Building Robust Wireless LAN for Industrial Control with DSSS-CDMA Cellphone Network Paradigm

Transcription

1 Buldg Robust Wreless LAN for Idustral Cotrol wth DSSS-CDMA Cellphoe Networ Paradgm Qx Wag,XueLu, Wequ Che, Webo He, ad Marco Caccamo Departmet of Computer Scece, Uversty of Illos at Urbaa-Champag Emal: {qwag4, xuelu, webohe, ECECS Departmet, Uversty of Ccat Emal: Abstract Deployg Wreless LAN for Idustral Cotrol (IC- WLAN) has may beefts, such as moblty, low deploymet cost ad ease of recofgurato. However, the top cocer s robustess of wreless commucatos. Wreless cotrol loops must be mataed uder persstet adverse chael codtos, such as ose, large-scale path loss ad fadg. May electro-magetc terferece sources dustral evromets, e.g. electrc motor ad weldg, mae wreless commucato more challegg. The covetoal IEEE WLANs, whch are desged for provdg hgh badwdth stead of hgh robustess, are therefore approprate for IC-WLAN. O the other had, f the low data rate feature of dustral cotrol s fully exploted by the state-of-the-art Drect Sequece Spread Spectrum (DSSS) techology, much hgher robustess ca be acheved. We hereby propose usg DSSS-CDMA to buld IC-WLAN, ad explotg the low data rate feature of dustral cotrol loops for ehaced robustess. We carred out fe-graed physcal layer smulatos ad Mote Carlo comparsos. The results show that DSSS-CDMA IC- WLAN provdes much hgher robustess tha IEEE WLAN, so that relable wreless dustral cotrol loops are made feasble. The DSSS-CDMA IC-WLAN scheme also opes up a ew problem space for terdscplary study, volvg real-tme schedulg ad resource maagemet, commucato, etworg ad cotrol. I ths paper, we study the resource maagemet problems o maxmzg robustess ad mmzg cotrol utlty loss. Aalytcal resource optmzato solutos are gve. 1. Itroducto Recetly, there are creasg efforts deployg Wreless LANs (WLAN) for dustral cotrol [28][37][38][16][22]. Idustral Cotrol WLAN (IC- WLAN) has may desrable features, such as exteded mechacal freedom ad moblty, low deploymet cost ad ease of recofgurato, ad s therefore mportat for the deploymet of robots, automatc vehcles ad dstrbuted maufacturg systems. At a larger scale, deployg IC-WLAN s also part of the efforts toward ubqutous computg. However, IC-WLAN also rases ew challeges, amog whch, probably the most outstadg oe s ts robustess. WLAN commucatos are heretly more vulerable tha wred LAN commucatos due to may reasos, e.g. multple-access coteto, Rado Frequecy (RF) terferece, large-scale path loss ad fadg [31]. IC- WLANs pose much hgher robustess requremets tha covetoal WLANs for offce or home use. For most offce or home WLAN applcatos, t s acceptable to have a few secods or eve mutes of RF terferece, such as accdetally turg o a ucooperatve devce whch broadcasts at the same RF bad as the exstg WLAN. The terfered wreless coectos just eed to bacoff tll the terferece s over ad the retrasmt. I fact, ths s the stadard behavor adopted by IEEE WLAN [2][3][4][7], whch s the owadays domat WLAN scheme. However, for dustral cotrol, such behavor s ofte ot acceptable. Most dstrbuted dustral cotrol loops are real-tme, the bacoff behavor wll cause deadle msses, result performace losses, halts/resets of maufacturg ppeles or defects products. For example, a couple of hudred mlsecods of bacoff s eough to mae a verted pedulum fall. Therefore, for most dustral cotrols, commucatos must be mataed eve uder adverse chael codtos. To acheve such robustess, cables of dustral cotrol wred LANs are shelded to protect agast terferece. For wreless LAN, provdg smlar sheldg s more ecessary ad challegg. The terferg RF sources ca be other rado devces that use the same RF

2 bad, tured o sde or eve outsde of the factory; or the may Electro-Magetc Iterferece (EMI) sources, such as electrc motors or electrc weldg [39][13]. The EMI s especally challegg, ot oly because t s commo to dustral evromets, but also because t s persstet, e.g. a electrc motor ca ru for hours or days. Also, due to heavy obstructos, the wreless medums of dustral evromets are ow to suffer much more serous largescale path loss ad fadg tha other door evromets [31]. Geercally speag, RF terferece, large-scale path loss ad fadg reduce the Sgal-to-Nose Rato (SNR) of the wreless chael. Whe the SNR s lower tha a certa threshold, the Bt Error Rate (BER) rses over the acceptable lmt, ad the wreless coecto may brea dow. Therefore, for IC-WLAN, t s ecessary to provde hghest SNR possble to mata wreless cotrol loops uder adverse codtos. A promsg soluto les the state-of-the-art Drect Sequece Spread Spectrum (DSSS) techology. DSSS has the desrable feature that allows tradeoffs betwee data throughput versus SNR. Specfcally, a smaller data throughput (for cotrol loops, t meas slower samplg/actuatg rates ad smaller pacet szes) correspods to a hgher SNR ad vce versa. O the other had, dustral cotrol loops usually cosst of low-data-throughput stable traffcs [27]. For example, due to decades of developmet ad wde deploymet of step motors [23][25], for most cotemporary dustral mechacal cotrols, fe-graed hgh-rate cotrols are carred out locally, whereas cotrol traffcs betwee dstrbuted odes are mostly coarse-graed, cosstg of low samplg/actuatg rates ad small pacet szes. Typcally, the samplg/actuatg rates betwee dstrbuted odes are aroud or below 10Hz, sometmes eve aroud 1Hz; the pacet sze are usually aroud a couple of hudred bts. I ths paper, we are terested explotg the features of cotrol loop traffc ad DSSS techology to buld robust IC-WLANs. By carryg out fe-graed physcal layer smulatos ad Mote Carlo comparsos, we show that whe the low data rate feature of cotrol loops s fully exploted, DSSS ca acheve much hgher robustess tha what s provded by IEEE , so that wreless dustral cotrol s made practcal (see Secto 4). Specfcally, aroud 10 20dB ad 25 35dB robustess mprovemets are acheved compared to IEEE b ad IEEE a respectvely, whch are sgfcat accordg to commucato egeerg crtera. Meawhle, compared to the predomat ad hoc paradgm of IEEE WLAN, the covetoal DSSS- CDMA cellphoe etwor paradgm s also more desrable for IC-WLAN. That s, every WLAN s a cell, wth oe Basestato (BS) ad several Remote Statos (RS). Wreless commucatos oly tae place betwee a basestato ad a remote stato of the same cell. Iter-cell commucatos are carred out through wrele bacboes betwee basestatos. The reasos are as followg: ) Cotrol loop traffcs are ofte perodc traffcs wth low data rates, ad most cotrol logcs cur low computato. Therefore t s a commo ad ecoomc practce factores to have oe powerful cetralzed basestato cotrollg all maches a local area [27]. A lots of legacy systems are bult upo such basestato-cetered commucato paradgm. ) The basestato-cetered paradgm also maes cetralzed real-tme schedulg easy tomplemet. Ipractce, cetralzed real-tme schedulg s ofte more preferable tha dstrbuted real-tme schedulg due to ts robustess ad smplcty. ) Idustral cotrol applcatos are typcally deployed well-bult permaet facltes, where powerful wrele bacboes for ter-basestato commucatos are avalable. Therefore, most cases, the beefts of wreless commucatos (mechacal freedom, moblty, flexblty) are oly sgfcat at the last hop. A cellphoe etwor paradgm matches such demad. v) At the MAC layer, CDMA s more preferable because of ts ease of schedulg, overru solato ad low overhead. The ma cotrbuto of ths paper s to pot out that by fully explotg the low data rate feature of dustral cotrol loops, the covetoal DSSS-CDMA cellphoe etwor paradgm s a better approach for buldg robust IC-WLAN, although IEEE s the owadays predomat WLAN scheme. However, cotemporary DSSS- CDMA cellphoe etwors are stll more focused o provdg hgher data throughput stead of hgher robustess, ad o much effort has bee made to customze resource allocatos to the eeds of dustral cotrol. I ths paper, we study the resource maagemet ssues o DSSS-CDMA IC-WLAN. Aalytcal resource optmzato solutos are gve for maxmzg robustess ad mmzg cotrol utlty loss respectvely. Our study also show that the resource maagemet problems o DSSS-CDMA IC-WLAN are o-trval. I fact, DSSS-CDMA IC-WLAN scheme opes up a ew problem space for terdscplary study, volvg real-tme schedulg ad resource maagemet, commucato, etworg ad cotrol. The rest of the paper s orgazed as follows: Secto 2 gves bacgroud o DSSS techology. The DSSS-CDMA IC-WLAN s proposed Secto 3, together wth some aalytcal results o ts resource optmzato. Fe-graed physcal layer smulatos are carred out Secto 4 to llustrate the robustess of the DSSS-CDMA IC-WLAN, followed by more extesve Mote Carlo smulatos that compare the robustess wth IEEE WLANs. Related wors are dscussed Secto 5. Secto 6 cocludes the paper.

3 2. Bacgroud DSSS s a physcal layer modulato/demodulato scheme for dgtal commucato [35][33][18]. It reshapes basebad sgal to occupy a wder spectrum 1.Atthetrasmtter, a user data bt stream of bt rate r b (.e. bt durato def of T b = 1/r b )sscrambled wth a Pseudo Nose (PN) sequece of chp rate r c (.e. chp durato of T c = 1/r c ), def producg a chp stream of rate r c. r c s a postve teger multple of r b, the rato g def = r c /r b s called processg ga. At the recever, f the chp stream s descrambled wth the same PN sequece, the orgal data bt stream s recovered. If a dfferet PN sequece s appled, or the scramble/descramble PN sequeces are ot sychrozed, the orgal data bt stream caot be recovered, stead, a ose-le radom chp stream s geerated. To summarze, each PN sequece creates a DSSS data chael. Note although DSSS requres sychrozato betwee each trasmtter-recever par, sychrozatos betwee trasmtters are ot eeded f CDMA s deployed at the MAC layer. More detals o DSSS ad ts termologes are elaborated [36] Appedx I. At the MAC layer, most cotemporary dgtal wreless systems pc ether of the two alteratves: oe s Code Dvso Multple Access (CDMA), the other s Tme Dvso Multple Access (TDMA). If DSSS-CDMA s deployed, dfferet data bt streams scrambled wth dfferet PN sequeces are trasmtted parallel through the same RF bad. At the recever, by applyg dfferet PN sequeces, the teded data bt stream s fltered out. If DSSS-TDMA s deployed, dfferet data bt streams scrambled wth dfferet PN sequeces occupy o-overlappg tme slots, whch requres more sophstcated schedulg schemes. Though ether alteratve s feasble, we fd DSSS-CDMA to be more preferable for IC-WLANs for the followg reasos: ) ease of real-tme schedulg; ) heret solato betwee coectos; ) less commucato overhead, especally uder adverse chael codtos. The frst two pots are straghtforward ad terrelated: Uder CDMA archtecture, a coecto exclusvely occupes oe or several CDMA chaels (PNs),.e., at MAC layer ad above, CDMA chaels are ot shared. There s o eed to schedule dfferet real-tme coectos wth oe CDMA chael, ad the overru of oe real-tme coecto does ot affect ay other real-tme coectos. I cotrast, a TDMA scheme, tme slots must be scheduled to serve 1 For coveece, we refer to DSSS as a basebad modulato/demodulato scheme. Correspodgly, the modulato/demodulato scheme that shfts basebad sgal to/from RF bad s referred to as RF modulato/demodulato. Typcal RF modulato/demodulato schemes for DSSS ca be Quadrature Phase Shft Keyg (QPSK) or Bary Phase Shft Keyg (BPSK), both ca acheve same robustess ( sese of BER) wth same SNR per bt [35][20]. dfferet real-tme coectos, ad f a real-tme coecto overrus ts tme slot, subsequet real-tme coectos wll be affected. The thrd pot eeds more elaborato: Geercally speag, DSSS requres tme sychrozato betwee trasmtter ad recever. However, uder adverse chael codtos, TDMA curs much more tme sychrozato overhead tha CDMA. Further quattatve aalyss s gve [36] Appedx II. Quattatvely, a umber of mportat features of DSSS commucato s captured by ts Bt Error Rate (BER) upper boud (1), whch assumes QPSK RF modulato, ad per coecto plot toe [35][26] (dfferet mplemetato alteratves may affect detals of the formula, but wll ot cause fudametal dffereces): ( ) gp u P ber exp J + Ξ =1, u P + H h=1 A h + P u (1) where P ber s the BER; g s processg ga; J s the receved power of Exteral RF Iterferece (EI), whch specfcally refers to EMI, thermal ose ad the RF terferece from RF devces that are tured o accdetally or malcously. P ( =1...Ξ) s the receved power of CDMA chael, Ξ s the total umber of CDMA chaels. u s the teded chael, whose correspodg receved power s P u. Each trasmttg ode may sed out several CDMA chaels parallel. To ease the recepto, the ode may also trasmt a addtoal plot toe. I (1) the plot toe of trasmttg ode h (h =1,...,H)sof Ξ =1, u P + H h=1 A h s therefore the up- power A h. per boud of total Multple Access Iterferece (MAI),.e. the terferece caused by other CDMA chaels ad plot toes receved parallel wth the teded chael. Note P u also appears the deomator, addg up to the total terferece power. Ths s to provde a pessmstc estmato o Iter Symbol Iterferece (ISI), whch s usually a result of multpath fadg. To smplfy, we ca merge Ξ =1, u P ad P u together to be deoted as P.The gp u /(J + P + h A h) part shows the effectve SNR for the teded chael, where J + P + h A h represets the upper boud of ose power ad gp u represets effectve sgal power. The bgger the SNR, the smaller the probablty of bt error P ber.whep ber s below a certa threshold Θ ber, the wreless commucato s acceptable for dustral cotrol. Therefore, to mata a IC-WLAN chael fact meas to mata the SNR of the chael from droppg below a acceptable threshold Θ sr. Wthout error correcto codg, the Pacet Error Rate (PER) P per s: P per = 1 (1 P ber ) Lpt (2) Or equvaletly: P ber = 1 (1 P per ) 1/Lpt (3)

4 where L pt s the bt legth of the pacet. Whe errorcorrecto codg s deployed, such as covolutoal code [30], (2) wll have a more complcated form, but stll, P per decreases as P ber decreases. Geerally speag, a maxmum acceptable pacet error rate Θ per correspods to a maxmum acceptable bt error rate Θ ber, whch further maps to a mmum acceptable SNR Θ sr. Specfcally, we should mata: gp u J + Ξ P + H h A h Θ sr (4) = l Θ ber (5) (because of (1)) ( = l 1 (1 Θ per ) 1/Lpt) (6) (because of (3)) (1) mples that SNR of the teded chael ca be rased by creasg the processg ga g. Meawhle, g s defed as the rato of chp rate ad bt rate: g def = r c /r b. Usually, chp rate r c s fxed by hardware because of multpath effect ad hardware cost costrats [29][35], therefore rasg processg ga meas slowg dow user data bt rate r b. DSSS hereby provdes a mechasm to leverage betwee SNR ad data bt rate. Later o, we wll carry out resource optmzato aalyss based o the above equato/equatos (see Secto 3.2 ad 3.3). 3. DSSS-CDMA IC-WLAN Archtecture 3.1. The Overall Archtecture Accordg to the aalyss of Secto 1 ad 2, we choose to desg our IC-WLAN followg the DSSS-CDMA cellphoe etwor paradgm, whch deploys DSSS for physcal layer, ad CDMA for MAC. I each cell, there s a Basestato (BS). Basestatos of dfferet cells are coected va wrele. The wreless commucatos oly tae place betwee a basestato ad a Remote Stato (RS) of the same cell. Specfcally, the IC-WLAN archtecture s llustrated by Fg. 1. I ths paper, we focus o the sgle cell scearo. The whole RF bad s evely parttoed to two halves, oe for dowl (from basestato to remote statos) ad the other for upl (from remote statos to basestato) 2. Each coecto cossts of oe CDMA chael each drecto (dowl ad upl). Uless explctly deoted, coecto a..a. cotrol loop refers to both dowl ad upl of the coecto; a CDMA chael of 2 Ths s a commo practce for CDMA archtecture, so that the recever s ot terfered by the trasmtter of the same ode. Fgure 1. DSSS-CDMA IC-WLAN archtecture coecto also refers to both chaels dowl ad upl. Samplg/actuatg pacets are set cotuously each CDMA chael,.e., the bts of cosequet samplg/actuatg pacets of same cotrol loop form a cotuous bt stream, smlar to the cellphoe sesso patter. Therefore, the samplg/actuatg perod s the same as the pacet trasmsso perod Resource Plag for Maxmzed Robustess Based o gve codtos ad dfferet optmzato goals, varous resource plag problems ca be formulated. Sce robustess s ofte the top cocer for IC- WLAN, let us frst aalyze the optmal cofgurato for maxmzed robustess. Suppose the followg parameters are gve: The set of cotrol loops to be cluded the IC-WLAN are C def = {C 1,C 2,...,C N }. Each cotrol loop C ( = 1, 2,...,N) correspods to a mmum samplg/actuatg rate f m, a maxmum acceptable pacet error rate Θ per, ad a samplg/actuatg pacet bt legth L pt (suppose samplg/actuatg pacets are of same legth, f ot, paddgs are used to mae them the same). Each ode ca trasmt wth maxmum power of P max. I the upl, power balacg s carred out to deal wth ear-far problem [31]. Wthout loss of geeralty, suppose the covetoal QPSK RF modulato/demodulato ad per ode plot toe are deployed. The power level of a ode s plot toe s the same as that of a CDMA data chael. The hardware fxed chp rate s r c, whch correspods to a chp durato of T c =1/r c. The cofgurable parameters are the cotrol loops processg ga g, wth the followg value rage: 1 g g max,adg s a teger. =1, 2,...,N. (7) where g max s a hardware-mplemetato-depedet costat (e.g. f g s specfed by a usged byte the hardware, the g ca ot exceed 256).

5 O the other had, gve chp rate r c, pacet bt legth L pt, ad the chose processg ga g, the pacet rate f (.e. the umber of pacets that ca be trasmtted per secod) s: f = r c g L pt (8) The pacet rate should ot volate mmum samplg/actuatg rate requremet, that s: f f m g L pt r c f m r c g L pt f m, =1, 2,...,N. (9) As for the wreless medum, let J ( =0, 1, 2,...,N) be the exteral RF terferece power receved by ode (ode 0 refers to the basestato, ode 1, 2,...,N refer to the remote stato of coecto 1, 2,..., N respectvely). Suppose the dowl/upl power atteuato for coecto are α dow ad α up respectvely. The robustess requremet for coecto s that the pacet error rate should ot exceed a threshold Θ per, both dowl ad upl. Accordg to (6), the maxmum ca be mapped to a m- acceptable pacet error rate Θ per mum acceptable SNR Θ sr. Accordg to (4), the robustess requremet for dowl of coecto s formalzed as: g P r dl J + N =1 P r dl + A r dl Θ sr, (10) =1, 2,...,N. where P r dl ( =1, 2,...,N) s the receved power of CDMA dowl chael at remote stato. A r dl s the receved plot toe power at remote stato (for dowl, basestato s the oly ode that trasmts plot toe). Assume the basestato trasmsso power allocated to all N dowl chaels are equal, ad the plot toe s of same power level as a CDMA chael. Also, for maxmum robustess, the basestato should trasmt wth ts maxmum power P max, therefore: P r dl 1 = A r dl = P r dl 2 = α dow =...= P r dl N P max N +1 (11) Substtutg (11) to (10), the dowl robustess requremet s trasformed to: α dow g P max (N +1)(J + α dow P max ) Θsr, (12) =1, 2,...,N. Accordg to (4), the robustess requremet for upl of coecto shall be formalzed as: g P r upl J 0 + N =1 P r upl + N upl =1 Ar Θ sr, (13) =1, 2,...,N. where P r upl ( =1, 2,...,N) s the receved power of CDMA upl chael at the basestato. A r upl ( = 1, 2,...,N) s the (basestato s) receved power of plot toe trasmtted from remote stato. Because of power balacg, there should have: P r upl 1 = P r upl 2 =...= P r upl N (14) O the other had, assume for each remote stato ( = 1, 2,...,N), ts trasmsso power P t upl s equally dvded by upl chael ad the plot toe, the: P r upl = A r upl = α up P t upl, =1, 2,...,N. (15) 2 Lastly, each remote stato caot trasmt wth power larger tha P max,.e.: P t upl P max, =1, 2,...,N. (16) The trasmsso power of each remote stato P t upl should be maxmzed to crease SNR (ad therefore robustess), meawhle matag the costrats of (14) ad (16). Therefore, the remote stato that suffers the worst case upl power atteuato should trasmt wth power P max, ad all other remote statos should adjust ther trasmsso power accordg to (14). Formally: therefore where P t upl = P max, P r upl = A r upl = α up P max, (17) 2 =argm {1,2,...,N} {α up }. (18) Note operator argm x A {F (x)} returs x A,such that x A, F (x ) F (x). Deote (14), (15) P r upl P t upl P r upl = A r upl = αup P max α up = A r upl = α up = α up P t upl 2 (because of (17)) P max 2 = P r upl (19). (20) up def α = α up =m{αup 1,αup 2,...,αup N }, (21)

6 ad substtute (17) ad (20) to (13), the upl robustess requremet s trasformed to: α up g P max 2(J 0 + α up NP max ) Θsr, =1, 2,...,N. (22) Suppose the power atteuatos α dow 1, α dow, α up 1, αup 2,..., αup N are gve, α dow N (12) J def = J dow 2,..., ( α dow g P max ) (N +1)Θ sr α dow P max, =1, 2,...,N. (23) ( α up g P max ) (22) J 0 α up NP max J dow def = 2Θ sr J up, =1, 2,...,N. (24) J up ad therefore represets the maxmum tolerable exteral RF terferece for dowl ad upl of dow coecto respectvely. That s, whe J exceeds J, coecto s dowl wll have a pacet error rate over acceptable lmt Θ per ;whej up 0 exceeds J, coectos s upl wll have a pacet error rate over acceptable lmt.defe Θ per J m def = m{ J up 1 dow dow J 1, J, J up 2 J dow 2,..., N, up,..., J }, (25).e. J m represets the mmum exteral RF terferece power eeded to brea dow at least oe of the coectos. To maxmze robustess therefore meas to maxmze J m uder the costrats of (9) ad (7). It s easy to derve that J m s maxmzed whe g = m { r c /(L pt f m ),g max}. Smlarly, suppose J 0,J 1,J 2,...,J N are gve, the tolerable power atteuatos are maxmzed (.e. α dow 1, α dow 2,..., α dow N, α up 1, αup 2,..., αup N are mmzed) whe g =m { r c /(L pt f m ),g max}. Therefore, maxmum robustess s acheved whe each cotrol loop deploys maxmum processg ga possble, such that the pacet rate s slowed dow as much as possble. Formally, ths s summarzed by the followg proposto: Proposto 1 (Maxmum Robustess Cofgurato) To acheve maxmum robustess, a DSSS-CDMA IC- WLAN should pc g =m { r c /(L pt f m ),g max} ( =1, 2,...,N). I Secto 4, we shall see that by deployg processg ga accordg to Proposto 1, a DSSS-CDMA IC- WLAN ca acheve approxmately 10 20dB ad 25 35dB robustess mprovemet compared to IEEE b N ad IEEE a respectvely, whch are sgfcat mprovemets accordg to commucato egeerg crtera. The uderlyg reaso s that IC-WLAN does ot am at achevg hgh data throughput; stead, cotrol traffcs are of extremely low data throughput, whch ca be exploted by DSSS to acheve much hgher robustess Resource Plag for Mmzed Utlty Loss The resource plag Secto 3.2 does ot cosder cotrol performace. Smply speag, gve mmum allowed samplg/actuatg rate f m ad maxmum acceptable pacet error rate Θ per, Proposto 1 gves the cofgurato that wll tolerate maxmumexteralrf terferece J worst or worst case power atteuato α worst.ifthewreless chael codto s better tha the worst case (represeted by J worst or α worst ), stcg to the cofgurato deoted by Proposto 1 s wasteful. Hgher cotrol performace ca be acheved wth less robust cofgurato. Specfcally, suppose the followg codtos are gve: α dow ad α up are the dowl/upl power atteuato for coecto respectvely. Deote α up = m{α up 1,αup 2,...,αup N }. J s the exteral RF terferece receved at the RS of coecto ( =1, 2,...,N). J 0 s the exteral RF terferece receved at the BS. The hardware fxed chp rate s r c. The maxmum trasmsso power of each ode s P max. For a coecto ( = 1, 2,...,N), the samplg/actuatg pacet bt legth s L pt ; ts mmum allowed samplg/actuatg rate s f m ; ad maxmum acceptable pacet error rate for both dowl/upl s Θ per. It ca be reasoably assumed that the utlty loss fucto for each coecto ( =1, 2,...,N) s U (f, P per ),.e. a fucto of the coecto s samplg/actuatg rate f ad pacet error rate P per.itsalso reasoable to assume the global utlty loss fucto U to be U = N =1 U (f, P per ). Same as Secto 3.2, we stll assume the cofgurable varables are the process ga of each coecto g. Due to page lmts, we drectly gve our results: For clarty, the optmzato problem s restated as follows: m U = N =1 s.t.: g 1 ζ l g U ( f (g ), P per (g ) ) (26) ( r c f m L pt ) 1 (1 Θ per )1/Lpt (27) (28) 1 g g max (29) g s a teger, =1, 2,...,N. (30)

7 where f (g ) s defed (8); ζ s defed as follows: { def α dow P max ζ = m (N +1)(J + α dow P max ), α up P max } 2(J 0 + α up NP max, (31) ) ad P per = 1 (1 exp( ζ g )) Lpt. (32) Costrats (27) (30) ether result a empty set, whch meas o feasble g exsts; or are equvalet to the followg form: a g b,g s a teger, ad =1, 2,...,N, (33) where a, b are postve tegers. Specfcally, whe U are of the followg shapes, quasclosed-form aalytcal optmal soluto exsts: or U (f, P per)=w (f (1 P per U (f, P per )) β, )=w exp( β f (1 P per )) (34) where w ad β are postve weght ad sestvty coeffcets. f (1 P per ) s the expected umber of samplg/actuatg pacets that are delvered correctly per secod. Lemma 1 For fucto F (x) =exp(ζ x) (1+L pt ζ x), where L pt ad ζ have the same deftos as before ad L pt > 1, equato F (x) =0hasoeadolyoepostve root x >0. Theorem 1 (Mmum Utlty Loss Cofgurato) Suppose feasble g for the costrat set (33) exsts. Ad suppose x s the sgleto postve root to equato exp(ζ x) (1 + L pt ζ x)=0, L pt > 1 ad U comples wth ether form defed (34). The the optmal g for the utlty loss mmzato problem s: a whe x <a {, ( g argm = g { x, x } U f (g), P per(g))} whe a x b, b whe x >b where a ad b are defed (33). 4. Smulato ad Comparsos I ths secto, we are gog to demostrate the effectveess of our DSSS-CDMA IC-WLAN desg, ad compare t wth IEEE WLANs, whch are the curret predomat WLAN schemes. IEEE WLANs ca be further categorzed to IEEE b[4], a[3] ad g[7]. IEEE b/a/g share the same MAC layer specfcato (wth mor varatos), whle dffer ther physcal layers. IEEE b operates at the 2.4GHz RF bad ad deploys DSSS physcal layer 3. IEEE a operates at the 5GHz RF bad ad deploys Orthogoal Frequecy Dvso Multplexg (OFDM) [14][19][21] physcal layer. IEEE g s bascally the combato of b ad a. Usually, IEEE operates uder Dstrbuted Coordato Fucto (DCF) mode, whch carres out CSMA/CA ad MACAW [15] MAC protocol. DCF s therefore coteto/radom-bacoff based ad s ot desged for real-tme systems. However, IEEE also specfes the Pot Coordato Fucto (PCF) mode, where the basestato polls each remote stato. PCF s coteto-free ad s the scheme desged for real-tme systems. Therefore, we compare the performace of our proposed DSSS-CDMA IC-WLAN wth IEEE b/a PCF IC-WLAN Fe-graed Physcal Layer Smulato Frst we carry out fe-graed physcal layer smulato to demostrate the effectveess of the DSSS-CDMA IC- WLAN scheme. We buld our smulato evromet o top of J-Sm erel [9]. The scearo s depcted Fg. 2(a). The IC-WLAN cludes two coectos: coecto 1 ad 2, each cotrols a Iverted Pedulum (IP) [32], deoted as IP 1 ad 2 Fg. 2(a). Each IP s a remote stato of the IC-WLAN, whch perodcally seds bac IP state to the basestato. Based o the most up-to-date IP state, the basestato calculates the ext cotrol ad seds t bac to the IP. Wthout loss of geeralty, we assume the two IPs are the same, as show Fg. 2(b), where x s the posto of IP cart, θ s the agular devato of IP from vertcal posto, ad u s the velocty cotrol voltage appled to IP cart. The state trasto matrx ad cotrol matrx are also depcted the fgure 4. The IP cart moves alog the x axs to eep the IP stadg vertcally. The requremet s that the IP must ot fall, otherwse a hgh cost resettg procedure s curred. Specfcally, we must mata θ < π 6. The samplg/actuatg pacet legth are both 152 bts. Emprcally, we ow the IP s mmum samplg/actuatg rate s f m 1 = f m 2 =10Hz. We carry out smulato uder both DSSS-CDMA scheme ad IEEE b scheme. To mae a far comparso, both schemes occupy the same RF bad of GHz, whch s a typcal RF bad deployed by IEEE 3 IEEE b also deploys CCK for hgher data throughput modes, whch are less robust. 4 A addtoal heurstc s added for the cotrol: whe θ ad u are of opposte sgs, u s obvously a out-of-date cotrol (because of delay) ad therefore gored.

8 802.11b. For DSSS-CDMA, the RF bad s dvded to two halves: GHz for dowl ad GHz for upl. For IEEE b, the sgal occupes the whole RF bad, but pacets are tme dvsoed to dowl pacets ad upl pacets. Therefore, the chp rate for DSSS-CDMA ad IEEE b are rc cdma =5.5Mcps ad rc eee80211b =11Mcps respectvely. For DSSS-CDMA scheme, we pc the hardwaremplemetato-depedet processg ga upper boud to be g max = Accordg to Proposto 1, the processg ga that maxmze robustess s therefore g cdma = m{ , 1024} = Furthermore, wthout loss of geeralty, QPSK ad per-ode plot toe are deployed for RF modulato/demodulato; the receved power of all data chaels are the same, ad the plot toe s of same power level as a data chael trasmtted from the same ode. At the upl, power balace s carred out to deal wth the ear-far problem. For IEEE b scheme, the most robust mode of 1Mbps throughput s deployed, whch correspods to a processg ga of g eee80211b =11ad Dfferetal BPSK (DBPSK) RF modulato/demodulato. Aga, to be far, the IEEE b WLAN wors pure PCF, whch s the real-tme mode for IEEE WLAN. Uder such mode, the basestato bascally polls IP 1 ad IP 2 a roud rob patter. The cotrol pacet s set to the IP as the poll pacet, ad the sample pacet s set bac from IP as the acowledgemet pacet. To demostrate the robustess, a exteral RF terferece source s placed ear IP 1 (see Fg. 2(a)), whose power spectrum occupes the same RF bad that DSSS-CDMA ad IEEE b are usg. To be far, uder both DSSS- CDMA ad IEEE b schemes, the maxmum trasmsso power of all odes (clude basestato, remote statos ad the exteral RF terferece source) are 30dBm (dbm s the wdely used ut for sgal power. A sgal power of P watt s sad to be of 10 log 10 (P/0.001) dbm), whch s the maxmum trasmsso power allowed by FCC for IEEE b. The oly excepto s for DSSS- CDMA upls, where the trasmsso power must also comply wth the power balacg requremet to produce the same power level at the basestato. The power balacg requremet maes the stuato more pessmstc o the DSSS-CDMA sde. The smulato starts at tme 0sec ad eds at tme 30sec. The exteral RF terferece source s tured o at tme 5sec ad tured off at tme 15sec. The wreless medum stace s geerated accordg to the radom model descrbed Table 1. The model s typcal for door dustral evromets [31][34]. To deal wth multpath fadg, two-fger RAKE recevers [29][35] are deployed for both 5 Ths value s pced smply because cdmaoe [1] (the most prevalet DSSS-CDMA cellphoe etwor stadard North Amerca) complat hardware all supports a processg ga of at least Table 1. Wreless Medum Model Large-scale path loss Log-ormal shadowg model wth model β =4 6, σ =6.8dB Small-scale fadg model Raylegh Multpath max excess delay sec Addtve Whte Gaussa Spectral desty = 174dBm/Hz Nose β s the path loss expoet, σ s the log-ormal stadard devato. Typcally refers to thermal ose. DSSS-CDMA ad IEEE b odes. Fg. 3 shows the trace of θ. It ca be see that throughout the tme, uder DSSS-CDMA scheme, both IP 1 ad IP 2 rema farly stable, eve durg exteral RF terferece (5 15sec) perod. Ths shows the wreless cotrol loops are mataed uder adverse chael codtos. However, uder IEEE b, IP 1 always fall due to exteral RF terferece (every tme t falls, the IP s set to 0.5rad ad stay there for 0.2sec to restart). Note IP 2 uder IEEE b ca also survve exteral RF terferece because t s much closer to basestato tha to the exteral RF terferece source 6. (a) Node layout of fe-graed physcal layer smulato. (b) Iverted Pedulum Fgure 2. Smulated Scearo 4.2. Robustess Comparsos Next, we are terested carryg out more rgd comparsos o robustess, betwee the proposed DSSS- CDMA scheme ad IEEE schemes. Geerally speag, the ma objectve of IEEE WLANs s to provde hgh data throughput. Ths s a msmatch for most dstrbuted dustral cotrol loops, for whch, the data throughput demad may be extremely low 6 To mae the smulato more optmstc o the IEEE sde, stead of bacg off (whch wll certaly cause deadle mss o the realtme wreless cotrol loop), the PCF scheme eeps pollg the remote stato eve f t detects ose o the wreless medum.

9 θ (rad) of IP1, DS CDMA θ (rad) of IP1, b Tme (sec) Tme (sec) θ (rad) of IP2, DS CDMA θ (rad) of IP2, b Tme (sec) Tme (sec) Fgure 3. Smulato Results (θ trace) (typcal pacet legths are aroud bts, ad mmum acceptable samplg/actuatg rate ca be lower tha 10Hz, or eve aroud 1Hz), whereas the robustess demad s hgh. Idustral cotrol requres the perodcal samplg/actuatg messages be delvered eve whe there s persstet exteral RF terferece, ad mag the commucato l as robust as possble s ofte of top cocer. We shall show that by fully explotg the low data throughput feature, the DSSS-CDMA IC-WLAN scheme ca acheve much hgher level of robustess. We carry out Mote Carlo smulato to compare the robustess betwee DSSS-CDMA ad IEEE b/a schemes. Specfcally, each tral, a layout of the basestato ad remote statos s geerated, together wth the wreless medum stace. The quattatve robustess dcator s J m (.e. the mmum exteral RF terferece power eeded to brea dow at least oe of the wreless cotrol loops, see the defto (25)), whch s calculated ad compared betwee the DSSS-CDMA ad IEEE b/a IC-WLAN schemes. The dustral door evromet s supposed to be a squarearea of 20m 20m, wth the basestato located at the ceter. remote statos are uformly dstrbuted across the room, each correspods to a wreless cotrol loop. The value of vares from 1 to 100. Wthout loss of geeralty, all samplg/actuatg pacets are of 152 bts (same as the verted pedulum case, a typcal cotrol pacet sze), ad all cotrol loops have the same mmum acceptable samplg/actuatg rate f m. Specfcally, Two values of f m are tested: 1Hz ad 10Hz, whch are typcal for dstrbuted dustral cotrol loops. Every applcato layer samplg/actuatg pacet must be delvered wth success probablty of o less tha Foragve, f m ad Table 2. Physcal Layer Settgs for Comparsos Max tras power RF bad DSSS-CDMA vs. IEEE 1watt b comparso 2.449GHz DSSS-CDMA vs. IEEE 800mw a comparso 5.795GHz Accordg to FCC regulato. Accordg to IEEE specfcato. Note RF badwdth also decdes basebad badwdth (.e. chp rate for DSSS ad bt rate for OFDM). IC-WLAN scheme, 200 trals are smulated. I each tral, a stace of remote stato layout ad a stace of the wreless medum are geerated. The wreless medum stace follows the radom model depcted Table 1. To mae far comparsos, physcal layer settgs of wreless devces are summarzed Table 2. Wthout loss of geeralty, the DSSS-CDMA scheme deploys QPSK wth per-ode plot toe for RF modulato/demodulato, receved power of all data chaels are the same, the plot toe s of same power level as a data chael trasmtted from the same ode. For IEEE b, the most robust 1Mbps DBPSK mode s assumed; ad for IEEE a, the most robust 6Mbpsmode s assumed, whch deploysbpsk ad 1/2 covolutoal codg for forward error correcto. For DSSS-CDMA scheme, we assume the hardwaremplemetato-depedet upper boud o processg ga g max s suffcetly large 7, so that the processg ga g for cotrol loop s pced to be g = r c /(f m L pt ) accordg to Proposto 1. Accordg to the gve pacet bt legth (152bt) ad RF badwdth lsted Table 2, whe f m =10Hz ad 1Hz, the correspodg processg ga are 3618 ad for DSSS-CDMA/IEEE b comparso, ad 9868, for DSSS-CDMA/IEEE a comparso 8. For IEEE b/a schemes, the pacet s retrasmtted as may tmes as possble throughout the samplg/actuatg perod so as to crease the chace of successful delvery. The upper boud of DSSS-CDMA BER uder specfed SNR s gve (1). For IEEE b 1Mbps mode, equalty (35) gves the lower boud of BER uder specfed SNR [20]. P 80211b ber 1 2 erfc gpu J (35) where g s the processg ga, P u s the receved sgal power, J s the receved total exteral RF terferece 7 Ths s a practcal assumpto: wth very lttle hardware cost crease, the PN sequece legth ca be creased expoetally, so that the upper boud o processg ga s creased expoetally wth t [18]. 8 Note RF badwdth s decded by chp rate r c, whch s fxed. Ay processg ga g ca be pced, but a bgger g correspods to a slower bt rate r b = r c/g.

10 power, ad erfc s the well-ow complemetary error fucto [20]. The IEEE a 6Mbps mode deploys BPSK ad 1/2 covolutoal code for forward error correcto, whch maes t hard to gve a closed-form BER-SNR formula 9. However, ts PER-SNR relatoshp ca be emprcally derved through Mote Carlo smulato. Based o these BER(PER)-SNR relatoshps, J m of DSSS- CDMA ad IEEE b/a schemes for the same wreless medum, ode layout ad applcato ca be calculated, whch are show Fg. 4(a) ad (b). Ths comparso s pessmstc o the DSSS-CDMA sde because of may reasos: Frst the upper boud of BER s used for DSSS- CDMA scheme, whle for IEEE b/a the lower boud of BER ad emprcal exact PER are used respectvely. Secod, (1), the teded sgal power P u s cluded as part of terferece to provde a (overly) pessmstc estmato o ISI; whle for IEEE b/a, ISI s assumed to be 0. Therefore, the actual tolerable exteral RF terferece power for DSSS-CDMA should be o less tha what s plotted Fg. 4(a) ad (b), ad the actual tolerable exteral RF terferece power for IEEE b/a should be o greater tha what s plotted Fg. 4(a) ad (b). From Fg. 4, t s obvous that DSSS-CDMA ca tolerate much hgher exteral RF terferece power tha correspodg IEEE schemes. Whe f m = 10Hz ad 1Hz, DSSS-CDMA acheves approxmately 10dB ad 20dB mprovemet o robustess tha IEEE b. Compared to IEEE a, the mprovemet o robustess s approxmately 25dB ad 35dB respectvely. Ths s because our DSSS-CDMA scheme fully explots the low data rate feature of dustral cotrol loops by settg processg ga accordg to Proposto 1. We see whe the data rate demad of cotrol loop s smaller (.e. wth smaller f m ), larger processg ga ca be deployed, ad the correspodg tolerable exteral RF terferecepower s creased. It s worth otg that for each fxed applcato settg, the tolerable exteral RF terferece power goes dow whe the umber of cotrol loops () creases. Ths s tutvely correct. Because the Shao boud of formato theory [17] bascally says whe the applcato layer data throughput creases, ad the sgal power s fxed (for dowl, there s always oly oe basestato), the tolerable ose power decreases. 5. Related Wor ad Dscusso Oe of the major purposes of ths paper s to show DSSS- CDMA cellphoe etwor paradgm s more approprate for IC-WLAN compared to the domat IEEE schemes for WLAN. Itutvely, the commucato patters of dustral cotrol loop ad cellphoe voce sesso share 9 A upper boud exsts [30], but t should ot be used for a far comparso or a comparso that s pessmstc o the DSSS-CDMA sde. Mea of 10log (J m / 0.001) (dbm), 10.e. average m jammg power dbm Mea of 10log 10 (J m / 0.001) (dbm),.e., average m jammg power dbm DS CDMA, f m = 1Hz, g = b, f m = 1Hz DS CDMA, f m = 10Hz, g = b, f m = 10Hz (a) Comparso wth IEEE b DS CDMA, f m = 1Hz, g = a, f m = 1Hz DS CDMA, f m = 10Hz, g = a, f m = 10Hz (b) Comparso wth IEEE a Fgure 4. Robustess comparso. J m (watt) s the mmum exteral RF terferece power eeded to brea dow at least oe of the wreless cotrol loops. s the umber of wreless cotrol loops. Note the curves for DSSS-CDMA are lower bouds for J m, whle the curves for IEEE b/a are upper bouds. may smlartes. They are both of low data rate, stable regular traffc, ad last for log durato a sesso-le patter. The ma dfferece s the hgh robustess cocer for dustral cotrol loops, whch calls for better explotato of the low data rate feature to acheve as much robustess as possble. The curret CDMA cellphoe etwor archtectures have ot yet focused o such demad. For example, the recet 3-G stadards [5][12] are more focused o provdg hgher data throughputs (whch meas smaller processg ga) to compete wth IEEE o home/offce applcatos doma. Nevertheless, t would be easy to buld our proposed DSSS-CDMA IC-WLAN o top of the may exstg CDMA cellphoe etwor archtectures, e.g. cdmaoe [1], cdma2000 [5], W-CDMA

11 [12], TD-SCDMA [11] etc. The techologes eeded by our scheme are already mature, specfcally, the capablty of provdg multple recofgurable CDMA chaels, processg ga optos ad power levels are already stadard practces supported by most cotemporary CDMA cellphoe chp sets, such as QualComm CSM6800, CSM6700, CSM5500 [10][24] etc. The major modfcato pedg s to better customze the cofgurable optos ad the resource maagemet strateges accordg to the dustral cotrol eeds. We have show that DSSS-CDMA IC-WLAN ca acheve much hgher robustess tha IEEE WLANs [2][3][4][7], whose robustess levels are fxed, ad ot desged for adverse chael codtos. However, f applcato-depedet processg ga cofgurato s provded for IEEE b, ts robustess ca also be greatly mproved. Ths s exactly the DSSS-TDMA IC-WLAN approach, whch we have already dscussed Secto 2, ad come to the cocluso that t s less preferable tha DSSS- CDMA. However, DSSS-TDMA IC-WLAN s stll a feasble approach, whch merts further study. Recetly, Wreless Persoal Area Networ (WPAN) also emerges as a scheme for smaller etwors tha WLAN. The ma MAC/physcal layer stadards are IEEE [6] (Bluetooth) ad IEEE [8]. Geercally speag, ther MAC schemes are both TDMA based. They are amg at low power short rage commucatos. The the ma focus s to acheve hgher power savg tha IEEE , stead of sgfcatly hgher robustess. Also, at the physcal layer, Frequecy Hoppg Spread Spectrum (FHSS) [31] ad DSSS share smlar aalytcal characterstcs. Uder may crcumstaces, FHSS ad DSSS are terchageable. However, FHSS s less advatageous tha DSSS for ts hardware cost ad system complexty. Ad dgtal wreless FHSS-CDMA (Bluetooth s fudametally a FHSS-TDMA scheme) systems are ot as wdely avalable as DSSS-CDMA systems. I the ed, t s worth otg that we caot acheve fte robustess. The goal s to mata wreless cotrol loop commucatos uder as harsh chael codtos as possble, stead of becomg totally mmue to adverse chael codtos. 6. Cocluso The top cocer for buldg Idustral Cotrol Wreless LAN (IC-WLAN) s robustess. Wreless chael codtos ca vary sgfcatly. Power atteuato may chage drastcally because of large-scale path loss ad fadg. Cotedg RF devces may be tured o accdetally or malcously. For dustral evromets, the stuato s eve worse because of varous EMI sources such as electrc motor ad weldg, ad serous large-scale path loss ad fadg because of heavy obstructos. Wreless cotrol loops must be mataed uder all these adverse chael codtos, stead of bacg off. Ths maes the IEEE WLANs, whch s maly desged for offce/home bursty data traffcs, approprate for IC-WLAN. O the other had, dustral cotrol loop traffcs are mostly regular sustaed traffcs wth extremely low data rates. DSSS techology ca therefore acheve hgh robustess by deployg hgh processg ga. Accordg to fe-graed physcal layer smulatos ad Mote Carlo comparsos, we show that by fully explotg the low data rate feature of dustral cotrol loops, a DSSS IC-WLAN ca provde sgfcatly hgher robustess tha IEEE WLAN. At the MAC layer, ether CDMA or TDMA ca be deployed, however, CDMA s more preferable for ts ease of schedulg, overru solato, ad low overhead for regular sustaed traffcs. Therefore, we clam that by fully explotg low data rate feature of dustral cotrol loops, DSSS-CDMA s a more approprate scheme for IC-WLAN. That s, we bascally pots out a ew applcato doma where the CDMA cellphoe etwor paradgm would preval aga due to ts uque characterstcs. Though some modfcatos are eeded, t promsg to buld our proposed DSSS-CDMA IC-WLAN scheme o top of the may cotemporary CDMA cellphoe etwor archtectures. O the other had, we also foresee a large umber of real-tme QoS ad resource maagemet problems to be addressed for the DSSS-CDMA IC-WLAN scheme. We study the optmal resource cofgurato to acheve maxmum robustess ad mmum utlty loss, ad gve aalytcal closed-form solutos. The resource maagemet problem ca be more complcated whe more maeuverable varables are troduced, such as power allocato, chaels per coecto ad cotrol loops to be cluded the IC-WLAN (see [36] Appedx IV for a dscusso o resource optmzato whe per coecto power allocato s ueve). Geerally speag, DSSS techology provdes the mechasm to combe the may varables, such as data rate, real-tme schedule, utlty, power, umber of cotrol loops ad robustess as a whole. May problems are to be explored, such as effcet plag/optmzato algorthms, capactyboud, utlty boud, robustessboud, coexstece of regular low throughput data traffc ad bursty hgh throughput data traffc etc. Also, the stuato wll be more complcated for multple cells. We are terested carryg out further studes all these drectos. 7. Acowledgemet The frst author s supported by Vodafoe Fellowshp. Ths research s also supported by ( alphabetcal order) MURI N , NSF ANI , NSF CCR-

12 , NSF CCR , NSF CCR , ad ONR N We especally tha Prof. Lu Sha for hs sghtful feedbac that helped to acheve the results of ths wor. We also tha Prof. Veugopal Veeravall, Prof. Bruce Haje, Prof. Chrstoforos Hadjcosts, Prof. Rog Zheg, Taya Creshaw ad aoymous revewers for ther commets. We tha Ng L for provdg assstace o usg J-Sm, ad Qgbo Zhu for provdg assstace o carryg out smulatos o hghperformace computer clusters. Refereces [1] TIA/EIA/IS Std [2] IEEE Std [3] IEEE Std a [4] IEEE Std b [5] TIA/EIA/IS CDMA 2000 Seres, Release A (2000) [6] IEEE Std [7] IEEE Std g [8] IEEE Std [9] Drcl j-sm [10] Qualcomm cdma techologes [11] Td-scdma forum [12] Umts forum [13] G. Ato, S. Crsta, et al. A predcto model for electromagetc terfereces radated by a dustral power drve system. IEEE Trasactos o Idustry Applcatos, 35(4), [14] A. R. S. Baha ad B. R. Saltzberg. Mult-Carrer Dgtal Commucatos: Theory ad Applcatos of OFDM. Kluwer Academc/Pleum Publshers, [15] V. Bharghava, A. Demers, S. Sheer, ad L. Zhag. MACAW: A meda access protocol for wreless LAN s. Proceedgs of the Coferece o Commucatos Archtectures, Protocols ad Applcatos (ACM SIGCOMM 1994), pages , [16] S. Cavaler ad D. Pao. A ovel soluto to tercoect feldbus systems usg IEEE wreless LAN techology. Comput. Stadards Iterfaces, 20(1):9 23, [17] T. M. Cover ad J. A. Thomas. Elemets of Iformato Theory. Wley-Iterscece, [18] R. C. Dxo. Spread Spectrum Systems wth Commercal Applcatos. Wley-Iterscece, Aprl [19] L. Hazo, W. Webb, ad T. Keller. Sgle- ad mult-carrer quadrature ampltude modulato: prcpes ad applcatos for persoal commucatos, WLANs ad broadcastg. Joh Wley & Sos, Ltd., [20] S. Hay. Commucatos Systems. Wley, thrd edto, [21] J. Hesala ad J. Terry. OFDM Wreless LANs: A Theoretcal ad Practcal Gude. Sams Publshg, [22] S. Jag. Wreless commucatos ad a prorty access protocol for multple moble termals factory automato. IEEE Tras. Robot. Automat., 14: , [23] R. B. Keburtz. The step motor - the ext advace cotrol systems. IEEE Trasactos o Automatc Cotrol, 9(1), [24] L. Korowajczu, B. de Souza Abreu Xaver, A. M. F. Flho, et al. Desgg cdma2000 Systems. Wley, [25] B. Kuo. Theory ad Applcatos of Step Motors. West Publshg Compay, [26] A. Muqattash ad M. Kruz. Cdma-based mac protocol for wreless ad hoc etwors. Proceedgs of the 4th ACM Itl Symposum o Moble Ad Hoc Networg ad Computg (MobHoc 2003), pages , [27] E. A. Parr. Idustral Cotrol Hadboo. Idustral Press, thrd edto, [28] N. J. Ploplys, P. A. Kawa, ad A. G. Alleye. Closedloop cotrol over wreless etwors. IEEE Cotrol Systems Magaze, 24(3):58 71, Jue [29] R. Prce ad P. E. G. Jr. A commucato techque for multpath chaels. Proceedgs of the IRE, 46: , [30] J. G. Proas ad M. Saleh. Commucato Systems Egeerg. Pretce Hall, secod edto, [31] T. S. Rappaport. Wreless Commucatos: Prcples ad Practce. Pretce Hall, secod edto, [32] L. Sha, X. Lu, M. Caccamo, ad G. Buttazzo. Ole cotrol optmzato usg load drve schedulg. Proc. of 39th IEEE Coferece o Decso ad Cotrol, 5: , Dec [33] M. K. Smo, J. K. Omura, et al. Spread Spectrum Commucatos Hadboo, Electroc Edto. McGraw-Hll, [34] D. Tse ad P. Vswaath. Fudametals of Wreless Commucato. (to be publshed by) Cambrdge Uversty Press, draft edto, [35] A. J. Vterb. CDMA: Prcples of Spread Spectrum Commucato. Pretce Hall, Aprl [36] Q. Wag, X. Lu, W. Che, W. He, ad M. Caccamo. Techcal report o buldg robust wreless la for dustral cotrol wth dsss-cdma cellphoe etwor paradgm. cdma tr.pdf, [37] H. Ye ad G. Walsh. Real-tme mxed-traffc wreless etwors. IEEE Tras. Id. Electro., 48(5), [38] H. Ye, G. Walsh, ad L. Bushell. Wreless local area etwors the maufacturg dustry. Proc. Amerca Cotrol Cof., pages , [39] F. Zhag. Ivestgato of Electromagetc Iterferece of PWM Motor Drves Automotve Electrcal Systems. Number TR MIT Laboratory for Electromagetc ad Electroc Systems, 1999.