha i handles and no o he overall working ime. Oher imporan properies are scalabiliy and adapabiliy of nework opology, in erms of number of nodes and heir densiy. As a maer of fac, some nodes may eiher be urned off or join he nework aferward or finally, hey could be moved o a differen locaion. hus, he MAC layer proocol migh be able o face hese evens wih an higher degree of responsiveness and reconfigurabiliy. Finally, several user-oriened aribues, including fairness, laency, hroughpu and bandwidh uilisaion, need o be aken ino accoun even if hey could be considered secondary wih respec o our applicaion purposes, since, in conras o ypical WLAN proocols, MAC proocols designed for WSN usually rade-off performance for cos (Langendoen and Halkes, 25). aking he IEEE Disribued Coordinaion Funcion (DCF) (IEEE sandard 82.11, 1999) as a saring poin, several more energy efficien echniques have been proposed in lieraure o avoid excessive power wasing when a node are in he so-called idle mode, ha is, a sae in which i coninuously sample he channels waiing for a message heir circuiry. hey are based on he receiver periodical urning on he radio o deec he presence of a packe preamble and if no signal is sensed he radio is urned-off again unil he nex sample. As a consequence, a low power sae, in which he ransceiver is disabled and only a low level clock is acive, migh be inroduced, duy cycling he radio. Furhermore, he sender can simply wai unil he momen he receiver is abou o sample he channel and send a packe wih an ordinary preamble, hus saving energy boh a he sender and a he receiver. In paricular, according o he WiseMAC proocol (El-Hoiydi e al., 23) nodes mainain he schedule offses of heir neighbours hrough piggy backed informaion on he ACK of he underlying CSMA proocol. However, WiseMAC suffers he disadvanages of being subjec o raffic flucuaions and moreover, o no efficienly managing message broadcasing due o he absence of synchronisaion procedures. Deriving from he classical conenion-based scheme, several proocols (S-MAC (Ye e al., 22), -MAC (van Dam and Langendoen, 23) and DMAC (Lu e al., 24)) have been proposed o address he idle lisening overhead by synchronising he nodes and implemening a duy cycle wihin each slo. A he beginning of a slo, all nodes wakeup and any node wishing o ransmi a message mus conend for he channel. his increases he probabiliy of collision in comparison o he random organisaion of he energy-efficien CSMA proocols. o miigae he increased collision overheads S-MAC and -MAC include an RS/CS handshake, while DMAC avoids his proocol overhead. Finally, hese approaches differ in heir way of deciding when and how o swich back from acive o sleep mode. In paricular, in S-MAC proocol, neighbouring nodes form virual clusers so as o se up a common sleep schedule. If wo neighbouring nodes reside in wo differen virual clusers, hey wake up a he lisen periods of boh clusers. On he oher hand, in -MAC, he lisen period ends when no acivaion even has occurred for a cerain ime hreshold. Finally, DMAC scheme performs as an improved Sloed Aloha algorihm in which slos are assigned o he ses of nodes based on a daa gahering ree. Resoring o he above consideraions, a class of MAC proocols, paricularly suied for a fla nework opology, as ha considered in Goodfood projec (hp://www. goodfood-projec.org), has been derived, aking he benefis of boh WiseMAC and S-MAC schemes. In paricular, i joins he power saving capabiliy, due o he inroducion of a duy cycle, ogeher wih he advanages provided by he offse scheduling, wihou an excessive signaling overhead. I is basically based on periodical rendez vous synchronisaion of all possible couples of nodes by means of MAC layer messages o allow boh communicaions and power saving. Furher, o avoid excessive signalling overhead especially when he number of neighbours becomes larger proocol enhancemen is proposed. More deails abou proocol issues, relaed inyos modules for he implemenaion and he provided advanages are given wihin he nex secion. 2.2 Proposed scheme Synchronous ransmission asynchronous recepion MAC An alernaive approach o he problem of providing an energy efficien MAC soluion is represened by he proposed scheme ha is characerised as i follows. Following El-Hoiydi e al. (23), each node migh be eiher ino an idle mode, in which i remains for a ime inerval l (lisening ime) or in an energy saving sleeping sae for a s (sleeping ime). he ransiions beween saes are synchronous wih a period called frame equal o f = l + s pariioned in wo subinervals; as a

4 32 F. Chii e al. Figure 1 SAR MAC characerisaion for four differen nodes consequence, a duy cycle funcion can be also inroduced: Figure 2 Phase evaluaion for wo differen cases d = l l + s (1) X τ Φ 1 X s l hough he proocol is able o manage a specific duy cycle for each node, i has been assumed o be he same for all he nodes as so i usually happens. o provide full communicaion capabiliies o he nework, all he nodes need o be weakly synchronised or equivalenly, o be aware a leas of he awakening ime of all heir neighbours. o his end, a node sends frame by frame one synchronisaion message o each of is neighbour nodes known o be in he lisening mode (synchronous ransmission), as explained in Figure 1. Insead, during he se up phase in which each node is discovering he nework opology, he conrol messages are asynchronously broadcased. On he oher hand, is neighbours periodically awake and ener he lisening sae independenly (asynchronous recepion), as i is again presened in Figure 1. he header of he synchronisaion message conains he following fields: a node unique idenifier, he message sequence number and he phase, ha is, he ime inerval afer which he sender eners a new frame period claiming o be in he lisening saus waiing for boh he synchronisaion and daa messages from is neighbours. he phase φ is evaluaed according o he Figure 2 as i follows: φ 1 = τ + S (2) if he node is in he sleeping mode, where τ is he ime remaining o he end of he acual lisening period (or he nex one if he node is in he sleeping sae). Conversely, if he moe is in he lisening saus, φ is calculaed as: φ 2 = τ l (3) In order o fully characerise he performance of he Synchronous ransmission Asynchronous Recepion (SAR) MAC approach, energeic consumpion needs o be analyically described. According o our approach, i has been evaluaed in mah o be in compliance wih he power baery supply. 1 f he energy spen for SAR MAC operaion (C) wihin one frame period can bed characerised as i follows: C = c Rx d f + c sleep [ f (1 d) N pk ] + NCx Φ 2 τ l [mah] (4) where c sleep and c Rx represen he sleeping and he receiving coss [ma] and C x is he single packe ransmission coss [mah], pk is he synchronisaion packe ime lengh [s] and finally N is he number of neighbours. In (4) only f and d have been menioned as hey are he parameers ha could be varied in pracice. In effec, he previous equaion migh be generalised as he erm C x is differen if a node aemps o ransmi a packe while i is in he sleeping sae. In his case, i firs urns on he radio, hen performs a channel sensing whose duraion is relaed o he adoped MAC policy, ransmis he packe and finally i urn off he radio. he mos predominan facor is clearly he cos of he channel sensing ha usually akes wice he packe duraion. When he following inequaliy hold: hen: N pk f C = c Rx d f + c sleep f (1 d) + NC x [mah] (5) he proocol cos normalised o he synchronisaion ime is finally: C = c Rx d + c sleep (1 d) + NC x [ma] (6) f f As highlighed in able 1, i usually happens ha c x c sleep c Rx, in where c x = C x / pk and

5 N Design and applicaion of enhanced communicaion proocols 33 pk is he packe ransmission ime [s], assumed equal o 1 ms as wors case. his means ha he major conribuion o he overall cos is represened by he lisening period ha he SAR MAC proocol ries o suiably minimise. Neverheless, an advanced approach is presened below, aiming a minimising also he cos of packe ransmission for densely deployed WSN or wih a high raffic load facor. able 1 Parameer c Rx c sleep C x c x Power consumpion values Enhanced SAR (SAR+) MAC Value =1 ma =.1 ma =3 mah =.1 ma his approach inroduces an improvemen in he SAR MAC as only one synchronisaion packe is mulicased o all he neighbour nodes belonging o a subse, ha is, such ha hey are joinly awake for a ime inerval greaer han Rx. his leads o an addiional advanage, as he number of neighbours increases allowing beer performance wih respec o scalabiliy and a power saving oo. Besides, he synchronisaion overhead is reduced wih a consequen collisions lowering if no carrier sensing is performed. However, in he proposed approach, a Carrier Sensing Muliple Access (CSMA) scheme is adoped, allowing a benefi in erms of energy consumpion and laency lowering. Under his hypohesis and moreover, supposing ha he number of subses is K, hen he normalised cos migh be expressed as: C = c Rx d + c sleep (1 d) + KC x [ma] (7) f f Since K N, he normalised cos resuls o be remarkably lowered, especially if number of nodes and duy cycle ge higher, even if he laer case is inherenly power consuming. 2.3 SAR MAC proocol performance analysis he evaluaion of he proposed MAC proocol effeciveness migh ake ino accoun wo differen aspecs: firs of all he energeic consumpion or equivalenly, he nework life-ime and also he capabiliy of synchronise he nodes according o applicaion requiremens, ha is in a weak sense. In Figure 3 he normalised cos versus he number of neighbour nodes is shown for he S-MAC and SAR MAC schemes. I is worh noicing ha he performance of he proposed proocol is beer wih respec o he exising approach for a number of neighbour nodes greaer han 7. For a number of neighbour nodes exremely high (no repored in Figure 3), he SAR+ proocol presens he advanage of a addiional power consumpion reducion, as he mulicas effec becomes relevan. However, SAR+ is able o lower he collision probabiliy by reducing he synchronisaion messages. Finally, in Figure 4 he normalised coss of S-MAC and SAR approaches are compared wih respec o he duy cycle duraion for a number of neighbour nodes equal o 8. I is possible o noice ha for d<3.5% he proposed proocol provide a significan gain. As far as he second aspec is concerned, i is worh noicing ha a disribued nework, in which nodes coninuously swich on and off, involves he considerable ask of synchronism mainaining among neighbouring nodes: small hardware olerances ha are negligible in an isolaed device become relevan in long life neworks, where moes have o preserve muual communicaion. In paricular, nodes migh face iming misalignmens (drif) arising from several reasons: differen Finie Sae Machines (FSMs), echnologies and implemenaions for each node or in addiion no equal dynamic behaviours due o differen sensing messages handling. As a resul, wihin a shor period, i is likely he lack of synchronisaion for he nodes o occur. Figure 3 Normalised cos versus number of neighbour nodes s Neighbour Number

6 34 F. Chii e al. Figure 4 Normalised cos versus duy cycle duraion Normalised cos [ma].1 1.% 1.5% 2.% 2.5% 3.% 3.5% 4.% 4.5% 5.% 5.5% 6.% Duy-cycle. SAR MAC proocol effecively overcomes his problem, since i does no rely upon a sric disribued synchronisaion: he nodes are no expeced o behave exacly in he same way for wha ime issues are concerned, bu he capabiliies provided by he communicaion proocol permi o evaluae when and whom a message migh be ransmied o. he dynamic phase esimaion, performed every MAC period according o (2) or (3), allows each node o rack and compensae he ime drifs of is neighbours wihou addiional complexiy. Finally, i is worh noicing ha he randomness of he awakening imes of all he nodes, due o he deploymen procedures, is able o reduce he collision probabiliy over he shared medium. o highligh he benefis yielded by he SAR MAC, wo differen es cases are presened. In he firs case, wo similar nodes wih he same FSM and moreover wih he same f and duy cycle equal o 6 sec and 1%, respecively, are invesigaed. In Figure 5 he ime diagram concerning he values of he phase φ ransmied by each node is presened, poining ou he presence of a phase drif for each node due o a no ideal duy cycle. However, his is upper bounded by he communicaion proocol, since he neighbour node upon he recepion of a synchronisaion message is able o compensae for i by anicipaing or delaying he sending of is nex synchronisaion message o ha neighbour. Besides, his posiive aiude is beer highlighed in Figure 6 where he drif saisics for he nodes are compared. In paricular, i is possible o show by means of numerical processing ha hese approximaely random variables are boh Gaussian wih zero mean and he same variance, hus proving he goodness of our approach. o fully characerise he sysem performance, a scenario comprising wo nodes wih differen behaviours was also invesigaed. In paricular, an ordinary node wih f and duy cycle equal o 6 sec and 1%, respecively and a maser node always in lisening mode (i.e. wih full duy cycle equal o 1%) have been aken ino accoun. In his case, he ransmied phase values show a greaer variaion range, as ransmied phase i is poined ou in Figure 7, since he swiching o he high power sae from he low power sae considerably effecs he accuraeness of he duy cycle. Neverheless, he SAR MAC proocol is again able o limi is negaive impac hrough a compensaion performed by each node every synchronisaion period. In Figure 7 i is possible o noice ha he phase values for he join synchronisaion messages are always compensaed and heir summaion is equal o a consan values, ha is he phase value of he maser node affeced by a unnoiceable drif. his propery is beer explained in Figure 8 where he saisics of he phase drifs for he maser node and he node are depiced. In paricular, i is possible o noice ha he ordinary node is effeced by a remarkable phase drif, whose mean value is abou 5 ms, whils he maser node mean phase drif is on he opposie equal o 5 ms, hus poining ou he ime compensaion provided by he SAR MAC proocol. hese advanages is more eviden by noicing ha a complee phase roaion akes abou 6 sec for an ordinary node (see Figure 7, while, basing on he above menioned mean value, a phase roaion period would be equal o 3 sec if his node was compleely isolaed, ha is, wihou he ime synchronisaion provided by he maser node). Figure 5 ransmied Phase ransmied phase for a couple of similar nodes ime [hh/mm/ss] Phase Node 1 Phase Node 2

7 ransmied Phase Design and applicaion of enhanced communicaion proocols 35 Figure Saisics of he phase drif for similar nodes: (a) node 1 and (b) node 2 Moreover he sysem could beer face several impairmens hroughou locally adaping he mos relevan parameers (i.e. he lisening and he sleeping ime); in paricular, hey could be varied in a ransparen way and he updaed values migh be delivered wihin one frame period. Finally, no sric hardware requiremens in erms of clock accuraeness are necessary. Relaive Frequency Figure 8.1 Saisics of he phase drif for differen nodes: (a) maser node and (b) ordinary node Phase drif values [ms] (a) Relaive Frequency Phase drif values [ms] Phase drif values [ms] (a) Relaive Frequency (b) Relaive Frequency Figure 7 7 ransmied phase for a couple of differen nodes ime [hh/mm/ss] Phase Maser Node Phase Ordinary Node Phase drif values [ms] (b) However, a deep invesigaion on he proposed SAR MAC robusness in he presence of differen operaive condiions need o be addressed. Specifically, wo differen es cases have been carried ou wihin our sysem: 2.4 SAR MAC proocol generalisaion Generally speaking, he proposed SAR MAC scheme belongs o he class of disribued fla approaches, as each node migh deermine is own frame period and duy cycle, independenly from is neighbours. his implies ha he communicaion infrasrucure is se up by means of progressively joining all he nodes wo-by-wo. hus he nework migh be carried ou in a modular way, wihou he need of a cenralised auhoriy coordinaing all he nodes. es 1: several nodes wih differen duy cycles, while he same f, as described in able 2. es 2: several nodes wih differen f, while he same l, as summarised in able 3. Wih reference o he firs scenario, he ransmied phase ime diagram for several couple of nodes wih differen duy cycles have been derived for all he possible cases. In Figure 9, he maser node (d = 1%) and he ordinary node 22 (d = 1%) has been considered for wha he phase drif analysis is concerned. I migh be noiced ha he phase

8 ransmied Phase ransmied Phase 36 F. Chii e al. drif is perfecly racked, even if here is no perfec ime alignmen since he duy cycles are no he same. his advanage can be also poined ou in Figure 1, where he phase drif is repored for a couple of ordinary nodes, respecively, nodes 21 and 25 having quie differen values of duy cycle, in paricular d 21 = 2% and d 25 = 7%. deployed. In his case, o he purpose of evaluaing he phase drif compensaion, i is no useful o refer o a ime diagram as nodes have boh differen d and f values, hence i is presen a ime shifing and a mulienvelope effec due o several modulo- f operaions. As a consequence, he ime diagram is no easy o be inerpreed. able 2 Parameers for es 1 ID l s f d 1 sec sec 1 sec 1% 21 2 sec 98 sec 1 sec 2% 22 1 sec 99 sec 1 sec 1% 23 3 sec 97 sec 1 sec 3% 24 5 sec 95 sec 1 sec 5% 25 7 sec 93 sec 1 sec 7% Figure 1 ransmied phase in he case of differen d: (a) maser node 21 and (b) ordinary node 25 able 3 Parameers for es 2 ID l s f d 3 sec 17 sec 11 sec 1% 21 3 sec 33 sec 36 sec 8.3% 22 3 sec 47 sec 5 sec 6% 23 3 sec 59 sec 62 sec 4.8% 24 3 sec 71 sec 74 sec 4.5% 25 3 sec 99 sec 12 sec 2.9% ime [hh/mm/ss] (a) Figure 9 ransmied phase in he case of differen d: (a) maser node and (b) ordinary node 22 (b) ransmied Phase ime [hh/mm/ss] (a) Oherwise, he SAR MAC conneciviy mainenance migh be beer evaluaed by considering he neighbours able size saisics as far as each node is concerned. In paricular, in Figure 11, he probabiliy densiy funcion (pdf) of having a cerain number of neighbour nodes wihin he MAC able is shown. I is worh noicing ha a generic node never ges disconneced and ha he average able size is almos close o four elemens, hus confirming he goodness of he proposed approach o increase he nework conneciviy, while reducing energy consumpion wihou remarkable overhead increasing. 3 Nework layer proocol ime [hh/mm/ss] (b) o complee he analysis of he SAR MAC proocol performance, he second es case has been aken ino accoun, in which 6 nodes wih independen f values are 3.1 Applicaion requiremens he provisioning of remoe environmenal monioring capabiliies, as expeced wihin he GoodFood Projec, esablishes several requiremens ha communicaions proocols may saisfy. Some of hem are relaed wih he end-user applicaions characerisics, while he ohers

9 Design and applicaion of enhanced communicaion proocols 37 depend on he node s hardware resricions. Firs of all, he daa reporing mehod is essenially ime-driven Al-Karaki and Kamal (24) as applicaions require periodic daa monioring. In fuure sysem releases i will be likely o suppor also query-driven scheme in which he gaeway or anoher node in he nework is able o generae a specific query o anoher node, o acquire a cerain parameer value. Figure SAR MAC able size saisics An addiional requiremen is represened by he need of faul olerance, since a node may fail for several reasons. In his case, MAC and rouing proocols migh cooperae o quickly se up a new end-o-end pah. Besides a cerain amoun of redundancy migh be inroduced o allow daa delivery even in he case of node s rouble due o physical damage or environmenal inerference. his is due o he fac ha he communicaion proocol is unreliable, meaning ha no explici acknowledgemen is provided in order o lower he signalling overhead and o preserve he available bandwidh. his affecs he nework layer performance raher han MAC layer. Finally, scalabiliy issues may be addressed o make he rouing scheme able o operae in he correc way despie he number of deployed nodes ha can be on he order of hundreds o sense a wide area. Moreover, he rouing proocol migh cope wih conemporaneous ransmissions, occurring when several nodes joinly ener he lisening sae rying o conrol he nework congesion. o end wih, he power consumpion is o be deeply considered in designing rouing proocol, as usually nodes have limied energy supply and limied compuing power. In a mulihop WSN, each node plays a dual role as daa sender and daa rouer: he energy needed o perform compuaions and delivering informaion migh dramaically reduce he node s lifeime, hus impacing on he nework lifeime. As a resuls, energy-conserving forms of communicaion and compuaion are of primary imporance. sricly layered archiecure, sofware layers inerac more closely by means of informaion sharing. In paricular, we refer o a proacive algorihm belonging o he class of link-sae proocol ha enhance he capabiliies of he Link Esimaion Paren Selecion (LEPS) proocol (inyos, XXXX). I is based on periodically sending a conrol message o neighbour nodes o carry on informaion needed for building and mainaining he local rouing able, depiced in Figure 12. In paricular, i could be noiced he presence of several possible desinaions (e.g. Sink 1 and Sink 2 ), for each of hem possible nex hop candidaes wih he relaed parameers are lised. However, our approach resors boh o he signalling inroduced by he MAC layer, ha is, he synchronisaion message and by he nework layer, ha is, he ping message, wih he aim of minimising he overhead and make he sysem more adapive in a cross-layer fashion. In paricular, he parameers ransmied along a MAC synchronisaion message, wih period f, are he following: nex hop o reach he he gaeway, ha is, he MAC address of he one hop neighbour disance o he gaeway in erms of number of needed hops phase he schedule ime a which he neighbour become lisening evaluaed according o (2) or (3) in order o minimise he ime laency link qualiy esimaion as he raio of correcly received and he expeced synchronisaion messages from a cerain neighbour. Figure 12 Rouing able general srucure On he oher hand, he parameers relaed o long-erm phenomena are carried ou by he Nework layer messages (ping), wih period p f, in order o avoid unnecessary conrol raffics and hus, reducing congesion. Paricularly, hey are: baery level, ha is, an esimaion of he energy available a ha node congesion level in erms of he raio beween he number of packes presen in he local buffer and he maximum number of packes o be sore in. Once, he rouing able has been filled wih hese parameers, i is possible o derive he proper meric by means of a weighed summaion of hem. 3.2 Proposed soluions In order o evaluae he capabiliy of he proposed MAC scheme in esablishing effecive end-o-end communicaions wihin a WSN, a rouing proocol has been inroduced and inegraed according o he cross-layer design principle (Shakkoai e al., 23), where, in conras o a 3.3 Overall performance analysis es-case scenario he reference scenario considered in performing he simulaions is relaed o he vineyard area seleced for he exploiaion of he GoodFood projec Pilo Side a he

10 38 F. Chii e al. Universiy of Florence s owned Monepaldi farm. In his sie, 15 nodes will be deployed in 1, m 2 area by he end of 25 and a he presen 9 nodes are fully operaive. Each node could be endowed wih sensors monioring hree differen environmens: plan, roo soil and canopy amosphere. In able 4, he characerisics of each sensed environmen in erms of involved sensors and acquisiion period are presened. On he whole nodes, he sensors for soil emperaure and moisure, leaf emperaure and air emperaure/humidiy will be implemened. able 4 Moniored parameers in he vineyard Subsysem Parameers Q [min] Roo soil Soil emperaure 15 Soil moisure 15 Canopy amosphere Air emperaure 15 Air humidiy 15 Plan Leaf emperaure 15 Shoo diameric growh 15 he power consumpion values for he adoped opimised plaform are hose repored in able 1; besides, l = 4 sec, S = 56 sec and consequenly, d = 6.7%, moreover, he deployed baeries are C-ype primary lihium baeries wih 7 Ah 3 V. Finally, he applied communicaions proocol is described in Secion 3.2. he whole sysem is sill running unaended beginning from he summer season; packes are colleced by he maser node and heir correc delivering is moniored by a specific applicaion hosed wihin he remoe server (logger) ha sores and processes daa Experimenal resuls Firs of all, has been measured he node s energeic consumpion by averaging i over he whole nework: i resuls o be equal o 1.44 mah. On he oher hand, is esimaed value, based on modelling he nework behaviour as a FSM and considering he pracical power consumpion values as in able 1, is equal o 1.15 mah, poining ou a good agreemen wih he experimenal resuls. As a consequence, he node life ime is greaer han Seven monhs in he case of above menioned baery adopion, ha allows he nework o operae for a farming season wihou mainenance, as i is claimed by he user requiremens. hen he Packe Delivery Raio (PDR), ha is, he raio beween he number of daa packes correcly received by he maser node and he number of sen packes esimaed according o he daa sampling rae, has been verified o be equal o.948. I could be due o link failures or he residual packe collisions rae, no correcly handled by he SAR MAC proocol, as packes are reransmied when he receiver is he sleeping mode. his allow he WSN uilisaion for accurae environmenal monioring and implicily, validae he efficiency of he adoped communicaion proocol. o he purpose of evaluae he Qualiy of Service (QoS) of he cross-layer soluion, i has been measured he average number of hops needed o deliver a packe o is desinaion. I has been made feasible by concaenaing he idenifier of inermediae nodes in a proper field of he packe header. I is equal o 2.83; his circumsance implies ha almos all he daa packes are no subjec o loops ha involve a leas four hops and he rouing funcionaliies are preserved. Finally, he measured mean delivery delay, ha is he difference beween he packe delivering ime and he insan of is firs sending, resuls o be equal o 6.1 sec, as he phase is randomly seleced wih uniform disribuion and he queue waiing ime migh be negleced for a low loaded condiion. his value represens a reasonable laency for he considered applicaion in which slow varying phenomena are moniored. 4 Conclusion his paper deals wih he proposal of an efficien cross-layer communicaion proocol, allowing remarkable performance, in erms of low power consumpion and QoS, for a WSN operaing in a criical scenario like oudoor environmenal monioring. A his aim an energy efficien MAC proocol, namely SAR MAC, has been proposed and properly inegraed wihin a rouing scheme, minimising he signalling overhead. he proocol has been analyically characerised and generalised o he case of compleely independen nodes in erms of frame period and duy cycle values. Furher, i is modified in order o cope wih densely deployed neworks, giving rise o he Enhanced SAR MAC (SAR+). he performance of he overall mulihop communicaion soluion has been evaluaed, by implemening i on moes operaing in he operaive scenario relaed o he EU Inegraed Projec GoodFood. In paricular, i has been also described a pracical implemenaion of he Pilo Sie, in erms of adoped plaforms and archiecures on which he esing procedure has been performed. he experimenal resuls highlighed a noiceable performance as far as energeic consumpion, he node life ime, he daa delivering efficiency and laency are concerned. his allows he applicaion of he WSN under invesigaion o he field of environmenal monioring. Acknowledgemens he auhors would like o express heir graiude o Professor Gianfranco Manes for his fruiful commens and discussions. A paricular hank he members of he EU Inegraed Projec FP6-IS IP GoodFood WorkPackage 7 for heir supporing in his research. References Al-Karaki, J.N. and Kamal, A.E. (24) Rouing echniques in wireless sensor neworks: a survey, IEEE Communicaion on Mag, December, No. 6, pp Akyildiz, I.F., Su, W., Sankarasubramaniam, Y. and Cayirci, E. (22) A survey on sensor neworks, IEEE Communicaion on Mag, Augus, Vol. 4, No. 8, pp

11 Design and applicaion of enhanced communicaion proocols 39 CC115 (26) Available a: hp:// El-Hoiydi, A., Decoignie, J-D., Enz, C. and Le Roux, E. (23) WiseMAC, an ulra low power MAC proocol for he WiseNE wireless sensor nework, Proceedings of Sensor Sysem 3, June. Hać, A. (23) Wireless Sensor Neworks Designs, John Wiley & Sons. IEEE sandard (1999) Wireless LAN medium access conrol (MAC) and physical layer (PHY) specificaions. Langendoen, K. and Halkes, G. (25) Energy-efficien medium access conrol, echnical Repor, Delf Universiy of echnology. Lu, G., Krishnamachari, B. and Raghavendra, C. (24) An adapive energy-efficien and low-laency MAC for daa gahering in sensor neworks, Proceedings of WMAN 4, April. Maoli, V., Mondini, A., Razeeb, K.M., O Flynn, B., Murphy, F., Bellis, S., Collodi, G., Manes, A., Pennacchia, P., Mazzolai, B. and Dario, P. (25) Developmen of a programmable sensor inerface for wireless nework nodes for inelligen agriculural applicaions, Proceedings of IE 5, June. MICA2 Series (XXXX) Available a: Producs. hp:// Shakkoai, S., Rappapor,.S. and Karlsson, P.C. (23) Cross-layer design for wireless neworks, IEEE Communicaion on Mag, Ocober, Vol. 41, No. 1, pp inyos (XXXX) Available a: hp:// van Dam,. and Langendoen, K. (23) An adapive energy-efficien MAC proocol for wireless sensor neworks, Proceedings of Sensor Sysem 3, November. Ye, W., Heidemann, J. and Esrin, D. (22) An energy-efficien MAC proocol for wireless sensor neworks, Proceedings of INFOCOM 2, June. Noe 1 As elecromagneic power can be measured in VA, i follows ha energy migh be expressed in VA s. hus, once a cerain volage has been fixed (depending on he seleced baery) energy can be characerised wihou ambiguiy.

12 31 F. Chii e al. Appendix In order o poin ou he mulicas gain provided by he proposed SAR+ approach, we derive he average number of subse K, ha is, he average number of synchronisaion messages delivered. We suppose ha here are N nodes and he average number of nodes belonging o a subse is n. As a consequence, K migh be expressed as: N K = (8) n o evaluae n we refer o Figure A1: in paricular a couple of nodes, namely i and j, belongs o he same subse if: i j l Rx (9) where i and j are, respecively, he insans a which nodes i and j go ino he lisening sae, l is he lisening period and Rx is he minimum ime needed o receive a synchronisaion message. For he sake of generaliy, we suppose ha i and j are boh random variable uniformly disribued over he inerval (, f ]. As a resul, he probabiliy ha wo nodes belong o he same subse is: P 2 = 2 f ( l Rx ) (1) Generalising (1) for he case of n neighbours belonging o he same subse, we have: [ ] 2 n 1 P n = ( l Rx ) (11) f since he ime each node eners he lisening sae is a random variables. hen he average number of nodes belonging o a generic subse can be expressed as: N 1 [ ] 2 n 1 n = n ( l Rx ) (12) f n=2 By defining q = 2 f ( l Rx ), i yields: n = (N 1)qN Nq N 1 + 2q 3 q 2 4q + 2 (1 q) 2 (13) he derivaion of K is sraighforward by subsiuing (12) ino (8). Figure A1: Subse derivaion procedure for nodes i and j l i j i+l j+l x+ l