Joint MS-GSC Combining and Multi-User High Speed Data Scheduling

Size: px

Start display at page:

Download "Joint MS-GSC Combining and Multi-User High Speed Data Scheduling"

Ariel Park
5 years ago
Views:

1 Jot MS-GSC Combg ad Mult-User Hg Speed Data Scedulg Slm Be Halma*, Moamed-Slm Alou**, ad Kald A. Qaraqe** * usa Polytecc Scool, La Marsa, usa. ** Departmet of Electrcal Egeerg, exas A&M Uversty at Qatar, Educato Cty, Doa, Qatar. Emal: {alou, ald.qaraqe}@qatar.tamu.edu Abstract Explotg mult-user dversty s oe soluto to effcetly use te wreless medum. e basc prcple s to assg te resources to te user experecg te best cael codtos based o a feedbac formato provded by dfferet users. I ts paper, we propose tree multuser scedulg scemes wc are combatos betwee te swtc ad exame trasmsso sceme te etwor layer ad te jot mmum-selecto geeralzed selecto combg ad adaptve modulato te pyscal layer. We study te performace of te proposed scemes terms of spectral effcecy, feedbac load ad rate, ad average umber of combed ad estmated pats. Usg some selected umercal examples, we also compare te performace of our ewly proposed scemes agast te performace of classcal ad recetly publsed scedulg algortms. I. IRODUCIO Future wreless commucato systems must provde g data rates to respod to te growg demads of multmeda servces. O te oter sde, te wreless cael capacty s a scarce resource tat must be used effcetly to respod to tese desred requremets. Assumg a large umber of actve users a cell, oe way to effcetly use ts resource, tus extractg a certa dversty ga, s optmum scedulg betwee tese multple users experecg dfferet cael codtos. I fact, covetoal commucatos systems (GSM for example, tme s dvded to fxed sze tme slots ad eac of tese tme slots s assged to a fxed user regardless of ts cael codto. However, t s gly probable tat at ay gve pot tme, at least oe of te oter users as a better cael qualty. Mult-user dversty [] [] ga ca be aceved by scedulg te user avg te best cael codtos. I order to maage prorty betwee users, cael qualty formato (CQI must be fed bac to te base stato troug a dedcated upl cael. Suc formato must be sgalled regularly order to respod to cael tme varato. e secod ssue to be dealt wt s ow te cael codtos are exploted by te system. I fact, adaptve modulato ad dversty combg are two mportat tools to respod to g data rates demads. More specfcally adaptve modulato eables te acevemet of g spectral effceces uder good codtos of te wreless cael wle dversty combg maes ger trasmsso rates possble. However, dversty combg s costraed by te lmted lfetme of te user equpmet battery wc as led to te adopto of adaptve dversty combg tecques [] []. Mmum-selecto geeralzed selecto combg (MS-GSC [] comes as oe of tese metods. It s based o te geeralzed selecto combg (GSC sceme. I partcular, t selects te best L c pats amog te L avalable pats but stops te combg operato we reacg te target sgal to ose rato (SR reducg as a cosequece te umber of actve maxmum rato combg (MRC braces ad as suc te processg power cosumpto. It s clear tat adaptve modulato ca beeft from dversty combg troug te mproved cael qualty but te more terestg pot s tat bot adaptve combg ad adaptve modulato use predetermed tresolds ter operato. Based o ts observato, we combe tese two cocepts ad develop ad aalyze ts paper some jot adaptve combg ad g speed data access scemes. We proposed tree scedulg scemes tat are combatos betwee te swtc ad exame trasmsso (SE sceme [] te etwor layer ad te adaptve modulato ad MS-GSC combg te pyscal layer: ( te feedbac effcet sceme ( amg at mmzg te feedbac load o te detrmet of spectral effcecy. ( te badwdt effcet sceme ( amg at acevg g badwdt effcecy wtout tag to cosderato te processg power cosumpto ssue, ad ( te badwdt effcet power greedy sceme amg at acevg te gest badwdt effcecy wt te mmum processg power cosumpto. For tese tree scemes uder cosderato, we evaluate ter average spectral effcecy (quatfed terms of average umber of bts/s/hz, average feedbac load (quatfed terms of te mea of te umber of probed users, te average feedbac rate (quatfed terms of te mea umber of feedbac bts per scedulg operato, ad te processg power cosumpto (quatfed terms of te average umber of estmated ad combed pats per sceduled user. e remader of ts paper s orgazed as follows. e ext secto gves a geeral bacgroud o te cael ad system models. e tree followg sectos deal wt te mode of operato ad te performaces of te proposed scemes. e sxt secto s devoted to some selected umercal results tat llustrate te performace of te scemes uder cosderato. II. GEERAL BACKGROUD A. System Model We cosder a base stato (BS servg K users, eac avg L dversty braces. We cosder a dscrete tme 98 //$. c IEEE ICICS Autorzed lcesed use lmted to: exas A M Uversty. Dowloaded o Jauary, at : from IEEE plore. Restrctos apply.

2 mplemetato for te proposed scemes. I partcular, sort guard perods are perodcally serted to te te trasmtted sgal. Durg tese guard perods te base stato probes eac of te users order to select te most approprate oe accordg to te scedulg sceme. Durg te probg operato, users perform a set of operatos cludg pat estmato, combed SRs comparso, ad modulato dex determato followed by CQI feedbac wc s te bass for te scedulg sceme. After decdg te modulato mode, te trasmtter ad te selected recever are cofgured accordgly trougout te subsequet data burst trasmsso. Cocerg te recever, ad due to complexty ad processg power costrats, we assume tat t ca ot combe more ta Lc dversty pats (L c <L. As suc ts recever cooses te most approprate pats to reac te desred modulato dex uder a certa bt error rate (BER costrat. B. Cael Model o reac mult-pat dversty, we cosder a frequecy selectve cael. e dversty pats represet te dfferet resolvable mult-pats wdebad code dvso multple access W-CDMA or ultra wde bad UWB systems We assume a bloc fadg cael, wc, te sum of te guard perod ad te data burst tme slot s rougly equal to te cael coerece tme. us, durg eac tme slot we experece te same cael codtos,.e te faded sgal ampltude remas rougly costat durg te data burst ad te prevous guard perod. Fally, sgals amog users are assumed to be depedet ad detcally dstrbuted (..d. ad te sgals of te L multpats at eac user ed are assumed to experece Rayleg type of fadg. C. Adaptve Modulato e process of adaptve modulato starts by examg te cael SR at te recever sde. s SR rage s separated to + fadg regos, eac of tem s assocated wt a partcular quadrature ampltude modulato QAM sgal costellato. More specfcally, te dvded SR regos are < + defed by te followg SR tresolds: were s te SR tresold for -QAM. Gve tese SR tresolds ad gve te SR estmated at te recever, ts recever determes te modulato dex, sucas < ad forms va te feedbac pat te trasmtter about te modulato dex to be used. Assumg te M-QAM sgallg, te SR tresolds for a target bt error rate ca be determed usg te followg equato gve [, Eq (8] + = l(ber( ; =,,...,; ( were BER s te specfed target BER. D. otatos I te followg sectos, we wll adopt te followg otatos ad relatos. We deote by Γ = l:l te sum l= of te frst ordered pat SRs :L :L... :L ad ts statstcs are gve [8, capter 9]. e jot probablty desty fucto (PDF of Γ, Γ ad Γ Lc s sow to be gve by p Γ,Γ,Γ Lc (x, y, z =p Γ, :L,Γ + (y, x y, z xu(z xu(x y x U x +(L c (x y z U y (, were U(. s te ut step fucto ad Γ + = Lc l=+ l:l. Fally, te jot PDF of Γ, :L ad Γ +, p Γ, :L,Γ +,s gve [9, Eq (] for L c = L ad [, Eq (9] for L c <L. III. FEEDBACK EFFICIE SCHEME A. Mode of Operato Durg te guard perod, te BS starts by radomly probg oe user amog te K users. s user wll ru te MS-GSC algortm order to determe te maxmum costellato sze t ca reac wt respect to te target BER. More specfcally, at te user level, we operate as follows. Frst, we set as te tresold SR. If ts tresold ca be reaced, te recever sets to te modulato dex. Oterwse, t reduces te modulato dex ad repeats te prevous operato. s process cotues utl te user fds a acceptable modulato dex or t eve fals to reac te SR tresold correspodg to te bary modulato. I te frst case, te user feedbacs b = log ( + bts dcatg te modulato dex, were. s te floor operator. Wle te latter case, t feedbacs oly oe egatve acowledge (ACK bt. At te BS level, f a postve respose s receved from a partcular user, ts user wt te correspodg modulato dex s servced. Oterwse, t probes aoter user wt te same procedure descrbed before. Fally, f te BS does ot fd a acceptable user, t buffers te data. B. Performace Aalyss Average spectral effcecy: Based o te mode of operato of ts sceme, t ca be sow tat te average spectral effcecy, η, s gve by [] η= K = + PΓLc [ ] +( P ΓLc K, ( were P ΓLc (. s te cumulatve dstrbuto fucto (CDF of Γ Lc. Average feedbac load: e average feedbac load, F, ca be sow to be gve by [] F = K ( P ΓLc (PΓLc + K K.( = Average feedbac rate per scedulg operato: Autorzed lcesed use lmted to: exas A M Uversty. Dowloaded o Jauary, at : from IEEE plore. Restrctos apply.

3 e average feedbac rate per scedulg operato R ca be sow to be gve by [] K (b + ( P ΓLc (PΓLc + K K.( Average umber of combed pats per sceduled user: e average umber of combed pats per sceduled user, c, ca be be sow to be gve by [] were Pr [ c =] = K Pr [ c = l] C = L c l= P» P = l Pr [ c = l], ( + +( P Γ, P» P = K Pr [ c = L c]= K = p Γ,Γ,Γ Lc (x, y, zdxdydz + p Γ,Γ,Γ Lc (x, y, zdxdydz( +(P Γl PΓl, <l<l c; P» P + = p ΓLc,Γ Lc (x, ydxdy + PΓLc. Average umber of estmated pats per user It ca be sow tat te average umber of estmated pats per user, E, ca be wrtte as E = L F. (8 K B. Performace Aalyss Average spectral effcecy: e average spectral effcecy, η, of ts sceme ca be sow to be gve by [] K η= = [ + [ ] [ + ] K K ] P ΓLc [ ] + ( PΓ Lc [ ] [ ].(9 Average feedbac load: e average feedbac load, F, ca be sow to be gve by [] F = K = ( P ΓLc [ ](P Γ Lc [ ] + K [ ]K. ( Average feedbac rate per scedulg operato: Based o te mode of operato of ts sceme, te average feedbac rate per scedulg operato, R, ca be sow to be gve by [] K (b + ( P ΓLc [ ](PΓ Lc [ ] + Kb [ ]K.( Average umber of combed pats per sceduled user: e average umber of combed pats per sceduled user, c, ca be sow to be also gve by (, were P Pr [ c =] = K Pr [ c = l] = K P Pr [ c = L K P c]= + K P ( P Γ (P Γl ( P ΓLc., P Γl P ΓLc K, <l<l c, Average umber of estmated pats per user: e umber of estmated pats per user, E, s stll gve by (8. ( IV. BADWIDH EFFICIE SCHEME A. Mode of Operato Durg te guard perod, te BS starts by radomly probg oe user amog te K users. s user rus te MS-GSC algortm settg as te tresold SR wc s te SR eablg to reac te maxmum costellato sze uder a certa BER costrat. If ts user succeeds provdg te maxmum costellato sze, t feedbacs oe bt correspodg to a postve acowledge (ACK ad tus t wll be cose as te served user durg te ext data burst. Oterwse, t feedbacs b = log ( + bts formg te BS of te maxmum costellato sze t ca reac after combg all te GSC L c braces. I te secod case, te BS probes aoter user wc performs te same operato as te prevous user. s procedure cotues utl te BS fds a user avg a combed SR greater ta te or all te users fal to aceve tat codto. I te latter case te BS cooses te user wt te maxmum provded costellato sze, ad f tere s a outage, t buffers te data. V. BADWIDH EFFICIE POWER GREEDY SCHEME A. Mode of Operato I a frst step, te BS probes te users for te gest modulato dex. I te begg, te BS starts by radomly probg oe user amog te K users. s user rus te MS- GSC algortm settg as te tresold SR. If ts user succeeds provdg te maxmum costellato sze, t feedbacs oe bt correspodg to a ACK ad t s served by te BS te ext data burst. Oterwse, t feedbacs oe bt correspodg to a ACK. I te latter case, te BS probes aoter user wc rus as well te MS-GSC algortm wt te tresold correspodg to te maxmum costellato sze ad te feedbacs te approprate acowledge. s operato cotues utl te BS fds a user avg a combed SR greater ta or all te users fal to aceve tat codto. I te latter case te BS reduces te costellato sze (.e. t sets as te ew tresold SR our case ad repeats te probg operato as descrbed before for ts ewly set tresold. e BS cotues probg utl a acceptable user wt approprate cael codtos s selected. I te case of a outage, te BS buffers te data. Autorzed lcesed use lmted to: exas A M Uversty. Dowloaded o Jauary, at : from IEEE plore. Restrctos apply.

4 B. Performace Aalyss Average spectral effcecy ad feedbac load: Based o te mode of operato of ts sceme, te average spectral effcecy ad feedbac load are exactly te same as tose of te sceme. Average feedbac rate per scedulg operato: Based o te mode of operato of ts sceme, t ca be sow tat Pr[ ] =P ΓLc [ ]( ( P ΓLc [ ], Pr[ K + ]=P ΓLc [ + ] (K P ΓLc [ ] ( [ + ] P ΓLc [ ], =,..., ; =,...,K ; Pr[ K] =P ΓLc [ ](K [ ] PΓ Lc [ ] +P ΓLc [ ]K, ad te resultg average feedbac rate s tus gve by K = = ( (K + Pr[ K + ]. ( Average umber of combed pats per sceduled user: It ca be sow tat te average umber of combed pats per sceduled user, c, s gve by (, were t ca be sow tat [] Pr [ c =] = P K P [ ] + K = R R + pγ,γ,γ (x, y, zdxdydz Lc +( P Γ K P, Pr [ c = l] = P K P [ ] + K = R R + p Γ,Γ,Γ Lc (x, y, zdxdydz + K P (P Γl P Γl Pr [ c = L P K P c]= [ ] + K = R R + p ΓLc,Γ Lc (x, ydxdy + P ΓLc K P (, <l<l c,. Average umber of estmated pats per user: e average umber of estmated pats per user, E,salso gve by (8. VI. UMERICAL RESULS I ts secto we wll mae a comparso of te tree proposed scemes as well as a comparso betwee tese ew scemes ad wt te recetly publsed oes. Fg. sows tat te spectral effcecy of te sceme s lower ta te spectral effcecy of te orzotal swtc ad exame trasmsso (HSE []. Fg. sows tat te average umber of estmated pats per user for te sceme s greater ta tat of te vertcal swtc ad exame trasmsso (VSE []. I fact, accordg to te mode of operato of te MS-GSC, t s ecessary to estmate ad ra all te L avalable pats weever we probe a ew user, cotrarly to wat s performed te VSE wc wll estmate just as may pats as eeded (but ot ecessarly combe te best oes. Fg. sows also tat te ad scemes aceve te same spectral effcecy ad te best possble comparso wt HSE ad. However, ts comes at te expese of a g feedbac load ad umber of estmated pats per user as llustrated Fg. ad Fg.. Seeg tat te ad scemes provde te maxmum spectral effcecy, we also compare tem wt a sceme avg te same mode of operato but wc adopts MRC stead of MS-GSC. I oter words, we a user s probed, t performs full L-pats MRC combg operato to determe te spectral effcecy tat ca be reaced. As t s sow Fg., te performace of te tree scemes are ot very dfferet terms of spectral effcecy. However, as llustrated Fg., te ad scemes offer cosderable savgs terms of umber of combed pats we compared to MRC. I addto t s clear from Fg. tat te sceme reduces cosderably te umber of combed pats compared to te sceme ad s as suc terestg from a battery power effcecy perspectve. Fally, we ca see from Fg. tat te sceme aceves te best performace from a feedbac perspectve but ts comes at te expese of a cosderable loss terms of spectral effcecy, as llustrated Fg.. REFERECES [] D. Gesbert ad M.-S. Alou, Selectve mult-user dversty, Proc. IEEE Iteratoal Symp. o Sgal Processg ad Iformato ecology (ISSPI, Dresde, Germay, Dec., pp.. [] B. Holter, M.-S. Alou, G. E. Øe, ad H.-C. Yag, Multuser swtced dversty trasmsso, Proc. IEEE Ve. ecol. Cof. (VC, vol., Los Ageles, Calfora, USA, Sept., pp. 8. [] Y. A. Hart, A. ewf, ad M.-S. Alou, Multuser dversty wt quatzed feedbac, IEEE ras. Wreless Commu., vol., pp., Ja.. [] M.-S. Alou ad H.-C. Yag, Mmum estmato ad combg geeralzed selecto combg (MEC-GSC, Proc. IEEE It. Sym. Iform. eory (ISI, Adelade, Australa, Sept., pp.. [] H.-C. Yag, ew results o ordered statstcs ad aalyss of mmumselecto geeralzed selecto combg (GSC, IEEE ras. Wreless Commu., vol., pp. 8 88, July. [] H.-C. Yag ad M.-S. Alou, MRC ad GSC dversty combg wt a output tresold, IEEE rasactos Vecular ecology, vol., pp. 8 9, May. [] M.-S. Alou ad A. J. Goldsmt, Adaptve modulato over aagam fadg caels, Kluwer J. Wreless Commucatos, vol., pp. 9, May. [8] M. K. Smo ad M.-S. Alou, Dgtal Commucato over Geeralzed Fadg Caels: a Ufed Approc to Performace Aalyss. ew Yor, Y: Jo Wley & Sos,. [9] H.-C. Yag,. Belaj, ad M.-S. Alou, Performace aalyss of jot adaptve modulato ad dversty combg over fadg caels, IEEE ras. Commu, vol., pp. 8, Mar.. [] S. Co, M.-S. Alou, K. A. Qaraqe, ad H.-C. Yag, Soft adover overead reducto by RAKE recepto wt fger replacemet, Proc. IEEE Iteratoal Cof. o Commu. Systems (ICCS, Sgapore, ov.. [] S. B. Halma, M.-S. Alou, ad K. A. Qaraqe, Jot MS-GSC combg ad g speed dowl pacet access scedulg, submtted to IEEE ras. Wreless Commu. [] K.-H. Par, Y.-C. Ko, ad M.-S. Alou, Jot adaptve combg ad multuser dow-l scedulg, Proc. IEEE Vecular ecology Cof. (VC, Dubl, Irelad, Apr., pp.. Autorzed lcesed use lmted to: exas A M Uversty. Dowloaded o Jauary, at : from IEEE plore. Restrctos apply.

5 9 Average spectral effcecy (bt/s/hz 8 MRC HSE VSE Average SR per Pat (db Fg.. Average spectral effcecy versus te average SR for te MRC,,, HSE,, ad VSE scemes we L=, L c =, K=, =8, ad BER =.. Average umber of Combed pats MRC Average SR per Pat (db Fg.. Average umber of combed pats per sceduled user versus te average SR for te,,, ad MRC scemes we L=, L c =, K=, =8, ad BER =.. Average feedbac load 8 Average SR per Pat (db Fg.. Average feedbac load versus te average SR for te,, ad scemes we L=, L c =, K=, =8, ad BER =.. Average feedbac rate per scedulg operato 8 Average SR per Pat (db Average umber of estmated pats per user HSE VSE Average SR per Pat (db Fg.. Average umber of estmated pats per user versus te average SR for te,, HSE,, ad VSE scemes we L=, L c =, K=, =8, ad BER =.. Fg.. Average feedbac rate per scedulg operato versus te average SR for te,, ad scemes we L=, L c =, K=, =8, ad BER =.. Autorzed lcesed use lmted to: exas A M Uversty. Dowloaded o Jauary, at : from IEEE plore. Restrctos apply.

Module 6. Channel Coding. Version 2 ECE IIT, Kharagpur

Module 6. Channel Coding. Version 2 ECE IIT, Kharagpur Module 6 Chael Codg Lesso 36 Coded Modulato Schemes After readg ths lesso, you wll lear about Trells Code Modulato; Set parttog TCM; Decodg TCM; The modulated waveform a covetoal ucoded carrer modulato