Optimal Adaptive Data Transmission over a Fading Channel with Deadline and Power Constraints
|
|
- Sophie Lester
- 5 years ago
- Views:
Transcription
1 Opmal Adapve Daa Transmsson over a Fadng Channel wh Deadlne and Power Consrans Muraza Zafer and yan Modano aboraory for Informaon and Deson Sysems Massahuses Insue of Tehnology Cambrdge, MA 239, USA emal:{muraza,modano}@m.edu Absra We onsder opmal rae onrol for energy-effen daa ransmsson over me-varyng (fadng hannels wh sr deadlne onsrans. Spefally, he senaro onsss of a ransmer wh B uns of daa ha mus be ransmed by deadlne T over a wreless hannel. The ransmer an onrol he ransmsson rae over me by varyng he ransmsson power subje o expeed shor-erm power lms. The expended power depends on boh he (hosen ransmsson rae and he presen hannel ondon and he objeve s o adap he rae over me and n response o he hangng hannel ondons so ha he oal energy os s mnmzed. We presen a novel onnuous-me formulaon of he problem; usng sohas onrol heory and lagrangan dualy, we oban explly he opmal rae onrol poly. We hen presen an llusrave smulaon example omparng he energy oss of he opmal and he full power poles. I. INTRODUCTION Daa serves n modern ommunaon sysems are evolvng from radonal emal and web daa ransfers o more enhaned serves suh as vdeo and real-me mulmeda sreamng [, delay onsraned daa/fle ransfers, hgh hroughpu web aess, and, Voe-over-IP (VoIP whh brngs voe ommunaon no he realm of daa serves (eg. xv-do, WMAX. All hese advanemens requre enhaned Qualy of Serve (QoS whh ranslaes no srer delay and hroughpu requremens on ommunaon. Communaon over wreless hannels adds anoher dmenson of omplexy assoaed wh me-varyng hannel ondons and sary of resoures. Among oher resoure lmaons, energy onsumpon s an mporan onern n sysem desgn and s an ave area of researh n wreless neworks [2. nergy effeny has numerous advanages n effen baery ulzaon of moble deves, nreased lfeme of sensor and ad-ho neworks, and superor ulzaon of lmed energy soures n saelles. As ransmsson energy onsues he bulk of he ommunaon energy expendure, s mperave o mnmze hs os o aheve sgnfan energy savngs, heneforh, our fous n hs paper wll be on ransmsson energy os. In hs work, we address he above wo ssues under a spef seng and oban opmal poles o ransm daa wh delay onsrans over a fadng hannel. Ths work was suppored by NSF ITR gran CCR-3254, by DARPA/AFOSR hrough he Unversy of Illnos gran no. F and by ONR gran number N4664. Modern wreless deves are equpped wh hannel measuremen and rae adapve apables [3. Channel measuremen allows he ransmer-reever par o measure he fade sae usng a pre-deermned plo sgnal whle rae onrol apably allows he ransmer o adjus he relable ransmsson rae over me. Suh a onrol an be aheved n varous ways ha nlude adjusng he power level, symbol rae, odng rae/sheme, onsellaon sze and any ombnaon of hese approahes; furher, he reever an dee hese hanges drely from he reeved daa whou he need for an expl rae hange onrol nformaon [4. Wh presen ehnology, ransmsson rae an be adaped very rapdly n me over mllseond duraon me-slos [3. These apables, hus, provde a unque opporuny o ulze dynam rae onrol algorhms o opmze sysem performane. For a ransmer-reever par, he power-rae funon defnes he relaonshp ha governs he amoun of ransmsson power requred o relably ransm a a eran rae. Two fundamenal aspes of hs funon, whh are exhbed by mos enodng/ommunaon shemes and hene are ommon assumpons n he leraure [5 [, are as follows. Frs, for a fxed b error probably and hannel sae, he requred ransmsson power s a onvex funon of he ommunaon rae as shown n Fgure (a. Ths mples (from a sraghforward applaon of Jensen s nequaly ha ransmng daa a low raes over longer duraon s more energy effen as ompared o hgh rae ransmssons. Seond, he wreless hannel s me-varyng whh shfs he onvex power-rae urves as a funon of he hannel sae as shown n Fgure (b. As good hannel ondons requre less ransmsson power, explong hs varably over me by adapng he rae n response o he hannel ondons leads o redued energy os. Thus, ulzng rae onrol apables and he above wo aspes of power-rae urves, we an mnmze energy os whle also sasfyng delay onsrans. We onsder a ransmer wh B uns of daa ha mus be ransmed by deadlne T over a wreless hannel. The hannel sae (fadng s sohas and modelled as a general Markov proess. The ransmer an onrol he ransmsson rae over me by varyng he ransmsson power subje o shor-erm power lms. The expended power depends on boh he hosen ransmsson rae and he presen hannel ondon /6/$2. 26 I 93
2 Power, P Fgure (a onvex nreasng rae, r Power, P Fgure (b onvex nreasng mprovng hannel sae rae, r Fg.. Transmsson power as a funon of he rae and he hannel sae; (a fxed hannel sae, (b varable hannel sae. (Fgure (b. The objeve a he ransmer s o dynamally adap he rae over me and n response o he hangng hannel ondons suh ha he ransmsson energy os s mnmzed and he deadlne onsran s me. We formulae he problem n onnuous-me; usng he sohas onrol heory and lagrangan dualy, we oban smple expl/losedform formulas for he opmal ransmsson rae as a funon of he amoun of daa remanng, he presen me relave o he deadlne and he presen hannel sae. Transmsson power/rae onrol s an ave area of researh n ommunaon neworks n varous dfferen onexs. Adapve nework onrol and shedulng has been suded n he onex of nework sably [3, average hroughpu [4, average delay [5, [ and pake drop probably [5. Ths leraure onsders average mers ha are measured overannfne me horzon and hene do no drely apply for delay onsraned/real-me daa. Furher, ransmsson adapaon smply based on seady sae dsrbuons does no suffe and o onsder sr deadlnes one needs o ake no aoun he sysem dynams over me, hus, nrodung new hallenges and omplexy no he problem. Reen work n hs dreon nludes [6 [9, [, [2. The work n [6 suded offlne formulaons under non-ausal knowledge of fuure hannel saes and devsed heurs onlne poles usng he opmal offlne soluon. The works n [8, [9, [ suded formulaons whou fadng and n parular our work n [ used a alulus approah o oban mnmum energy poles wh general arrval urves and QoS onsrans. Ths paper s an exenson of he work n [2 where we onsdered he same problem whou expl expeed power onsrans. The addonal omplexy arsng due o he power lm onsrans s addressed usng a lagrangan dualy approah. II. PROBM STUP We onsder a onnuous-me model of he sysem. Clearly, suh a model s an approxmaon of he aual sysem bu he assumpon s jusfed sne n prae he ommunaon slo duraons (on he order of mse are muh smaller han pake delay requremens (on he order of s of mse; hus, one an vew he sysem as vrually operang n onnuous-me. Suh a model makes he problem mahemaally raable and yelds smple soluons. To apply he resuls obaned here, one would hen smply dsreze he soluon as done for he smulaons n Seon IV. A. Transmsson Model e h denoe he hannel gan, P ( he ransmed sgnal power and P rd ( he reeved sgnal power a me. We make he ommon assumpon [5, [6, [8 [ ha he requred reeved sgnal power for relable ommunaon (wh a eran fxed b-error probably s onvex n he rae,.e. P rd ( =g(r(. Sne he reeved sgnal power s gven as P rd ( = h 2 P (, he requred ransmsson power o aheve rae r( s gven by, P ( = g(r( ( ( where ( = h 2 and g(r s a non-negave onvex nreasng funon for r. The quany ( s referred o as he hannel sae a me. Is value a me s assumed known eher hrough predon or dre hannel measuremen bu evolves sohasally n he fuure. I s worh emphaszng ha he relaonshp n ( nludes muh more generaly han dsussed above. For example, ( ould n fa represen a ombnaon of sohas varaons n he sysem and (unonrollable nerferene from oher ransmer-reever pars, as long as he power-rae relaonshp n ( holds. We furher assume ha g(r belongs o he lass of monomal funons, namely, g(r =kr n,n>,k > (n, k R. Whle hs assumpon resrs he heoreal generaly of he problem, serves several purposes. Frs, leads o smple losed form soluons ha an be appled n prae. Seond, for mos praal ransmsson shemes, g( s desrbed numerally and s exa analyal form s unknown. In suh suaons, one an oban he bes approxmaon of ha funon o a monomal funon and apply he resuls hus obaned. Thrd, monomals form he frs sep owards sudyng a polynomal g( whh would hen apply o a general g( usng he polynomal expanson. Also, noe ha for values of n lose o, g(r models a lnear power-rae urve whh s a wdely suded model. Fnally, whou loss of generaly hroughou he paper we ake he onsan k =, sne, as wll be evden, any oher value of k smply sales he problem whou affeng he resuls. B. Channel Model We onsder a general onnuous me dsree sae spae Markov model for he hannel sae proess. Markov proesses onsue a large lass of sohas proesses ha exhausvely model a wde se of fadng senaros and here s subsanal leraure on hese models [6, [7 and her applaons o ommunaon neworks [7, [8. Denoe he hannel sae proess as C( and he sae spae as C. e ( denoe a sample pah and = (, Cbe a parular realzaon a me. Sarng from sae, lej be he se of all saes ( o whh he hannel an ranson when he sae hanges. e λ denoe he hannel ranson rae from sae o, hen, he sum ranson rae a whh he hannel jumps ou of sae s, λ = J λ. Clearly, he expeed me ha C( spends n sae s /λ and one an vew λ as he oherene me of he hannel n sae. 932
3 queue B Fg. 2. server ( daa n buffer B x( Shema desrpon of he sysem. Now, defne λ= sup λ and a random varable, Z(, as, { Z( = /, wh prob. λ /λ, J (2, wh prob. λ /λ Wh hs defnon, we oban a ompa desrpon of he proess evoluon as follows. Gven a hannel sae, heres an xponenally dsrbued me duraon wh rae λ afer whh he hannel sae hanges. The new sae s a random varable whh s gven as C = Z(. Clearly, from (2 he ranson rae o sae J s unhanged a λ, whereas wh rae λ λ here are ndsngushable self-ransons. Noe ha here s no generaly los wh hs new desrpon as yelds a sohasally denal senaro and he selfransons are ndsngushable over any sample pah. The represenaon smply helps n noaonal onvenene xample: Consder he sandard Glber-llo hannel model [7 ha has wo saes b and g denong he bad and he good hannel ondons respevely. The wo saes orrespond o a wo level quanzaon of he hannel gan. If he measured hannel gan s below some value, he hannel s labelled as bad and ( s assgned an average value b, oherwse ( = g for he good ondon. e he ranson rae from he good o he bad sae be λ gb and from he bad o he good sae be λ bg.eγ = b / g, and usng he earler noaon, λ = max(λ bg,λ gb. For sae g we have, { γ, wh prob. λ gb /λ Z( g = (3, wh prob. λ gb /λ To oban Z( b, replae γ wh /γ and λ gb wh λ bg n (3. C. Problem Formulaon As menoned earler, he ransmer has B uns of daa and a deadlne T by whh he daa needs o be sen. e x( denoe he amoun of daa lef n he queue and ( be he hannel sae a me. The sysem sae an be desrbed as (x,,, where he noaon means ha a me, wehavex( =x and ( =. er(x,, denoe he hosen ransmsson rae for he orrespondng sysem sae (x,,. Sne he underlyng proess s Markov, s suffen o resr aenon o ransmsson poles ha depend only on he presen sysem sae [2. Clearly hen, (x,, s a Markov proess. The sysem s deped n Fgure 2. Gven a poly r(x,,, he sysem evolves n me as a Peewse-Deermns-Proess (PDP as follows. We are gven x( = B and ( =. Unl, where s he frs me nsan afer =a whh he hannel hanges, T me he buffer s redued a he rae r(x(,,. Hene, over he nerval [,, x( sasfes he ordnary dfferenal equaon, dx( = r(x(,, (4 d quvalenly, x( =x( r(x(s,,sds, [,. Now, sarng from he new sae (x(,,, he above proedure repeas unl = T s reahed. A me T, he daa ha mssed he deadlne (amoun s assgned a g(/ penaly os of for some >. Ths peular os an be vewed n he followng wo ways. Frs, smply represens a spef penaly funon where an be adjused and n parular made small enough so ha he daa ha msses he deadlne s small. Ths wll ensure ha wh good soure-hannel odng, he enre daa an be reovered even g(/ f msses he deadlne. Seond, noe ha s he amoun of energy requred o ransm daa n me wh he hannel sae beng. Thus, s he small me wndow n whh he remanng daa s ompleely ransmed ou assumng ha he hannel sae does no hange over ha perod. In fa, vewng T + as he aual deadlne, hen models a small buffer wndow n whh unlmed power an be used o mee he deadlne, albe a an assoaed os. e he nerval [,T be paroned no equal perods 2 and denoe P as he shor-erm expeed power onsran a he ransmer. Then, over eah paron he power [ onsran requres ha he expeed energy os,, g(r(x(s,,sds s less han P (T/. Noe ha T/ s he duraon of eah paron nerval. Clearly, by varyng, he me sale of he paron an be vared and he power onsran an be made eher more or less resrve. A ransmsson poly r(x,, mus also sasfy he followng addonal requremens, (a r(x,, <, (non-negavy (b r(x,, =,fx =(no daa lef o ransm 3. e Φ denoe he se of r(x,, ha sasfy he above requremens. We say ha a poly r(x,, s admssble f r(x,, Φ and sasfes he power onsran on all he paron nervals. Denoe he opmzaon problem as (P; an now be summarzed as follows, (P nf r( Φ subje o g(r(x(s,,sds + g( g(r(x(s,,sds PT T (. g(r(x(s,,sds PT Sne g( s srly onvex, makng smaller nreases he penaly os. 2 xensons o arbrary szed parons s farly sraghforward and suh a generaly s omed for mahemaal smply. 3 Addonal ehnal requremens are, r( be loally lpshz n x (x > and peewse onnuous n, o ensure ha (4 has a unque soluon. 933
4 The expeaons above are ondoned on (x, 4, he sarng values a =. For he analyss, we wll keep he general noaon x bu s value n our ase s smply, x = B. III. OPTIMA POICY We onsder a lagrangan dualy approah o solve he problem n (P. The bas seps nvolved n suh an approah are o form he lagrangan, oban he dual funon ha depends on he lagrange mulplers, maxmze he dual funon wh respe o he lagrange mulplers and show ha here s no dualy gap; ha s, maxmzng he dual funon gves he opmal os for he onsraned problem. However, here are mporan sublees n problem (P whh make non-sandard. Frs, he doman of he rae funons r( s a funonal spae whh makes (P an nfne dmensonal opmzaon, and, seond, (P s a sohas opmzaon and by hs we mean ha here s a probably spae nvolved over whh he expeaon s aken. In hs seon, we delve no he soluon deals proeedng along he seps oulned above. A. Dual Funon The nequaly onsrans n (P an be wren as, [ kt (k T g(r( ds PT, k =,..., (5 e ν = (ν,...,ν be he lagrange mulplers for hese power onsrans orrespondng o he parons of [,T. Sne hese are nequaly onsrans, he lagrange mulplers mus be non-negave,.e. ν,...,ν. Thelagrangan s hen gven as, g(r( H(r(, ν = ds + g( + [ kt ν k ( k= (k T g(r( ds PT Re-arrangng he above equaon, an be wren n he form, [ T ( + ν(sg(r( g( H(r(, ν = ds + (ν ν (PT/ (7 where ν(s akes value ν k over he k h paron nerval,.e. ν(s =ν k,s [ (k T, kt. As s he ase n dualy heory, he dual funon s he nfmum of H(r(, ν over Φ. The pon o noe here s ha he r( over whh hs mnmzaon s onsdered do no have o sasfy he power onsrans, hough oher requremens sll apply. Ths s beause he shor erm power onsrans (volaon have been added as a os n he objeve funon of he dual problem. Denong he dual funon as ( ν, we hus have, ( ν = nf H(r(, ν (8 r( Φ 4 To avod beng umbersome on noaon, we wll, hroughou, represen ondonal expeaon whou an expl noaon bu raher menon he ondonng parameer when here s ambguy. (6 One of he neresng properes of he dual funon s ha gves a lower bound o he opmal os n (P. Ths sandard propery s referred o as weak dualy and apples n our ase as well. I s summarzed n he lemma below; he proof s dre and omed for brevy. emma : e (x, be he sarng sae a =and denoe J(x, as he opmal os for (P. Then, for all ν, wehave,( ν J(x, Before we proeed o srong dualy whh nvolves maxmzng ( ν over ν, we solve he mnmzaon n (8 and oban he dual funon. valuang he dual funon: The approah we adop o evaluae he dual funon s o vew he problem n sages orrespondng o he parons and solve for he opmal rae funons n eah of he paron nerval wh he neessary boundary ondons a he edges. An mmedae observaon from (7 shows ha he effe of he lagrange mulplers s o mulply he nsananeous power funon g(r( wh a me-varyng funon ( + ν(s. Thus, he dfferene over he varous nervals s n a dfferen mulplave faor o he os funon, whh for he k h nerval s, +ν(s =ν k. Sne (8 nvolves a mnmzaon over r( for fxed lagrange mulplers ν, he seond erm n (7,.e. (ν+...+νpt, s rrelevan for he mnmzaon and we wll negle for now. Defne, Hν(x, r, = ( + ν(sg(r( ds + g( (9 H ν (x,, = nf r( Φ Hr ν(x,, ( where he expeaon n (9 s ondoned on he sae (x,,. Saed smply, Hν(x, r, s he os-o-go funon for poly r(, sarng from sae (x,, and H ν (x,, s he opmal os-o-go funon sarng from sae (x,,. Relang bak o (7, Hν(x r,, s he expeaon erm n (7 and H ν (x,, s he mnmzaon of hs erm over Φ. Clearly from (7 and (8, havng solved for H ν (x,,, we hen oban he dual funon as smply, ( ν =H ν (x,, (ν +...+ν PT. In he proess of obanng H ν (x,,, we wll also oban he opmal rae funons for he lagrange mulpler ν. Now, fous on he k h paron nerval so ha [ (k T, kt and onsder a small nerval [, +h, whn hs paron. e some poly r( be followed over [, + h and he opmal poly hereafer, hen usng Bellman s prnple [9 we have, H ν (x,, =mn r( { +h ( + ν k g(r(x(s,,s ds } +H ν (x +h, +h,+ h ( where x +h s shor-hand for x( + h and he expeaon s ondoned on (x,,. The lef sde above s he opmal os f he opmal poly s followed rgh from he sarng sae (x,,, whereas on he rgh sde, he expresson whn he mnmzaon brake s he oal os wh poly r( 934
5 beng followed over [, +h and he opmal poly hereafer. Removng he mnmzaon gves he nequaly, +h ( + ν k g(r(x(s,,s H ν (x,, ds + [H ν (x +h, +h,+ h (2 Rearrangng, dvdng by h and akng he lm h gves, A r H ν (x,, + ( + ν kg(r (3 ( (+νk g(r( ds The above follows sne +h s h (+ν kg(r where r s he value of he ransmsson rae a me,.e. r = r(x,,, and, A r H ν (x,, s defned as A r H H ν (x,, = lm ν(x +h, +h,+h H ν(x,, h h. The quany A r H ν (x,, s alled he dfferenal generaor of he Markov proess (x(,( for poly r( and nuvely, s a naural generalzaon of he ordnary me dervave for a funon ha depends on a sohas proess. An elaborae dsusson on hs op an be found n [9 [2. For our ase, usng he me evoluon as n (4, he quany A r H ν (x,, an be evaluaed as, A r H ν (x,, = H ν r H ν x +λ( z [H ν (x, Z(, H ν (x,, (4 where z s he expeaon wh respe o he Z varable. Now, n he above seps from (2-(3 f poly r( s replaed wh he opmal poly r (, here s equaly hroughou and we ge, A r H ν (x,, + ( + ν kg(r = (5 Hene, for a gven sysem sae (x,,, he opmal ransmsson rae, r, s he value ha mnmzes (3 and he mnmum value of he expresson equals zero. Thus, over he k h paron nerval wh [ (k T, kt,wegehe followng opmaly ondon, [ ( + νk g(r mn r [, + A r H ν (x,, = (6 Subsung A r H ν ( from (4, we see ha (6 s a paral dfferenal equaon n H ν (x,,, also referred o as he Hamlon-Jaob-Bellman (HJB equaon. mn r [, { ( + νk g(r + H ν r H ν x } +λ( z [H ν (x, Z(, H ν (x,, = (7 In summary, he above argumens sae ha f H ν ( s suffenly smooh, sasfes he opmaly ondon n (7 over all he paron nervals k =,..., and he opmal rae funons are he orrespondng r values ha mnmze (7. The boundary ondons for H ν ( are as follows. A = T, H ν (x,, T = g( x, sne sarng n sae (x, a me T, he opmal os smply equals he penaly os. Over eah paron nerval, we requre ha H ν ( s onnuous a he edges, so ha he funons evaluaed for he varous nervals are onssen. An mporan avea o noe s ha he opmaly ondon alone doesn suffe and we need suffeny argumens o verfy ha a soluon of he PD n (7 s he opmal soluon. Ths s ndeed he ase, bu, he verfaon heorems are very dealed and omed for brevy. We now presen he resuls for he funon, H ν (x,,, and he orrespondng opmal rae funon, denoed as rν(x,,, where he subsrp ν s used o ndae expl dependene on he lagrange mulplers, ν. Theorem I gves he soluon whh s furher explaned laer, bu frs, we need some addonal noaon regardng he hannel proess. e here be m hannel saes n he Markov model and denoe he varous saes Cas, 2,..., m. Gven a hannel sae,he values aken by he random varable Z( (defned n (2 are denoed as {z j }, where z j = j /. The probably ha Z( =z j s denoed as p j. Clearly, f here s no ranson from sae o j, p j =. Theorem I: Consder he mnmzaon n ( wh g(r = r n, (n >,n R. For k =,..., and [ (k T, kt (kh paron nerval, H ν (x,, = ( + ν kx n (f k, =,...,m (8 ( n rν(x, x, = f k, =,...,m (9 ( where over he k h nerval, {f k (} m = s he soluon of he followng OD sysem, (f k ( = λf k ( n + λ m p j (f k ( n n z j (f k (2 ( n j. (f k m( = λf k m( n + λ n j= m j= p mj (f k m(n z mj (f k ( n (2 j The followng boundary ondons apply; f k =, f (T = ( + ν n, (a he deadlne and f k =,..,, ( f k kt ( = +νk n +ν k+ f k+ ( kt, (a he paron boundares. The dual funon n (8 s hen gven as, ( ν = ( + ν x n (f ( n (ν ν PT (22 Proof: The proof s omed for lengh onsderaons bu an be heked ha he soluon sasfes he opmaly ondon. The above soluon an be undersood as follows. There s a se of funons {f k (} for he parons and he m hannel saes;.e. for eah paron nerval, k, here are m funons {f k (} m = for he orrespondng hannel saes. The subsrp refers o he hannel sae whle he supersrp refers o he paron nerval. Now, gven ha he presen me les n he k h nerval, he rae funon has he smple losed form x expresson f k ( as gven n (9 whle H ν( s as gven n (8. The funons {f k (} m = for he kh nerval are he 935
6 soluon of he sysem of OD n (2-(2 wh he boundary ( ondon a he rgh edge of he nerval gven as, f k kt = ( +νk n ( kt +ν k+. Ths ensures ha Hν (x,, s on- f k+ nuous a he paron edge, = kt. For he h nerval he boundary ondon s, f (T = ( + ν n ; hs ensures ha a = T, H ν (x,,t= g( x = xn.now, n he funons {f k (} an be evaluaed sarng a he h nerval o oban {f (} m = and hen {f (} m = usng he boundary ondon above and proeedng bakwards o he frs nerval. In omplee generaly, a losed form soluon for he sysem of OD as gven above s dfful o oban, however, here s a speal ase whh an be solved n losed form as dsussed nex. Neverheless, n he general ase, he sysem of OD an be easly solved numerally usng sandard ehnques wh mnmal ompuaonal requremen. An mporan pon o noe s ha hs ompuaon needs o be done offlne before he sysem operaon. One he {f k (} are known, he losed form sruure of he poly n (9 warrans no furher ompuaon. Consan Drf Channel Model: Under a speal sruure n he Markov hannel model whh we refer o as he onsan drf hannel, he funons f k ( are ndependen of he hannel sae (.e. f k ( =f k (, and he ommon funons {f k (} k= an be obaned n losed form. The parular assumpon on he hannel model s ha he expeed value of /Z( s ndependen of he hannel sae,.e. [/Z( = β (a onsan. Sne = Z(, sarng n sae, he nex ranson sae sasfes [ = [ Z( = β/. Thus, f we look a he proess /(, he above assumpon means ha over he nerval of neres, he expeed value of he nex sae (gven he presen sae / s a onsan mulple of he presen sae. We refer o β as he drf parameer of he hannel proess. If β>, he proess /( drfs upwards n an expeed sense, f β =, here s no expeed drf and f β<, he drf s downwards. In prae, hs ould be a good model for slow fadng hannels whh over he deadlne nerval are drfng owards mprovng or worsenng ondons. Theorem II: Consder he mnmzaon n ( wh g(r =r n and he onsan drf hannel model wh parameer β. Fork =,...,, [ (k T, kt, H ν (x,, = ( + ν kx n (f k ( n (23 rν(x, x, = f k (24 ( e η = λ(β n, hen, f k ( =( + ν k n e η(t + { k ( +νk η +ν j= j T ( j (e η( e η( T ( j } + ( e η( kt η B. agrange Dualy From emma, we see ha gven a lagrange veor ν, he dual funon s a lower bound o he opmal os of he onsraned problem, P. Thus, makes sense o maxmze ( ν over ν. Theorem III below, saes ha srong dualy holds or ha maxmzng ( ν over ν gves he opmal os of P, and, ha f P has an opmal poly, hen he opmal rae funon s he same as ha obaned n Theorem I wh ν = ν, he maxmzng lagrange veor. As n emma, le J(x, be he opmal os of (P sarng a =n sae (x,, where x [,, C. Noe ha for (P, he sarng sae s known and hene s fxed for he opmzaon. Problem (P s feasble sne a poly ha does no ransm any daa and smply nurs he penaly os s an admssble poly. Is os s fne and hene J(x, s fne. Theorem III: (Srong Dualy Consder he dual funon defned n (8 for ν, hen, we have, J(x, = max ( ν (25 ν and he maxmum on he rgh s aheved by some ν. If (P has an opmal soluon, whh we denoe as r (x,,, hen, r (x,, s he mnmzng r( n (8 for ν = ν. Proof: The proof follows from he lagrange dualy resul n [22 and s omed here for lengh onsderaons. Ineresngly, he dual funon s onave [22 whh makes he maxmzaon n (25 muh smpler as here are no ssues of loal maxma and a dre graden searh algorhm would numerally yeld ν. For our ase, he dual funons for a general ν, are gven n Theorems I (general markov hannel and II (onsan drf hannel. Whle a losed form soluon of ν s dfful o oban, one an easly oban ν numerally usng sandard ehnques. C. Opmal Poly for (P The opmal poly for problem (P an now be obaned by ombnng Theorems I and III and s gven as follows. For k =,..., and [ (k T, kt (kh paron nerval, r (x,,=rν (x,,= x f k, =,...,m (26 ( where he funons {f k (} are evaluaed wh ν = ν. As menoned earler n Seon III-A, he ompuaon for ν and {f k (} needs o be done offlne before he daa ransmsson. In prae, f he ransmer has ompuaonal apables, hese ompuaons an be arred ou a = for he gven problem parameers, oherwse, he ν and {f k (} an be pre-deermned and sored n a able n he ransmer memory. Havng known {f k (}, he losed form sruure of he opmal poly as gven n (26 warrans no furher ompuaon and s smple o mplemen. A me, he ransmer looks a he amoun of daa n he buffer, x, he hannel sae,, he paron nerval k n whh les and ompues he rae for he ommunaon slo as smply x f k (. 936
7 Fg. 3. xpeed oal os 3 2 FullP Opmal Inal daa, B Toal os omparson of he opmal and he full power poly. IV. SIMUATION RSUTS In hs seon, we onsder an llusrave example and presen energy os omparsons for he opmal and he Full Power (FullP poly. In FullP poly, he ransmer always ransms a full power, P, and so gven he sysem sae (x,, he rae s hosen as, r(x,, =g (P =(P /n, for g(r =r n. The smulaon seup s as follows. The hannel model s he G model as desrbed earler n Seon II-B, wh parameers λ bg =, λ gb =3/7, g =and b =.2; hus, λ = max(λ bg,λ gb =and γ = b / g =.2. I an be easly heked ha wh he above parameers, n seady sae he fraon of me spen n he good sae s.7 and.3 n he bad sae. The deadlne s aken as T =and he number of paron nervals as =2. The power-rae funon s, g(r =r 2 and he value of n he penaly os funon s aken as. whh s.% of he deadlne; hus, a me wndow of.% s provded a T. To smulae he proess, ommunaon slo duraon s aken as d = 3 mplyng ha here are T/ 3 = slos over he deadlne nerval. For eah slo, he ransmsson rae s ompued as gven by he orrespondng poly and he oal os s obaned as he sum of he energy oss n he slos plus he penaly os. xpeaon s hen aken as an average over he sample pahs. Fgure 3 s a plo of he expeed oal os of he wo poles wh he nal daa amoun B vared from o.the value of P s hosen suh ha a B =5, even wh bad hannel ondon over he enre deadlne nerval, he enre daa an be served a full power. Ths mples, P = γ (5/T 2 =.25 (5/T s he rae requred o serve 5 uns n me T. Thus, B 5 gves he regme n whh full power always mees he deadlne and B>5s he regme n whh daa s lef ou whh hen nurs he penaly os. I s evden from he plo ha he opmal poly gves a sgnfan gan n he oal os (noe ha he y-axs s on a log sale and a around B = FullP poly nurs almos mes he opmal os. Thus, dynam rae adapaon an yeld sgnfan energy savngs. V. CONCUSION We onsdered energy effen ransmsson of daa over a fadng hannel wh deadlne and power onsrans. Spefally, we addressed he senaro of a wreless ransmer wh shor-erm power lm onsrans, havng B uns of daa ha mus be ransmed by deadlne T over a fadng hannel. Usng a novel onnuous-me formulaon and lagrangan dualy, we oban n losed form he opmal ransmsson poly ha dynamally adaps he rae over me and n response o he me-varyng hannel varaons o mnmze he ransmsson energy os. Ths work opens varous neresng researh dreons whh nlude daa ransmsson wh mulple deadlnes and exensons o senaros nvolvng onrol of mulple ransmers havng deadlne onsrans. ACKNOWDGMNTS The auhors would lke o aknowledge Asuman Ozdaglar and Devavra Shah for helpful dsussons on he work. RFRNCS [ A. K. Kasaggelos, Y. senberg, F. Zha, R. Berry and T. Pappas, Advanes n effen resoure alloaon for pake-based real-me vdeo ransmsson, Proeedngs of he I, vol. 93, no., Jan. 25. [2 A. phremdes, nergy onerns n wreless neworks, I Wreless Communaons, vol. 9, ssue 4, pp , Augus 22. [3 A. Jalal, R. Padovan, R. Pankaj, Daa hroughpu of CDMA-HDR a hgh effeny hgh daa rae personal ommunaon wreless sysem, I Vehular Tehnology Conf., vol. 3, 2. [4. Tsaur, D. ee, Closed-loop arheure and proools for rapd dynam spreadng gan adapaon n CDMA neworks,infocom, 24. [5 R. Berry, R. Gallager, Communaon over fadng hannels wh delay onsrans, I Tran. on Informaon Theory, vol. 48, no. 5, May 22. [6 A. l Gamal, C. Nar, B. Prabhakar,. Uysal-Bykoglu and S. Zahed, nergy-effen shedulng of pake ransmssons over wreless neworks, I Infoom 22, pp , 22. [7 A. Fu,. Modano, J. Tsskls, Opmal energy alloaon for delay onsraned daa ransmsson over a me-varyng hannel, I INFO- COM 23, vol. 2, pp 95-5, Aprl 23. [8 M. Khojasepour, A. Sabharwal, Delay-onsraned shedulng: power effeny, fler desgn and bounds, I INFOCOM 24, Marh 24. [9 P. Nuggehall, V. Srnvasan, R. Rao, Delay onsraned energy effen ransmsson sraeges for wreless deves, I INFOCOM 22. [ B. Collns, R. Cruz, Transmsson poles for me varyng hannels wh average delay onsrans, Alleron onf. on omm., onrol and ompung, Monello, I, 999. [ M. Zafer,. Modano, A Calulus Approah o Mnmum nergy Transmsson Poles wh Qualy of Serve Guaranees, Proeedngs of he I INFOCOM 25, vol., pp , Marh 25. [2 M. Zafer,. Modano, Connuous-me Opmal Rae Conrol for Delay Consraned Daa Transmsson, Alleron onf. on omm., onrol and ompung, Monello, Sep. 25. [3. Tassulas and A. phremdes, Sably properes of onsraned queueng sysems and shedulng poles for maxmum hroughpu n mulhop rado neworks, I Transaons on Auoma Conrol, vol. 37, no. 2, De [4 X. u,. Chong, N. Shroff, A framework for opporuns shedulng n wreless neworks Compuer Neworks, 4, pp , 23. [5 B. Aa, Dynam power onrol n a wreless sa hannel subje o a qualy of serve onsran, Operaons Res. 53 (25, no 5, [6 F. Babh and G. ombard, A Markov model for he moble propagaon hannel, I Transaons on Vehular Tehnology, vol. 49, no., pp , Jan. 2. [7. N. Glber, Capay of burs-nose hannel, Bell Sys. Teh. J.,vol. 39, no. 9, pp , Sep. 96. [8 A. Goldsmh and P. Varaya, Capay, muual Informaon and odng for fne sae markov hannels, I Trans. Informaon Theory, 996. [9 W. Flemng and H. Soner, Conrolled Markov Proesses and Vsosy Soluons, Sprnger-Verlag, 993. [2 M. Davs, Markov Models and Opmzaon, Chapman and Hall, 993. [2 B. Oksendal, Sohas Dfferenal quaons, Sprnger, 5 h edn., 2. [22 D. uenberger, Opmzaon by veor spae mehods, John Wley & sons,
Lecture Notes 4: Consumption 1
Leure Noes 4: Consumpon Zhwe Xu (xuzhwe@sju.edu.n) hs noe dsusses households onsumpon hoe. In he nex leure, we wll dsuss rm s nvesmen deson. I s safe o say ha any propagaon mehansm of maroeonom model s
More informationPendulum Dynamics. = Ft tangential direction (2) radial direction (1)
Pendulum Dynams Consder a smple pendulum wh a massless arm of lengh L and a pon mass, m, a he end of he arm. Assumng ha he fron n he sysem s proporonal o he negave of he angenal veloy, Newon s seond law
More informationProblem Set 3 EC2450A. Fall ) Write the maximization problem of the individual under this tax system and derive the first-order conditions.
Problem Se 3 EC450A Fall 06 Problem There are wo ypes of ndvduals, =, wh dfferen ables w. Le be ype s onsumpon, l be hs hours worked and nome y = w l. Uly s nreasng n onsumpon and dereasng n hours worked.
More informationECON 8105 FALL 2017 ANSWERS TO MIDTERM EXAMINATION
MACROECONOMIC THEORY T. J. KEHOE ECON 85 FALL 7 ANSWERS TO MIDTERM EXAMINATION. (a) Wh an Arrow-Debreu markes sruure fuures markes for goods are open n perod. Consumers rade fuures onras among hemselves.
More informationV.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS
R&RATA # Vol.) 8, March FURTHER AALYSIS OF COFIDECE ITERVALS FOR LARGE CLIET/SERVER COMPUTER ETWORKS Vyacheslav Abramov School of Mahemacal Scences, Monash Unversy, Buldng 8, Level 4, Clayon Campus, Wellngon
More informationComputational results on new staff scheduling benchmark instances
TECHNICAL REPORT Compuaonal resuls on new saff shedulng enhmark nsanes Tm Curos Rong Qu ASAP Researh Group Shool of Compuer Sene Unersy of Nongham NG8 1BB Nongham UK Frs pulshed onlne: 19-Sep-2014 las
More informationCOMPUTER SCIENCE 349A SAMPLE EXAM QUESTIONS WITH SOLUTIONS PARTS 1, 2
COMPUTE SCIENCE 49A SAMPLE EXAM QUESTIONS WITH SOLUTIONS PATS, PAT.. a Dene he erm ll-ondoned problem. b Gve an eample o a polynomal ha has ll-ondoned zeros.. Consder evaluaon o anh, where e e anh. e e
More informationCS286.2 Lecture 14: Quantum de Finetti Theorems II
CS286.2 Lecure 14: Quanum de Fne Theorems II Scrbe: Mara Okounkova 1 Saemen of he heorem Recall he las saemen of he quanum de Fne heorem from he prevous lecure. Theorem 1 Quanum de Fne). Le ρ Dens C 2
More informationRegularization and Stabilization of the Rectangle Descriptor Decentralized Control Systems by Dynamic Compensator
www.sene.org/mas Modern Appled ene Vol. 5, o. 2; Aprl 2 Regularzaon and ablzaon of he Reangle Desrpor Deenralzed Conrol ysems by Dynam Compensaor Xume Tan Deparmen of Eleromehanal Engneerng, Heze Unversy
More informationDynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005
Dynamc Team Decson Theory EECS 558 Proec Shruvandana Sharma and Davd Shuman December 0, 005 Oulne Inroducon o Team Decson Theory Decomposon of he Dynamc Team Decson Problem Equvalence of Sac and Dynamc
More informationSolution in semi infinite diffusion couples (error function analysis)
Soluon n sem nfne dffuson couples (error funcon analyss) Le us consder now he sem nfne dffuson couple of wo blocks wh concenraon of and I means ha, n a A- bnary sysem, s bondng beween wo blocks made of
More informationGraduate Macroeconomics 2 Problem set 5. - Solutions
Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K
More informationSequential Unit Root Test
Sequenal Un Roo es Naga, K, K Hom and Y Nshyama 3 Deparmen of Eonoms, Yokohama Naonal Unversy, Japan Deparmen of Engneerng, Kyoo Insue of ehnology, Japan 3 Insue of Eonom Researh, Kyoo Unversy, Japan Emal:
More informationOn One Analytic Method of. Constructing Program Controls
Appled Mahemacal Scences, Vol. 9, 05, no. 8, 409-407 HIKARI Ld, www.m-hkar.com hp://dx.do.org/0.988/ams.05.54349 On One Analyc Mehod of Consrucng Program Conrols A. N. Kvko, S. V. Chsyakov and Yu. E. Balyna
More informationCS 268: Packet Scheduling
Pace Schedulng Decde when and wha pace o send on oupu ln - Usually mplemened a oupu nerface CS 68: Pace Schedulng flow Ion Soca March 9, 004 Classfer flow flow n Buffer managemen Scheduler soca@cs.bereley.edu
More informationChapter 6: AC Circuits
Chaper 6: AC Crcus Chaper 6: Oulne Phasors and he AC Seady Sae AC Crcus A sable, lnear crcu operang n he seady sae wh snusodal excaon (.e., snusodal seady sae. Complee response forced response naural response.
More informationDual Approximate Dynamic Programming for Large Scale Hydro Valleys
Dual Approxmae Dynamc Programmng for Large Scale Hydro Valleys Perre Carpener and Jean-Phlppe Chanceler 1 ENSTA ParsTech and ENPC ParsTech CMM Workshop, January 2016 1 Jon work wh J.-C. Alas, suppored
More information( ) () we define the interaction representation by the unitary transformation () = ()
Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger
More informationEEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment
EEL 6266 Power Sysem Operaon and Conrol Chaper 5 Un Commmen Dynamc programmng chef advanage over enumeraon schemes s he reducon n he dmensonaly of he problem n a src prory order scheme, here are only N
More information)-interval valued fuzzy ideals in BF-algebras. Some properties of (, ) -interval valued fuzzy ideals in BF-algebra, where
Inernaonal Journal of Engneerng Advaned Researh Tehnology (IJEART) ISSN: 454-990, Volume-, Issue-4, Oober 05 Some properes of (, )-nerval valued fuzzy deals n BF-algebras M. Idrees, A. Rehman, M. Zulfqar,
More informationLecture 2 M/G/1 queues. M/G/1-queue
Lecure M/G/ queues M/G/-queue Posson arrval process Arbrary servce me dsrbuon Sngle server To deermne he sae of he sysem a me, we mus now The number of cusomers n he sysems N() Tme ha he cusomer currenly
More information2/20/2013. EE 101 Midterm 2 Review
//3 EE Mderm eew //3 Volage-mplfer Model The npu ressance s he equalen ressance see when lookng no he npu ermnals of he amplfer. o s he oupu ressance. I causes he oupu olage o decrease as he load ressance
More informationThe Maxwell equations as a Bäcklund transformation
ADVANCED ELECTROMAGNETICS, VOL. 4, NO. 1, JULY 15 The Mawell equaons as a Bäklund ransformaon C. J. Papahrsou Deparmen of Physal Senes, Naval Aademy of Greee, Praeus, Greee papahrsou@snd.edu.gr Absra Bäklund
More informationElectromagnetic waves in vacuum.
leromagne waves n vauum. The dsovery of dsplaemen urrens enals a peular lass of soluons of Maxwell equaons: ravellng waves of eler and magne felds n vauum. In he absene of urrens and harges, he equaons
More informationOptimal Replenishment Policy for Hi-tech Industry with Component Cost and Selling Price Reduction
Opmal Replenshmen Poly for H-eh Indusry wh Componen Cos and Sellng Pre Reduon P.C. Yang 1, H.M. Wee, J.Y. Shau, and Y.F. seng 1 1 Indusral Engneerng & Managemen Deparmen, S. John s Unversy, amsu, ape 5135
More informationOutput equals aggregate demand, an equilibrium condition Definition of aggregate demand Consumption function, c
Eonoms 435 enze D. Cnn Fall Soal Senes 748 Unversy of Wsonsn-adson Te IS-L odel Ts se of noes oulnes e IS-L model of naonal nome and neres rae deermnaon. Ts nvolves exendng e real sde of e eonomy (desred
More informationLecture 18: The Laplace Transform (See Sections and 14.7 in Boas)
Lecure 8: The Lalace Transform (See Secons 88- and 47 n Boas) Recall ha our bg-cure goal s he analyss of he dfferenal equaon, ax bx cx F, where we emloy varous exansons for he drvng funcon F deendng on
More informationPart II CONTINUOUS TIME STOCHASTIC PROCESSES
Par II CONTINUOUS TIME STOCHASTIC PROCESSES 4 Chaper 4 For an advanced analyss of he properes of he Wener process, see: Revus D and Yor M: Connuous marngales and Brownan Moon Karazas I and Shreve S E:
More informationIn the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!
ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal
More information[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5
TPG460 Reservor Smulaon 08 page of 5 DISCRETIZATIO OF THE FOW EQUATIOS As we already have seen, fne dfference appromaons of he paral dervaves appearng n he flow equaons may be obaned from Taylor seres
More informationEP2200 Queuing theory and teletraffic systems. 3rd lecture Markov chains Birth-death process - Poisson process. Viktoria Fodor KTH EES
EP Queung heory and eleraffc sysems 3rd lecure Marov chans Brh-deah rocess - Posson rocess Vora Fodor KTH EES Oulne for oday Marov rocesses Connuous-me Marov-chans Grah and marx reresenaon Transen and
More informationHEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD
Journal of Appled Mahemacs and Compuaonal Mechancs 3, (), 45-5 HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Sansław Kukla, Urszula Sedlecka Insue of Mahemacs,
More informationMechanics Physics 151
Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm H ( q, p, ) = q p L( q, q, ) H p = q H q = p H = L Equvalen o Lagrangan formalsm Smpler, bu
More informationVariants of Pegasos. December 11, 2009
Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on
More informationMechanics Physics 151
Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm Hqp (,,) = qp Lqq (,,) H p = q H q = p H L = Equvalen o Lagrangan formalsm Smpler, bu wce as
More informationMethod of Characteristics for Pure Advection By Gilberto E. Urroz, September 2004
Mehod of Charaerss for Pre Adveon By Glbero E Urroz Sepember 004 Noe: The followng noes are based on lass noes for he lass COMPUTATIONAL HYDAULICS as agh by Dr Forres Holly n he Sprng Semeser 985 a he
More informationPerformance Analysis for a Network having Standby Redundant Unit with Waiting in Repair
TECHNI Inernaonal Journal of Compung Scence Communcaon Technologes VOL.5 NO. July 22 (ISSN 974-3375 erformance nalyss for a Nework havng Sby edundan Un wh ang n epar Jendra Sngh 2 abns orwal 2 Deparmen
More informationComb Filters. Comb Filters
The smple flers dscussed so far are characered eher by a sngle passband and/or a sngle sopband There are applcaons where flers wh mulple passbands and sopbands are requred Thecomb fler s an example of
More informationA GENERAL FRAMEWORK FOR CONTINUOUS TIME POWER CONTROL IN TIME VARYING LONG TERM FADING WIRELESS NETWORKS
A GENERAL FRAEWORK FOR CONTINUOUS TIE POWER CONTROL IN TIE VARYING LONG TER FADING WIRELESS NETWORKS ohammed. Olama, Seddk. Djouad Charalambos D. Charalambous Elecrcal and Compuer Engneerng Deparmen Elecrcal
More informationFTCS Solution to the Heat Equation
FTCS Soluon o he Hea Equaon ME 448/548 Noes Gerald Reckenwald Porland Sae Unversy Deparmen of Mechancal Engneerng gerry@pdxedu ME 448/548: FTCS Soluon o he Hea Equaon Overvew Use he forward fne d erence
More informationOrdinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s
Ordnary Dfferenal Equaons n Neuroscence wh Malab eamples. Am - Gan undersandng of how o se up and solve ODE s Am Undersand how o se up an solve a smple eample of he Hebb rule n D Our goal a end of class
More informationGENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim
Korean J. Mah. 19 (2011), No. 3, pp. 263 272 GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS Youngwoo Ahn and Kae Km Absrac. In he paper [1], an explc correspondence beween ceran
More informationMechanics Physics 151
Mechancs Physcs 5 Lecure 0 Canoncal Transformaons (Chaper 9) Wha We Dd Las Tme Hamlon s Prncple n he Hamlonan formalsm Dervaon was smple δi δ Addonal end-pon consrans pq H( q, p, ) d 0 δ q ( ) δq ( ) δ
More informationOnline Appendix for. Strategic safety stocks in supply chains with evolving forecasts
Onlne Appendx for Sraegc safey socs n supply chans wh evolvng forecass Tor Schoenmeyr Sephen C. Graves Opsolar, Inc. 332 Hunwood Avenue Hayward, CA 94544 A. P. Sloan School of Managemen Massachuses Insue
More informatione-journal Reliability: Theory& Applications No 2 (Vol.2) Vyacheslav Abramov
June 7 e-ournal Relably: Theory& Applcaons No (Vol. CONFIDENCE INTERVALS ASSOCIATED WITH PERFORMANCE ANALYSIS OF SYMMETRIC LARGE CLOSED CLIENT/SERVER COMPUTER NETWORKS Absrac Vyacheslav Abramov School
More informationJohn Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany
Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy
More informationSuperstructure-based Optimization for Design of Optimal PSA Cycles for CO 2 Capture
Supersruure-asedOpmaonforDesgnof OpmalPSACylesforCO 2 Capure R. S. Kamah I. E. Grossmann L.. Begler Deparmen of Chemal Engneerng Carnege Mellon Unversy Psurgh PA 523 Marh 2 PSA n Nex Generaon Power Plans
More informationRelative controllability of nonlinear systems with delays in control
Relave conrollably o nonlnear sysems wh delays n conrol Jerzy Klamka Insue o Conrol Engneerng, Slesan Techncal Unversy, 44- Glwce, Poland. phone/ax : 48 32 37227, {jklamka}@a.polsl.glwce.pl Keywor: Conrollably.
More information, t 1. Transitions - this one was easy, but in general the hardest part is choosing the which variables are state and control variables
Opmal Conrol Why Use I - verss calcls of varaons, opmal conrol More generaly More convenen wh consrans (e.g., can p consrans on he dervaves More nsghs no problem (a leas more apparen han hrogh calcls of
More informationRobustness Experiments with Two Variance Components
Naonal Insue of Sandards and Technology (NIST) Informaon Technology Laboraory (ITL) Sascal Engneerng Dvson (SED) Robusness Expermens wh Two Varance Componens by Ana Ivelsse Avlés avles@ns.gov Conference
More informationJ i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.
umercal negraon of he dffuson equaon (I) Fne dfference mehod. Spaal screaon. Inernal nodes. R L V For hermal conducon le s dscree he spaal doman no small fne spans, =,,: Balance of parcles for an nernal
More informationSOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β
SARAJEVO JOURNAL OF MATHEMATICS Vol.3 (15) (2007), 137 143 SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β M. A. K. BAIG AND RAYEES AHMAD DAR Absrac. In hs paper, we propose
More informationTHE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS
THE PREICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS INTROUCTION The wo dmensonal paral dfferenal equaons of second order can be used for he smulaon of compeve envronmen n busness The arcle presens he
More informationReactive Methods to Solve the Berth AllocationProblem with Stochastic Arrival and Handling Times
Reacve Mehods o Solve he Berh AllocaonProblem wh Sochasc Arrval and Handlng Tmes Nsh Umang* Mchel Berlare* * TRANSP-OR, Ecole Polyechnque Fédérale de Lausanne Frs Workshop on Large Scale Opmzaon November
More informationEpistemic Game Theory: Online Appendix
Epsemc Game Theory: Onlne Appendx Edde Dekel Lucano Pomao Marcano Snscalch July 18, 2014 Prelmnares Fx a fne ype srucure T I, S, T, β I and a probably µ S T. Le T µ I, S, T µ, βµ I be a ype srucure ha
More information10. A.C CIRCUITS. Theoretically current grows to maximum value after infinite time. But practically it grows to maximum after 5τ. Decay of current :
. A. IUITS Synopss : GOWTH OF UNT IN IUIT : d. When swch S s closed a =; = d. A me, curren = e 3. The consan / has dmensons of me and s called he nducve me consan ( τ ) of he crcu. 4. = τ; =.63, n one
More informationMethods of Improving Constitutive Equations
Mehods o mprovng Consuve Equaons Maxell Model e an mprove h ne me dervaves or ne sran measures. ³ ª º «e, d» ¼ e an also hange he bas equaon lnear modaons non-lnear modaons her Consuve Approahes Smple
More informationIncreasing the Probablility of Timely and Correct Message Delivery in Road Side Unit Based Vehicular Communcation
Halmsad Unversy For he Developmen of Organsaons Producs and Qualy of Lfe. Increasng he Probablly of Tmely and Correc Message Delvery n Road Sde Un Based Vehcular Communcaon Magnus Jonsson Krsna Kuner and
More informationAn introduction to Support Vector Machine
An nroducon o Suppor Vecor Machne 報告者 : 黃立德 References: Smon Haykn, "Neural Neworks: a comprehensve foundaon, second edon, 999, Chaper 2,6 Nello Chrsann, John Shawe-Tayer, An Inroducon o Suppor Vecor Machnes,
More informationNotes on the stability of dynamic systems and the use of Eigen Values.
Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon
More informationA Deterministic Algorithm for Summarizing Asynchronous Streams over a Sliding Window
A Deermnsc Algorhm for Summarzng Asynchronous Sreams over a Sldng ndow Cosas Busch Rensselaer Polyechnc Insue Srkana Trhapura Iowa Sae Unversy Oulne of Talk Inroducon Algorhm Analyss Tme C Daa sream: 3
More informationSurvival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System
Communcaons n Sascs Theory and Mehods, 34: 475 484, 2005 Copyrgh Taylor & Francs, Inc. ISSN: 0361-0926 prn/1532-415x onlne DOI: 10.1081/STA-200047430 Survval Analyss and Relably A Noe on he Mean Resdual
More informationCS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4
CS434a/54a: Paern Recognon Prof. Olga Veksler Lecure 4 Oulne Normal Random Varable Properes Dscrmnan funcons Why Normal Random Varables? Analycally racable Works well when observaon comes form a corruped
More informationLecture 11 SVM cont
Lecure SVM con. 0 008 Wha we have done so far We have esalshed ha we wan o fnd a lnear decson oundary whose margn s he larges We know how o measure he margn of a lnear decson oundary Tha s: he mnmum geomerc
More informationOn Convergence of Approximate Message Passing
On Convergene of Approxmae Message Passng Franeso Calagrone, Lenka Zdeborová Insu de Physque Théorque CEA Salay and URA 36, CNRS 99 Gf-sur-Yvee, Frane. Floren Krzakala Laboraore de Physque Sasque, Éole
More informationHow about the more general "linear" scalar functions of scalars (i.e., a 1st degree polynomial of the following form with a constant term )?
lmcd Lnear ransformaon of a vecor he deas presened here are que general hey go beyond he radonal mar-vecor ype seen n lnear algebra Furhermore, hey do no deal wh bass and are equally vald for any se of
More informationSolution of Unit Commitment Problem Using Enhanced Genetic Algorithm
Soluon of Un Commmen roblem Usng Enhaned Gene Algorhm raeek K. Snghal, R. Naresh 2 Deparmen of Eleral Engneerng Naonal Insue of ehnology, Hamrpur Hmahal radesh, Inda-77005 snghalkpraeek@gmal.om, 2 rnareshnh@gmal.om
More informationComparison of Differences between Power Means 1
In. Journal of Mah. Analyss, Vol. 7, 203, no., 5-55 Comparson of Dfferences beween Power Means Chang-An Tan, Guanghua Sh and Fe Zuo College of Mahemacs and Informaon Scence Henan Normal Unversy, 453007,
More informationTime-interval analysis of β decay. V. Horvat and J. C. Hardy
Tme-nerval analyss of β decay V. Horva and J. C. Hardy Work on he even analyss of β decay [1] connued and resuled n he developmen of a novel mehod of bea-decay me-nerval analyss ha produces hghly accurae
More informationP R = P 0. The system is shown on the next figure:
TPG460 Reservor Smulaon 08 page of INTRODUCTION TO RESERVOIR SIMULATION Analycal and numercal soluons of smple one-dmensonal, one-phase flow equaons As an nroducon o reservor smulaon, we wll revew he smples
More informationA Game-theoretical Approach for Job Shop Scheduling Considering Energy Cost in Service Oriented Manufacturing
06 Inernaonal Conferene on Appled Mehans, Mehanal and Maerals Engneerng (AMMME 06) ISBN: 978--60595-409-7 A Game-heoreal Approah for Job Shop Shedulng Consderng Energy Cos n Serve Orened Manufaurng Chang-le
More informationInter-Class Resource Sharing using Statistical Service Envelopes
In Proceedngs of IEEE INFOCOM 99 Iner-Class Resource Sharng usng Sascal Servce Envelopes Jng-yu Qu and Edward W. Knghly Deparmen of Elecrcal and Compuer Engneerng Rce Unversy Absrac Neworks ha suppor mulple
More informationUNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION
INTERNATIONAL TRADE T. J. KEHOE UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 27 EXAMINATION Please answer wo of he hree quesons. You can consul class noes, workng papers, and arcles whle you are workng on he
More informationJournal of Chemical and Pharmaceutical Research, 2014, 6(5): Research Article
Avalable onlne.jopr.om Journal o Chemal Pharmaeual Researh, 014, 6(5:44-48 Researh Arle ISS : 0975-7384 CODE(USA : JCPRC5 Perormane evaluaon or engneerng proje managemen o parle sarm opmzaon based on leas
More informationBandlimited channel. Intersymbol interference (ISI) This non-ideal communication channel is also called dispersive channel
Inersymol nererence ISI ISI s a sgnal-dependen orm o nererence ha arses ecause o devaons n he requency response o a channel rom he deal channel. Example: Bandlmed channel Tme Doman Bandlmed channel Frequency
More informationExistence and Uniqueness Results for Random Impulsive Integro-Differential Equation
Global Journal of Pure and Appled Mahemacs. ISSN 973-768 Volume 4, Number 6 (8), pp. 89-87 Research Inda Publcaons hp://www.rpublcaon.com Exsence and Unqueness Resuls for Random Impulsve Inegro-Dfferenal
More informationOnline Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading
Onlne Supplemen for Dynamc Mul-Technology Producon-Invenory Problem wh Emssons Tradng by We Zhang Zhongsheng Hua Yu Xa and Baofeng Huo Proof of Lemma For any ( qr ) Θ s easy o verfy ha he lnear programmng
More informationImpact of Polarization Mode Dispersion on CPFSK Transmission Systems Using MZI Based Direct Detection Receiver
Impa of Polarzaon Mode Dsperson on CPFSK ransmsson Sysems Usng MZI Based Dre Deeon Reever M. S. Islam, Member, I, and S. P. Majumder, Member, I Absra An analyal approah s presened o deermne he mpa of sgnal
More informationScattering at an Interface: Oblique Incidence
Course Insrucor Dr. Raymond C. Rumpf Offce: A 337 Phone: (915) 747 6958 E Mal: rcrumpf@uep.edu EE 4347 Appled Elecromagnecs Topc 3g Scaerng a an Inerface: Oblque Incdence Scaerng These Oblque noes may
More informationarxiv: v1 [cs.sy] 2 Sep 2014
Noname manuscrp No. wll be nsered by he edor Sgnalng for Decenralzed Roung n a Queueng Nework Y Ouyang Demoshens Tenekezs Receved: dae / Acceped: dae arxv:409.0887v [cs.sy] Sep 04 Absrac A dscree-me decenralzed
More informationChapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are
Chaper 6 DEECIO AD EIMAIO: Fundamenal ssues n dgal communcaons are. Deecon and. Esmaon Deecon heory: I deals wh he desgn and evaluaon of decson makng processor ha observes he receved sgnal and guesses
More informationby Lauren DeDieu Advisor: George Chen
b Laren DeDe Advsor: George Chen Are one of he mos powerfl mehods o nmercall solve me dependen paral dfferenal eqaons PDE wh some knd of snglar shock waves & blow-p problems. Fed nmber of mesh pons Moves
More informationIMPLEMENTATION OF FRACTURE MECHANICS CONCEPTS IN DYNAMIC PROGRESSIVE COLLAPSE PREDICTION USING AN OPTIMIZATION BASED ALGORITHM
COMPDYN III ECCOMAS hema Conferene on Compuaonal Mehods n Sruural Dynams and Earhquake Engneerng M. Papadrakaks, M. Fragadaks, V. Plevrs (eds. Corfu, Greee, 5 8 May IMPLEMENAION OF FRACURE MECHANICS CONCEPS
More information( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model
BGC1: Survval and even hsory analyss Oslo, March-May 212 Monday May 7h and Tuesday May 8h The addve regresson model Ørnulf Borgan Deparmen of Mahemacs Unversy of Oslo Oulne of program: Recapulaon Counng
More informationFirst Order Approximations to Operational Risk Dependence. and Consequences
Frs Order Approxmaons o Operaonal Rsk Dependene and Consequenes Klaus Böker and Clauda Klüppelberg 2 May 20, 2008 Absra We nvesgae he problem of modellng and measurng muldmensonal operaonal rsk. Based
More informationCoordination and Concurrent Negotiation for Multiple Web Services Procurement
Proeedngs of he Inernaonal MulConferene of Engneers and Compuer Senss 009 Vol I IMECS 009, Marh 8-0, 009, Hong Kong Coordnaon and Conurren Negoaon for Mulple Web Serves Prouremen Benyun Sh, and Kwang Mong
More informationarxiv: v2 [quant-ph] 11 Dec 2014
Quanum mehanal uneranes and exa ranson ampludes for me dependen quadra Hamlonan Gal Harar, Yaob Ben-Aryeh, and Ady Mann Deparmen of Physs, Tehnon-Israel Insue of Tehnology, 3 Hafa, Israel In hs work we
More informationDEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL
DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL Sco Wsdom, John Hershey 2, Jonahan Le Roux 2, and Shnj Waanabe 2 Deparmen o Elecrcal Engneerng, Unversy o Washngon, Seale, WA, USA
More informationJoint Channel Estimation and Resource Allocation for MIMO Systems Part I: Single-User Analysis
624 IEEE RANSACIONS ON WIRELESS COUNICAIONS, VOL. 9, NO. 2, FEBRUARY 200 Jon Channel Esmaon and Resource Allocaon for IO Sysems Par I: Sngle-User Analyss Alkan Soysal, ember, IEEE, and Sennur Ulukus, ember,
More informationAdvanced Macroeconomics II: Exchange economy
Advanced Macroeconomcs II: Exchange economy Krzyszof Makarsk 1 Smple deermnsc dynamc model. 1.1 Inroducon Inroducon Smple deermnsc dynamc model. Defnons of equlbrum: Arrow-Debreu Sequenal Recursve Equvalence
More informationLinear Quadratic Regulator (LQR) - State Feedback Design
Linear Quadrai Regulaor (LQR) - Sae Feedbak Design A sysem is expressed in sae variable form as x = Ax + Bu n m wih x( ) R, u( ) R and he iniial ondiion x() = x A he sabilizaion problem using sae variable
More informationAppendix H: Rarefaction and extrapolation of Hill numbers for incidence data
Anne Chao Ncholas J Goell C seh lzabeh L ander K Ma Rober K Colwell and Aaron M llson 03 Rarefacon and erapolaon wh ll numbers: a framewor for samplng and esmaon n speces dversy sudes cology Monographs
More informationLi An-Ping. Beijing , P.R.China
A New Type of Cpher: DICING_csb L An-Png Bejng 100085, P.R.Chna apl0001@sna.com Absrac: In hs paper, we wll propose a new ype of cpher named DICING_csb, whch s derved from our prevous sream cpher DICING.
More information12d Model. Civil and Surveying Software. Drainage Analysis Module Detention/Retention Basins. Owen Thornton BE (Mech), 12d Model Programmer
d Model Cvl and Surveyng Soware Dranage Analyss Module Deenon/Reenon Basns Owen Thornon BE (Mech), d Model Programmer owen.hornon@d.com 4 January 007 Revsed: 04 Aprl 007 9 February 008 (8Cp) Ths documen
More informationON THE WEAK LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS
ON THE WEA LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS FENGBO HANG Absrac. We denfy all he weak sequenal lms of smooh maps n W (M N). In parcular, hs mples a necessary su cen opologcal
More informationGAME theory is a field of mathematics that studies conflict. Dynamic Potential Games with Constraints: Fundamentals and Applications in Communications
1 Dynamc Poenal Games wh Consrans: Fundamenals and Applcaons n Communcaons Sanago Zazo, Member, IEEE, Sergo Valcarcel Macua, Suden Member, IEEE, Malde Sánchez-Fernández, Senor Member, IEEE, Javer Zazo
More informationTight results for Next Fit and Worst Fit with resource augmentation
Tgh resuls for Nex F and Wors F wh resource augmenaon Joan Boyar Leah Epsen Asaf Levn Asrac I s well known ha he wo smple algorhms for he classc n packng prolem, NF and WF oh have an approxmaon rao of
More informationA Systematic Framework for Dynamically Optimizing Multi-User Wireless Video Transmission
A Sysemac Framework for Dynamcally Opmzng ul-user Wreless Vdeo Transmsson Fangwen Fu, haela van der Schaar Elecrcal Engneerng Deparmen, UCLA {fwfu, mhaela}@ee.ucla.edu Absrac In hs paper, we formulae he
More informationSolution. The straightforward approach is surprisingly difficult because one has to be careful about the limits.
ose ad Varably Homewor # (8), aswers Q: Power spera of some smple oses A Posso ose A Posso ose () s a sequee of dela-fuo pulses, eah ourrg depedely, a some rae r (More formally, s a sum of pulses of wdh
More informationSampling Techniques for Probabilistic and Deterministic Graphical models
Samplng ehnques for robabls and Deermns Graphal models ICS 76 Fall 04 Bozhena Bdyuk Rna Deher Readng Darwhe haper 5 relaed papers Overvew. robabls Reasonng/Graphal models. Imporane Samplng 3. Markov Chan
More information