Opportunism, Backpressure, and Stochastic Optimization with the Wireless Broadcast Advantage

Size: px
Start display at page:

Download "Opportunism, Backpressure, and Stochastic Optimization with the Wireless Broadcast Advantage"

Transcription

1 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER Opporunism, Backpressure, and Sochasic Opimizaion wih he Wireless Broadcas Advanage Michael J. Neely, Rahul Urgaonkar Universiy of Souhern California hp://www-rcf.usc.edu/ mjneely Absrac This paper provides a uorial reamen of recen sochasic nework opimizaion echniques, including Lyapunov nework opimizaion, backpressure, and max-weigh decision making. A new echnique of place holder bis ha improves delay for neworking problems wih general coss is also presened. An example applicaion is given for he problem of energy-aware scheduling and rouing in a wireless mobile nework wih channel errors and muli-receiver diversiy. The Diversiy Backpressure Rouing algorihm (DIVBAR, Neely and Urgaonkar 2006, 2008) is illusraed and simulaed in comparison o he Exremely Opporunisic Rouing sraegy (ExOR, Biswas and Morris 2005). I. INTRODUCTION This paper provides a uorial reamen of a recen heory of sochasic nework opimizaion. This heory provides online conrol sraegies for ime varying neworks wih general classes of penalies, rewards, and uiliy funcions. We firs describe he general echnique, which involves conceps of Lyapunov opimizaion, backpressure, and max-weigh decision making. This maerial is aken largely from [1], where a more deailed reamen is given. We hen provide a new exension of he heory ha uses place holder bis o improve congesion and delay in neworks wih general coss [2]. Finally, we presen an example applicaion for a wireless mobile ad-hoc nework wih muli-receiver diversiy. In his nework, a single wireless ransmission can be overheard by muliple receivers, each wih differen success probabiliies. The hroughpu and energy-opimal Diversiy Backpressure Rouing algorihm (DIVBAR) from [3] [4] is presened for his conex. An example simulaion ha compares DIVBAR o he Exremely Opporunisic Rouing algorihm (ExOR) of [5] is given, along wih a simulaion of an enhanced E- DIVBAR algorihm ha combines backpressure and shorespah echniques for furher delay improvemen. II. BACKGROUND The echnique of using Lyapunov drif o sabilize a mulihop packe radio nework was inroduced by Tassiulas and Ephremides in [6], where backpressure rouing and maxweigh scheduling principles are derived. This echnique has had wide success in developing dynamic algorihms for sabiliy in compuer neworks [7] [8] [9] [10], wireless sysems This maerial is suppored in par by one or more of he following: he DARPA IT-MANET program gran W911NF , he NSF gran OCE , he NSF Career gran CCF wih opporunisic scheduling [11] [12] [13], and mobile adhoc neworks [14]. Nework sabiliy can be achieved in hese sysems wihou requiring knowledge of raffic arrival raes or channel probabiliies. An exended Lyapunov drif heorem ha allows join sabiliy and performance opimizaion is developed in [15] [16] [17] o rea uiliy-opimal flow conrol in sochasic neworks. This is exended in [18] for energy opimizaion, and in [1] for a general class of sochasic penaly, reward, and uiliy merics. Performance of hese merics can be pushed o wihin O(1/V ) of opimaliy (where V is a conrol parameer ha can be made arbirarily large), wih a corresponding radeoff in average nework congesion and delay ha is O(V ). This can be applied o rea merics of reliabiliy [20], disorion [2], revenue [21] [22], and o rea neworks wih non-ergodic, non-repeaable mobiliy [23]. This sochasic nework opimizaion echnique is relaed o classical dualiy heory and saic convex programming (see [15] [14] [1] for discussion of he similariies, and [24] [25] for applicaions of backpressure o disribued compuaion). An ineresing observaion is ha sochasic problems are ofen easier o solve han corresponding saic problems, as hey do no require convexiy assumpions (he ime average objecive ends o convexify any non-convexiies of he problem). Relaed alernaive approaches o sochasic nework opimizaion are developed in [26] [27] [28]. Work in [26] considers scheduling in a wireless downlink (similar o [15] [16]), and uses a fluid model ransformaion o relae he sochasic problem o a saic problem. Opimaliy can be approached wih his mehod if he scaled queue backlogs are assumed o converge o appropriae Lagrange mulipliers. Work in [27] reas neworks wih more general penalies and rewards, and uses a relaed fluid model ransformaion ogeher wih a primal-dual algorihm from saic convex opimizaion heory. Work in [28] considers uiliy maximizaion via a sochasic gradien. This paper focuses on he echnique of [1]. This echnique does no require fluid model ransformaions, and direcly leads o he explici performance-delay radeoff [O(1/V ), O(V )]. While his radeoff holds for general neworks and merics, we noe ha i is no necessarily opimal for paricular sysems. Indeed, an improved energy-delay radeoff of [O(1/V ), O( V )] is shown o be opimal for wireless downlinks [29] [30], and a uiliy-delay radeoff of [O(1/V ), O(log(V ))] is shown o be opimal for sysems wih flow conrol [31] [32]. The work in [30] [31] [32] exends he sochasic nework opimizaion

2 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER echniques of [1] o achieve hese opimal delay radeoffs, alhough he resuling dynamic algorihms are more complex and, for breviy, shall no be considered here. III. STOCHASTIC NETWORK OPTIMIZATION A. A General Sochasic Nework Consider a sochasic queueing nework ha operaes in sloed ime wih normalized slos {0, 1, 2,...}. Le Q represen he se of all queues in he nework, and for each q Q, le Q q () be he curren backlog in queue q on slo (assumed o be non-negaive). The paricular unis of backlog, such as ineger-valued unis of packes or real-valued unis of bis, can be assigned as appropriae for a given problem. Le Q() = (Q q ()) q Q be he vecor of all queue backlogs. This vecor changes from slo o slo according o a parially conrollable sochasic process, which depends on a nework sae variable S() and a conrol acion I(). The nework sae variable S() represens an unconrollable source of randomness associaed wih he nework condiion on slo (such as node locaions or channel condiions), and is assumed o evolve independenly of any conrol acions aken in he nework. For convenience, hroughou his paper we assume ha S() is i.i.d. over slos and akes values on some absrac (possibly infinie) sae space S, alhough all of our resuls hold also when S() evolves according o a finie sae, irreducible (possibly non-aperiodic) Markov chain [12] [14] [20] [22]. A he beginning of slo, he nework conroller observes he curren S() and he curren Q(), and makes a conrol acion I() I S(), where I S() is an absrac (possibly infinie) se of all feasible conrol opions under nework sae S(). The conrol acion deermines how queue backlog is added, removed, or ransferred from one queue o anoher in he nework, and may involve a collecion of flow conrol, rouing, and ransmission rae scheduling decisions. On a given slo, he ransiion probabiliies associaed wih he queue vecor are deermined only by he curren Q(), S(), and I() I S() : Q() S(),I() Q( + 1) (1) The equaion (1) can be viewed as a Markov relaion. However, radiional Dynamic Programming and Markov Decision Theory approaches o his problem involve offline compuaions wih very high complexiy and require full a-priori knowledge of all sysem saisics. Our approach is quie differen and allows for simple online decision making, ofen wihou requiring he underlying probabiliy disribuions, provided ha he sysem has a paricular general srucure o be made precise in he nex secion. To begin, we assume he queueing dynamics saisfy he following inequaliy for all q Q: 1 Q q (+1) max[q q () Rq ou (I(), S()), 0]+Rq in (I(), S()) (2) 1 The inequaliy in (2) is ypical, as he Rq in ( ) and Rou q ( ) funcions should depend only on I() and S() and no on queue backlog. Hence, hese funcions represen ransmission raes offered by he nework on slo, alhough hese raes may no fully be uilized if here is lile or no backlog o send (see [1]). For example, we migh have Rq in () = A q() + Pa Ω q() Rou a (), where Aq() is a random amoun of exogenous arrivals, and Ω q() is he se of all queues a ha are currenly ransmiing o q. where Rq ou ( ) represens he (poenially random) amoun of backlog ha can be shifed ou of queue q under nework sae S() and conrol acion I(), and Rq in ( ) is he (poenially random) amoun ha can arrive (considering boh exogenous arrivals from he ranspor layer, and possible endogenous arrivals from oher queues). Such a srucure is ypical for single or muli-hop neworks, and in paricular i applies o all he neworks reaed in he references of his paper (see [1] for examples and more deails on he queueing dynamics for neworks). We furher assume ha he Rq ou ( ) and Rq in ( ) values have bounded condiional second momens, regardless of he conrol acion I() aken on slo. B. Sochasic Penalies We say ha he queueing nework is srongly sable if: lim sup 1 1 E {Q q (τ)} < for all q Q We consider he problem of minimizing he ime average of some general nework penaly funcion, subjec o nework sabiliy and o a collecion of addiional ime average penaly consrains. Specifically, le p(i(), S()) be a random penaly process associaed wih he curren nework sae S() and conrol acion I() for slo. Le x k (I(), S()) be addiional random penaly processes for k {1,..., K}. Noe ha p( ) and x k ( ) can be deerminisic funcions, in which case we do no require any convexiy or coninuiy assumpions, or can be random funcions. 2 We assume only ha hey are non-negaive. Define he following ime average expecaions: p 1 1 = lim sup x k = lim sup 1 1 E {p(i(τ), S(τ))} E {x k (I(τ), S(τ))} Le x =(x 1,..., x K ). Our objecive is o design a nework conrol algorihm ha achieves he following: Minimize: p Subjec o: 1) x x av 2) Nework Sabiliy where x av is a vecor of posiive average penaly consrains. For example, he penaly funcion p( ) migh represen he sum power expended in he nework, and he addiional penalies x k ( ) migh represen powers expended by each paricular node of he nework, which enforces average power consrains a each node [18]. Alernaively, some of he penalies migh represen signal processing disorions [2]. This framework can be used o rea rewards g(i(), S()) by defining a penaly process as a negaive reward process: p(i(), S()) = g max g(i(), S()), assuming ha rewards are upper bounded by a finie consan g max. 2 The work in [1] reas all penalies as deerminisic funcions, alhough random funcions p(i(), S()) can be reaed in his way by using a corresponding deerminisic funcion ˆp(I(), S()) =E {p(i(), S()) I(), S()}.

3 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER The above problem is in fac a special case of he more general sochasic nework opimizaion framework developed in [1]. Indeed, he work in [1] considers he problem wih K differen penalies x = (x 1,..., x K ) and M differen consrains, where he objecive is o minimize f 0 (x) subjec o consrains f m (x) b m for m {1,..., M}, where f 0 ( ), f 1 ( ),..., f M ( ) are general convex muli-variable funcions ha are non-decreasing in each variable. The objecive of minimizing a nonlinear funcion of a ime average is quie differen han minimizing he ime-average iself, and can be solved by inroducing auxiliary variables and flow sae queues, see [1] and [16] [17]. C. Virual Queues and Lyapunov Drif For each of he penaly consrains x k x k,av (for k {1,..., K}), we creae a virual queue Z k () wih updae equaion as follows: Z k ( + 1) = max[z k () x k,av, 0] + x k (I(), S()) (3) This queue is implemened only in sofware, and has a virual consan service rae of x k,av and a virual inpu process equal o he penaly x k (I(), S()) incurred by our conrol acion on slo. Ensuring ha such a queue is srongly sable implies ha he lim sup ime average expecaion of he inpu process is less han or equal o he virual service rae x k,av, and hence ensures ha he ime average penaly consrain x k x k,av is saisfied [18]. This ransforms he consrained opimizaion problem ino a pure problem of sabilizing (virual and acual) queues while minimizing he ime average objecive p. Such virual queues can be viewed as a sochasic version of a Lagrange muliplier, and were inroduced in [18] [19] o maximize nework hroughpu subjec o average power consrains. Le Z() = (Z 1 (),..., Z K ()) be he vecor of virual queues. Recall ha Q() is he vecor of acual queues, and define Θ() = (Q(), Z()) as he combined queue vecor. Define he following quadraic Lyapunov funcions: L Z (Θ()) = 1 K 2 Z k() 2, L Q (Θ()) = 1 2 q Q Q q() 2 L(Θ()) =L Z (Θ()) + L Q (Θ()) Define he one-sep condiional Lyapunov drif Z (Θ()), Q (Θ()), and (Θ()) as follows: Z (Θ()) Q (Θ()) (Θ()) = E {L Z (Θ( + 1)) L Z (Θ()) Θ()} = E {L Q (Θ( + 1)) L Q (Θ()) Θ()} = Z (Θ()) + Q (Θ()) (4) Squaring he equaion (3) and aking expecaions, i can be shown via a sandard argumen ha Z (Θ()) Z (Θ()), where (see, for example, [1]): Z (Θ()) =B Z K Z k ()E {x k,av x k (I(), S()) Θ()} (5) where B Z is a finie consan ha depends on he maximum second momen of he penaly processes, assumed o be upper bounded regardless of he conrol acion I(). Similarly, squaring (2) yields Q (Θ()) Q (Θ()), where: Q q ()E { Rq ou (I(), S()) Θ() } Q (Θ()) =B Q q Q + q Q Q q ()E { R in q (I(), S()) Θ() } (6) where B Q is a finie consan ha depends on he maximum second momens of he Rq ou ( ) and Rq in ( ) processes. D. Lyapunov Opimizaion and he Max-Weigh Algorihm Here we presen he basic Lyapunov opimizaion heorems required for our sochasic analysis (see proofs in [1]). Le Θ() = (Θ q ()) q D be any sochasic vecor of queue backlogs ha evolves in discree ime, where D is an index se for he queues. Le L(Θ()) be a nonnegaive funcion of he queue backlog vecor, and define (Θ()) =E {L(Θ( + 1)) L(Θ()) Θ()}. Theorem 1: (Lyapunov drif [1]) Le f() and g() be any sochasic processes, possibly relaed o Θ(). If for all and all Θ() we have: hen for all we have: (Θ()) E {g() f() Θ()} 1 1 E {f(τ)} 1 1 E {g(τ)} + E {L(Θ(0))} Now le x() be any (possibly negaive) sochasic penaly process associaed wih he sysem, and le x represen a arge penaly. Theorem 1 implies he following. Theorem 2: (Lyapunov Opimizaion [1] [18] [15]) If here are consans B 0, ɛ 0, V 0 such ha he following holds for all and all Θ(): (Θ()) + V E {x() Θ()} B ɛ q D hen: lim sup 1 1 Θ q () + V x x x + B/V q D E {Θ q(τ)} B+V (x x inf ) ɛ where x is he lim sup ime average expecaion of x() and x inf is he he lim inf ime average expecaion of x(). Thus, if he condiion of Theorem 2 holds, he ime average penaly x is a mos B/V beyond he arge x. If V is a conrol parameer ha can be chosen independenly of B, hen B/V can be made arbirarily small, wih a corresponding linear radeoff in ime average congesion. For our sysem, we can define x() = p(i(), S()). Hence, Theorem 2 suggess he conrol sraegy ha observes S(), Θ() on each slo and chooses I() I S() o minimize (subjec o a given Θ()): (Θ()) + V E {p(i(), S()) Θ()} (7) where he above expecaion includes averaging over all possible S() values. However, given ha we can firs observe S(), he opimal policy is o greedily minimize he inside of he expecaion subjec o S() knowledge (recall ha E {X} =

4 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER E Y {E {X Y }} for any random variables X and Y ). 3 Hence, we do no require knowledge of he S() probabiliies. Acually, achieving he minimum of (7) o wihin any addiive consan will suffice. Using (Θ()) as defined in he previous subsecion and noing ha (Θ()) Z (Θ())+ Q (Θ()) leads o he following: The Generalized Max-Weigh Policy: Every imeslo, observe S() and Θ(), and choose I() I S() o greedily minimize (subjec o a given Θ()): Z (Θ()) + Q (Θ()) + V E {p(i(), S()) Θ()} (8) wih Z ( ), Q ( ) defined in (5), (6). Then updae he virual queues Z k () for each k {1,..., K} according o (3) (he acual queue dynamics proceed according o (2)). Using he definiions of Z (Θ()) and Q (Θ()), he minimizaion of (8) is equivalen o choosing I() I S() o minimize: Q q ()[ q Q V ˆp(I(), S()) + K Z k ()ˆx k (I(), S()) ou ˆR q (I(), S()) ˆR in q (I(), S())] (9) where ˆp(I(), S()) =E {p(i(), S()) I(), S()}, and where ˆx k (I(), S()), ˆRou in q (I(), S()), and ˆR q (I(), S()) are similarly defined as he condiional expecaions given I() and S(). These funcions of I() and S() are assumed o be known. Noe ha for a given slo, he observed nework sae S() and queue backlogs Q() and Z() ac as known consans in he minimizaion of (9). Hence, his minimizaion does no require knowledge of he probabiliy disribuion of S(). Furher, oher han compuing expecaions associaed wih he ˆp( ), ˆx( ), ˆRou in q ( ), and ˆR q ( ) funcions, i does no require knowledge of any oher saisics ha affec he nework evoluion, such as raffic arrival raes. 4 IV. ALGORITHM PERFORMANCE Consider he following class of conrol policies ha choose I() I S() as a saionary and poenially randomized funcion only of he curren nework sae S(), and independen of queue backlog. We call such policies S-only policies. Noe ha, by he law of large numbers and he assumpion ha S() is i.i.d. over slos, his class of policies leads o well defined sysem ime averages. Le p be he infimum ime average penaly p incurred over he class of S-only algorihms I () ha saisfy he following: E {p(i (), S())} = p (10) E {x k (I (), S())} x k,av, k {1,..., K} E { R ou q (I (), S()) } E { R in q (I (), S()) }, q Q 3 Specifically, E {X Θ} = E S Θ {E {X Θ, S}} for any random variables X, Θ, and S. If X = X(I, S) (i.e., a random funcion ha depends on S and on an addiional, possibly random, inpu I), hen E {X(I, S) Θ} is minimized by he conrol inpu ha chooses I as he funcion of S and Θ ha minimizes E {X(I, S) Θ, S}. 4 The ˆR in q ( ) funcion may include an unconrollable exogenous arrival rae erm λ q ha is no needed o minimize (9), or may include a flow conrol erm ha can be minimized wih knowledge of he new arrivals A q() observed on slo, wihou requiring he disribuion of A q(), see [1]. In he laer case, A q() can be viewed as a par of he curren nework sae S(). We are implicily assuming he above ime average consrains are feasible, and hence p is he infimum ime average penaly over all feasible S-only algorihms. While our Generalized Max-Weigh Policy is no an S-only policy, we shall measure is opimaliy wih respec o p. For many nework objecives, such as hroughpu, average power, fairness, ec., opimaliy can be aained over he class of S-only algorihms, and hence p is he opimal performance over all conrol algorihms [14] [18]. For simpliciy, we assume ha he infimum p is achievable, in he sense of saisfying he consrains of (10), by a paricular S-only policy I (). 5 We mus furher make he following slackness assumpion: There exiss a consan ɛ > 0 ogeher wih an S-only policy I s () such ha: 6 E {x k (I s (), S())} + ɛ x k,av, k {1,..., K} E { Rq ou (I s (), S()) } ɛ + E { Rq in (I s (), S()) }, q Q Theorem 3: (Performance Theorem [1] [18]) Suppose here is a value p max < so ha 0 p(i(), S()) p max always. Under he Generalized Max-Weigh Policy wih any conrol parameer V 0, he nework is sable, he consrains x x av are saisfied, he lim sup ime average penaly p saisfies: lim sup 1 1 E {p(i(τ), S(τ))} p + (B Z + B Q )/V (11) and he ime average queue backlogs saisfy: 1 1 lim sup K E {Q q (τ)} + E {Z k (τ)} q Q (B Z + B Q + V p max )/ɛ (12) Thus, he V parameer can be chosen as desired o affec he [O(1/V ), O(V )] performance-congesion radeoff. Proof: Le I () be he S-only policy ha achieves all feasibiliy consrains (10) and yields E {p(i (), S())} = p. Le I MW () denoe he Max-Weigh policy. Le MW (Θ()) denoe he Lyapunov drif under he Max-Weigh policy I MW () (as defined in (4)), and le MW Z (Θ()) and MW Q (Θ()) denoe he corresponding drif bounds as defined in Secion III-C (equaions (5) and (6)). Le Z(Θ()) and Q(Θ()) be he corresponding drif bounds for he I () policy. Because I MW () (by definiion) minimizes (8), we have: MW (Θ()) + V E { p(i MW (), S()) Θ() } Z(Θ()) + Q(Θ()) + V E {p(i (), S()) Θ()} (13) Now noe ha because I () is S-only and S() is i.i.d. over slos, I () is independen of queue backlogs Θ(), and so: E {x k (I (), S()) Θ()} = E {x k (I (), S())} x k,av where he final inequaliy above is due o he inequaliy consrains of (10). Thus, plugging he inequaliy consrains 5 Else, we can prove he same performance resuls wih a limiing argumen using a sequence of policies I n() ha converge o he performance p. 6 This single ɛ-slackness assumpion ha holds for all consrains simulaneously can be relaxed by replacing i wih muliple ɛ-slackness assumpions ha hold for each consrain individually while all oher consrains are saisfied wih possible 0-slackness.

5 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER of (10) ino he definiions of Z(Θ()) and Q(Θ()) given in (5) and (6), i is a simple maer o verify: Z(Θ()) + Q(Θ()) B Z + B Q Define B =B Z +B Q. Using he above inequaliy in (13) yields: MW (Θ()) + V E { p(i MW (), S()) Θ() } B + V p This is in he exac form for using he Lyapunov Opimizaion Theorem (Theorem 2) wih ɛ = 0, x() = p(i MW (), S()), x = p, which proves he performance bound (11). Similarly, o derive he ime average bound on queue backlog, we can use he S-only policy I s () ha yields he slackness consrains. Noe ha he inequaliy (13) holds equally if we plug he I s () policy ino each of he hree erms on he righ hand side, raher han he I () policy (as I MW () minimizes he righ hand side over any oher conrol acion). I is again a simple maer o verify from he slackness consrains and he definiions of s Z(Θ()) and s Q(Θ()) (represening drif bounds under he I s () policy) ha: K s Z(Θ()) + s Q(Θ()) B ɛ Q q () Z k () + q Q which can be plugged ino he righ hand side of (13) o yield he resuling bound (12) upon applicaion of Theorem 2. A. On Disribued Implemenaion and Queue Groupings We noe ha disribued implemenaion of he Generalized Max-Weigh policy is possible when channels are orhogonal and/or ransmission opions are randomized [14] [18] [23]. Consan-facor hroughpu and performance resuls, where sabiliy regions are reduced by a consan facor, are guaraneed when a disribued algorihm comes wihin a consan facor of maximizing (9). This is shown in [15] for sabiliy, in [1] [17] for join sabiliy and hroughpu-uiliy maximizaion, and in [3] for sabiliy and power minimizaion. A differen approach o sable scheduling via maximal machings can also be shown o yield consan-facor resuls [33] [34] [35] [36]. Maximal scheduling is ofen analyzed using queue-grouped Lyapunov funcions ha can provide igher delay guaranees for reduced hroughpu regions [37] [38]. Queue groupings can also be used o prove order-opimal delay for he full-hroughpu max-weigh scheduling algorihm in wireless downlinks wih ON/OFF channels [39]. B. On Delay Reducion and Wors Case Backlog We noe ha nework congesion and delay can ofen be improved by direcly minimizing he drif expression (7), raher han is upper bound ha uses (Θ()) Z (Θ())+ Q (Θ()). The performance bounds of Theorem 3 hold also in his case (he proof is exacly he same). However, congesion is ypically beer because he drif expression (7) uses he acual amoun of backlog ransferred (ofen being a minimum of he offered ransfer rae and he acual backlog available), while he drif bound uses only he offered ransfer rae (he offered rae and he acual amoun ransferred are he same when queue backlog is sufficienly large). Alernaively, one can minimize any expression ha differs by only an addiive consan from eiher he drif expression (7) or is upper bound. This simply adds an addiive consan o he congesion bound of Theorem 3. Such a scenario arises if backlog updaes are delayed so ha acual queue backlogs differ by a consan from hose used in he conrol decision [1]. Queue backlogs can also be augmened by consans o achieve addiional desired behavior while mainaining sabiliy. For example, delay in muli-hop neworks can ofen be improved by adding a shores pah bias erm o each queue backlog. This is used in he Enhanced Dynamic Rouing and Power Allocaion (EDRPC) algorihm of [14] and he Enhanced Diversiy Backpressure Rouing (E-DIVBAR) algorihm of [3], where simulaions demonsrae considerable delay improvemen (see also Secion VI). We furher noe ha in sysems wih flow conrol, he max-weigh policy (or modificaions ha add a consan o he backpressure value) ofen allow wors case queue backlog guaranees [18] [23] [21]. V. PLACE HOLDER BITS FOR DELAY IMPROVEMENT Noe ha he iniial condiion of he queues does no affec he ime average penaly or queue backlog bounds of Theorem 2 or Theorem 3. Suppose now ha he nework dynamics under he Max-Weigh policy (9) saisfy he following propery: Propery 1: There is a non-negaive queue backlog vecor Θ 0 = (Q 0 ; Z 0 ) such ha if Θ(0) Θ 0, hen Θ() Θ 0 for all 0, where inequaliy is aken enrywise. If Propery 1 holds for a non-zero vecor Θ 0 = (Q 0, Z 0 ), we iniialize he virual queues o Z(0) = Z 0, and we iniialize he acual queues so ha here is no acual daa iniially in hese queues, bu here is an amoun Q 0 of fake daa or place holder daa. Specifically, suppose he acual queue backlog is given by ˆQ(), and define: Q() = ˆQ() + Q 0 The Max-Weigh algorihm uses he Q() values in is decisions (9). However, whenever i makes a ransmission decision, i ransmis only acual daa (no fake daa), whenever possible. The resuling sample pah of Q() is equivalen o ha of a sysem wih iniial condiion Θ(0) = (Q 0 ; Z 0 ), and hence yields he exac same ime average bounds on penaly p and queue backlog Q q as specified in Theorem 3. Furher, because Propery 1 holds, we will have (Q(); Z()) (Q 0 ; Z 0 ) for all, and so he acual queue backlog will never drop low enough o require ransmission of fake daa. I follows ha he acual backlog vecor ˆQ() will be exacly an amoun Q 0 lower han Q() for every insan of ime, which reduces he average queue backlog by exacly his amoun, wihou any change in he ime average penaly p. Thus, his fake daa is simply acing as a place holder o properly affec he performance opimizaion. We recenly developed his concep of place holder bis in [2] for a single-hop sysem (he reader is referred o more specific examples given here), alhough i applies equally well in his poenially muli-hop seing. For cos minimizaion problems, he Max-Weigh algorihm of (9) ypically does saisfy Propery 1 for a non-zero vecor Θ 0. Indeed, from

6 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER (9) i can be seen ha a queueing node q will ypically no ransmi any daa ou unless he amoun of backlog i currenly has, muliplied by he amoun i will ransmi ou, exceeds a hreshold equal o he penaly incurred by his ransmission muliplied by V. The backlog reducion of place holder daa can be quie significan, and can be successfully applied o energy minimizaion algorihms [18] [3] and oher relaed cos minimizaion algorihms o improve delay. VI. MULTI-RECEIVER DIVERSITY AND DIVBAR This sochasic opimizaion echnique is applied in our prior work [3] o minimize average power in a muli-hop wireless mobile nework wih muli-receiver diversiy. Specifically, a packe ransmission a a given node can poenially be received by a se of neighbors, wih heerogeneous success probabiliies ha depend on he curren opology sae S(). All successful receivers provide ACK feedback, and he ransmiing node sends a final message o indicae which one (if any) should ake responsibiliy for he packe. The deailed nework conrol variables are given in [3], where he Max-Weigh policy is applied o creae a Diversiy Backpressure Rouing Algorihm (DIVBAR) ha is hroughpu and energy opimal. In he case of a single commodiy, orhogonal channels, and no power opimizaion (i.e., V = 0), DIVBAR reduces o having each node ransmi whenever possible, and shifing forwarding responsibiliy o he successful receiver wih he larges differenial backlog (being he difference in queue backlog a he ransmier and receiver), reaining he packe a he ransmier if no receiver has a posiive differenial backlog. If power minimizaion is included (V > 0), he algorihm is more complex and requires insananeous success probabiliies of neighbor nodes under a given opology sae, bu does no require knowledge of raffic raes or of how he opology changes from one slo o he nex. If a mos one packe can be ransmied on a slo, and each ransmission requires 1 uni of power, i can be shown ha DIVBAR saisfies Propery 1 wih Q 0 = (max[ V 1, 0]) q Q (in unis of packes). Figure 1 illusraes a cell-pariioned mobile ad-hoc nework, from Example 2 of [3]. There are 9 source nodes: 3 saionary, 3 locally mobile, and 3 fully mobile. The locally mobile nodes are resriced o move in he shaded cells while he fully mobile nodes can move anywhere in he nework. There are 2 saionary sinks and packes can be delivered o eiher of hem (hus, his is a single commodiy scenario). The mobile Saionary Node Locally Mobile Node Fully Mobile Node Sink Fig. 1. A mobile nework wih wo sinks and heerogeneous mobiliy [3]. nodes perform a Markovian random walk over heir respecive regions (choosing each slo o eiher remain in heir curren cell or o move o an adjacen cell). We assume a mos one packe can be ransmied per cell, bu ransmissions in differen cells are orhogonal. Figure 2 illusraes he performance of DIVBAR for his example (see he deailed model descripion in [3]). This figure also includes new daa ha demonsraes he benefis of backlog reducion achieved by using place holder packes, a no addiional energy cos (as shown in Figure 3). Average Toal Congesion Average Toal Congesion vs V DIVBAR DIVBAR wih place holder packes V Fig. 2. Average oal congesion (i.e., oal packes summed over all queues in he nework) versus V under DIVBAR wih and wihou place holder packes on he example mobile nework of Fig. 1. Average power (per node) Average Power vs V DIVBAR DIVBAR wih place holder packes V Fig. 3. Average power versus V under DIVBAR wih and wihou place holder packes on he example mobile nework of Fig. 1. The wo curves are almos idenical, demonsraing no loss of energy efficiency wih place holder packes. Figure 4 reas a differen nework: he 2-commodiy saic nework from Example 1 in [3] (figure omied for breviy, see [3]). In he muli-commodiy case, DIVBAR selecs he opimal commodiy o ransmi via a rank-ordering ha depends on he curren backlog and recepion success probabiliies of neighboring nodes. Such a commodiy disincion is required for opimaliy, and is quie differen from oher diversiy algorihms (such as ExOR) ha do no opimize over commodiy selecion. We consider only sabiliy here, wihou power opimizaion (so ha V = 0), and compare o he ExOR algorihm of [5] ha uses he heurisic of shifing packes o he successful receiver ha is closes o he desinaion, where proximiy is measured wih respec o a shores pah meric.

7 ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS, PACIFIC GROVE, CA, OCTOBER Unlike DIVBAR, he ExOR algorihm is no hroughpuopimal, and in his example i becomes unsable a raes ha are a lile more han half he disance o he capaciy region boundary. However, is shores-pah properies allow ExOR o have lower delay in he low rae region. A combined sraegy E-DIVBAR ha uses backpressure and shores pah in he forwarding decision is also illusraed (his is relaed o he mixed backpressure/shores pah enhancemen in [14]). I is shown in [3] ha E-DIVBAR is also hroughpu opimal, and we see in his example ha i has lower delay han boh DIVBAR and ExOR across all inpu raes. Average Toal Occupancy (log scale) ExOR DIVBAR E DIVBAR Average Toal Occupancy vs. Inpu Rae Inpu Rae (packes/slo) Fig. 4. A comparison of DIVBAR, ExOR, and E-DIVBAR for he 2- commodiy saic nework example of [3]. REFERENCES [1] L. Georgiadis, M. J. Neely, and L. Tassiulas. Resource allocaion and cross-layer conrol in wireless neworks. Foundaions and Trends in Neworking, vol. 1, no. 1, pp , [2] M. J. Neely and A. Sharma. Dynamic daa compression wih disorion consrains for wireless ransmission over a fading channel. arxiv: v1, July 24, [3] M. J. Neely and R. Urgaonkar. Opimal backpressure rouing in wireless neworks wih muli-receiver diversiy. Ad Hoc Neworks (Elsevier), 2008, doi: /j.adhoc [4] M. J. Neely. Opimal backpressure rouing for wireless neworks wih muli-receiver diversiy. Proc. of Conf. on Informaion Sciences and Sysems (CISS), Invied paper on Opimizaion of Communicaion Neworks, March [5] S. Biswas and R. Morris. ȨxO Ŗ: Opporunisic muli-hop rouing for wireless neworks. Proc. of Sigcomm, [6] L. Tassiulas and A. Ephremides. Sabiliy properies of consrained queueing sysems and scheduling policies for maximum hroughpu in mulihop radio neworks. IEEE Transacaions on Auomaic Conrol, vol. 37, no. 12, pp , Dec [7] P. R. Kumar and S. P. Meyn. Sabiliy of queueing neworks and scheduling policies. IEEE Trans. on Auomaic Conrol, vol.40,.n.2, pp , Feb [8] N. Kahale and P. E. Wrigh. Dynamic global packe rouing in wireless neworks. Proc. IEEE INFOCOM, [9] N. McKeown, V. Ananharam, and J. Walrand. Achieving 100% hroughpu in an inpu-queued swich. Proc. IEEE INFOCOM, [10] E. Leonardi, M. Mellia, F. Neri, and M. Ajmone Marsan. Bounds on average delays and queue size averages and variances in inpu-queued cell-based swiches. Proc. IEEE INFOCOM, [11] L. Tassiulas and A. Ephremides. Dynamic server allocaion o parallel queues wih randomly varying conneciviy. IEEE Transacions on Informaion Theory, vol. 39, pp , March [12] L. Tassiulas. Scheduling and performance limis of neworks wih consanly changing opology. IEEE Trans. on Inf. Theory, May [13] M. Andrews, K. Kumaran, K. Ramanan, A. Solyar, and P. Whiing. Providing qualiy of service over a shared wireless link. IEEE Communicaions Magazine, vol. 39, no.2, pp , [14] M. J. Neely, E. Modiano, and C. E Rohrs. Dynamic power allocaion and rouing for ime varying wireless neworks. IEEE Journal on Seleced Areas in Communicaions, vol. 23, no. 1, pp , January [15] M. J. Neely. Dynamic Power Allocaion and Rouing for Saellie and Wireless Neworks wih Time Varying Channels. PhD hesis, Massachuses Insiue of Technology, LIDS, [16] M. J. Neely, E. Modiano, and C. Li. Fairness and opimal sochasic conrol for heerogeneous neworks. Proc. IEEE INFOCOM, March [17] M. J. Neely, E. Modiano, and C. Li. Fairness and opimal sochasic conrol for heerogeneous neworks. IEEE/ACM Transacions on Neworking, vol. 16, no. 2, pp , April [18] M. J. Neely. Energy opimal conrol for ime varying wireless neworks. IEEE Transacions on Informaion Theory, vol. 52, no. 7, pp , July [19] M. J. Neely. Energy opimal conrol for ime varying wireless neworks. Proc. IEEE INFOCOM, March [20] R. Urgaonkar and M. J. Neely. Opporunisic scheduling wih reliabiliy guaranees in cogniive radio neworks. Proc. IEEE INFOCOM, April [21] M. J. Neely. Opimal pricing in a free marke wireless nework. Proc. IEEE INFOCOM, May [22] L. Huang and M. J. Neely. The opimaliy of wo prices: Maximizing revenue in a sochasic nework. Proc. 45h Alleron Conf. on Communicaion, Conrol, and Compuing, Sep [23] M. J. Neely and R. Urgaonkar. Cross layer adapive conrol for wireless mesh neworks. Ad Hoc Neworks (Elsevier), vol. 5, no. 6, pp , Augus [24] M. J. Neely. Disribued and secure compuaion of convex programs over a nework of conneced processors. DCDIS Conf., Guelph, Onario, July [25] M. Andrews, K. Jung, and A. Solyar. Sabiliy of he max-weigh rouing and scheduling proocol in dynamic neworks and a criical loads. Proc. STOC, June [26] A. Eryilmaz and R. Srikan. Fair resource allocaion in wireless neworks using queue-lengh-based scheduling and congesion conrol. Proc. IEEE INFOCOM, March [27] A. Solyar. Maximizing queueing nework uiliy subjec o sabiliy: Greedy primal-dual algorihm. Queueing Sysems, vol. 50, pp , [28] J. W. Lee, R. R. Mazumdar, and N. B. Shroff. Opporunisic power scheduling for dynamic muliserver wireless sysems. IEEE Transacions on Wireless Communicaions, vol. 5, no.6, pp , June [29] R. Berry and R. Gallager. Communicaion over fading channels wih delay consrains. IEEE Transacions on Informaion Theory, vol. 48, no. 5, pp , May [30] M. J. Neely. Opimal energy and delay radeoffs for muli-user wireless downlinks. IEEE Transacions on Informaion Theory, vol. 53, no. 9, pp , Sep [31] M. J. Neely. Super-fas delay radeoffs for uiliy opimal fair scheduling in wireless neworks. IEEE Journal on Seleced Areas in Communicaions, Special Issue on Nonlinear Opimizaion of Communicaion Sysems, vol. 24, no. 8, pp , Aug [32] M. J. Neely. Inelligen packe dropping for opimal energy-delay radeoffs in wireless downlinks. Proc. of WiOp, Boson, April [33] X. Lin and N. B. Shroff. The impac of imperfec scheduling on crosslayer rae conrol in wireless neworks. Proc. IEEE INFOCOM, [34] X. Wu, R. Srikan, and J. R. Perkins. Scheduling efficiency of disribued greedy scheduling algorihms in wireless neworks. IEEE Transacions on Mobile Compuing, June [35] P. Chaporkar, K. Kar, X. Luo, and S. Sarkar. Throughpu and fairness guaranees hrough maximal scheduling in wireless neworks. IEEE Trans. on Informaion Theory, vol. 54, no. 2, pp , Feb [36] L. Lin, X. Lin, and N. B. Shroff. Low-complexiy and disribued energy minimizaion in muli-hop wireless neworks. Proc. IEEE INFOCOM, [37] S. Deb, D. Shah, and S. Shakkoai. Fas maching algorihms for repeiive opimizaion: An applicaion o swich scheduling. Proc. of 40h Annual Conference on Informaion Sciences and Sysems (CISS), Princeon, NJ, March [38] M. J. Neely. Delay analysis for maximal scheduling in wireless neworks wih bursy raffic. Proc. IEEE INFOCOM, April [39] M. J. Neely. Delay analysis for max weigh opporunisic scheduling in wireless sysems. Proc. 46h Annual Alleron Conf. on Communicaion, Conrol, and Compuing, Sep

Opportunistic Scheduling with Reliability Guarantees in Cognitive Radio Networks

Opportunistic Scheduling with Reliability Guarantees in Cognitive Radio Networks PROC. IEEE INFOCO, PHOENIX, AZ, APRIL 008 Opporunisic Scheduling wih Reliabiliy Guaranees in Cogniive Radio Neworks Rahul Urgaonkar, ichael J. Neely Universiy of Souhern California, Los Angeles, CA 90089

More information

Intelligent Packet Dropping for Optimal Energy-Delay Tradeoffs in Wireless Downlinks

Intelligent Packet Dropping for Optimal Energy-Delay Tradeoffs in Wireless Downlinks PROC. OF 4TH INT. SYMPOSIUM ON MODELING AND OPTIMIZATION IN MOBILE, AD HOC, AND WIRELESS NETWORKS (WIOPT), APRIL 2006 1 Inelligen Packe Dropping for Opimal Energy-Delay Tradeoffs in Wireless Downlinks

More information

A Primal-Dual Type Algorithm with the O(1/t) Convergence Rate for Large Scale Constrained Convex Programs

A Primal-Dual Type Algorithm with the O(1/t) Convergence Rate for Large Scale Constrained Convex Programs PROC. IEEE CONFERENCE ON DECISION AND CONTROL, 06 A Primal-Dual Type Algorihm wih he O(/) Convergence Rae for Large Scale Consrained Convex Programs Hao Yu and Michael J. Neely Absrac This paper considers

More information

Intelligent Packet Dropping for Optimal Energy-Delay Tradeoffs in Wireless Downlinks

Intelligent Packet Dropping for Optimal Energy-Delay Tradeoffs in Wireless Downlinks IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 54, NO. 3, PP. 565-579, MARCH 2009. 1 Inelligen Packe Dropping for Opimal Energy-Delay Tradeoffs in Wireless Downlinks Michael J. Neely Universiy of Souhern

More information

5. Stochastic processes (1)

5. Stochastic processes (1) Lec05.pp S-38.45 - Inroducion o Teleraffic Theory Spring 2005 Conens Basic conceps Poisson process 2 Sochasic processes () Consider some quaniy in a eleraffic (or any) sysem I ypically evolves in ime randomly

More information

Stable Scheduling Policies for Maximizing Throughput in Generalized Constrained Queueing Systems

Stable Scheduling Policies for Maximizing Throughput in Generalized Constrained Queueing Systems 1 Sable Scheduling Policies for Maximizing Throughpu in Generalized Consrained Queueing Sysems Prasanna Chaporar, Suden Member, IEEE, Saswai Sarar, Member, IEEE Absrac We consider a class of queueing newors

More information

An introduction to the theory of SDDP algorithm

An introduction to the theory of SDDP algorithm An inroducion o he heory of SDDP algorihm V. Leclère (ENPC) Augus 1, 2014 V. Leclère Inroducion o SDDP Augus 1, 2014 1 / 21 Inroducion Large scale sochasic problem are hard o solve. Two ways of aacking

More information

Optimal Server Assignment in Multi-Server

Optimal Server Assignment in Multi-Server Opimal Server Assignmen in Muli-Server 1 Queueing Sysems wih Random Conneciviies Hassan Halabian, Suden Member, IEEE, Ioannis Lambadaris, Member, IEEE, arxiv:1112.1178v2 [mah.oc] 21 Jun 2013 Yannis Viniois,

More information

Resource Allocation in Visible Light Communication Networks NOMA vs. OFDMA Transmission Techniques

Resource Allocation in Visible Light Communication Networks NOMA vs. OFDMA Transmission Techniques Resource Allocaion in Visible Ligh Communicaion Neworks NOMA vs. OFDMA Transmission Techniques Eirini Eleni Tsiropoulou, Iakovos Gialagkolidis, Panagiois Vamvakas, and Symeon Papavassiliou Insiue of Communicaions

More information

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB Elecronic Companion EC.1. Proofs of Technical Lemmas and Theorems LEMMA 1. Le C(RB) be he oal cos incurred by he RB policy. Then we have, T L E[C(RB)] 3 E[Z RB ]. (EC.1) Proof of Lemma 1. Using he marginal

More information

Vehicle Arrival Models : Headway

Vehicle Arrival Models : Headway Chaper 12 Vehicle Arrival Models : Headway 12.1 Inroducion Modelling arrival of vehicle a secion of road is an imporan sep in raffic flow modelling. I has imporan applicaion in raffic flow simulaion where

More information

Lecture 4 Notes (Little s Theorem)

Lecture 4 Notes (Little s Theorem) Lecure 4 Noes (Lile s Theorem) This lecure concerns one of he mos imporan (and simples) heorems in Queuing Theory, Lile s Theorem. More informaion can be found in he course book, Bersekas & Gallagher,

More information

Network Utility Maximization over Partially Observable Markovian Channels

Network Utility Maximization over Partially Observable Markovian Channels WIOPT 2011 1 Nework Uiliy Maximizaion over Parially Observable Markovian Channels Chih-ping Li, Suden Member, IEEE and Michael J. Neely, Senior Member, IEEE Absrac This paper considers maximizing hroughpu

More information

Delay Reduction via Lagrange Multipliers in Stochastic Network Optimization

Delay Reduction via Lagrange Multipliers in Stochastic Network Optimization IEEE TRANS. ON AUTOMATIC CONTROL, OL. 56, ISSUE 4, PP. 842-857, APRIL 2011 1 Delay Reducion via Lagrange Mulipliers in Sochasic Nework Opimizaion Longbo Huang, Michael J. Neely Absrac In his paper, we

More information

On the Stability Region of Multi-Queue Multi-Server Queueing Systems with Stationary Channel Distribution

On the Stability Region of Multi-Queue Multi-Server Queueing Systems with Stationary Channel Distribution On he Sabiliy Region of Muli-Queue Muli-Server Queueing Sysems wih Saionary Channel Disribuion Hassan Halabian, Ioannis Lambadaris, Chung-Horng Lung Deparmen of Sysems and Compuer Engineering, Carleon

More information

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation Course Noes for EE7C Spring 018: Convex Opimizaion and Approximaion Insrucor: Moriz Hard Email: hard+ee7c@berkeley.edu Graduae Insrucor: Max Simchowiz Email: msimchow+ee7c@berkeley.edu Ocober 15, 018 3

More information

Random Walk with Anti-Correlated Steps

Random Walk with Anti-Correlated Steps Random Walk wih Ani-Correlaed Seps John Noga Dirk Wagner 2 Absrac We conjecure he expeced value of random walks wih ani-correlaed seps o be exacly. We suppor his conjecure wih 2 plausibiliy argumens and

More information

STATE-SPACE MODELLING. A mass balance across the tank gives:

STATE-SPACE MODELLING. A mass balance across the tank gives: B. Lennox and N.F. Thornhill, 9, Sae Space Modelling, IChemE Process Managemen and Conrol Subjec Group Newsleer STE-SPACE MODELLING Inroducion: Over he pas decade or so here has been an ever increasing

More information

Mean-square Stability Control for Networked Systems with Stochastic Time Delay

Mean-square Stability Control for Networked Systems with Stochastic Time Delay JOURNAL OF SIMULAION VOL. 5 NO. May 7 Mean-square Sabiliy Conrol for Newored Sysems wih Sochasic ime Delay YAO Hejun YUAN Fushun School of Mahemaics and Saisics Anyang Normal Universiy Anyang Henan. 455

More information

Solutions for Assignment 2

Solutions for Assignment 2 Faculy of rs and Science Universiy of Torono CSC 358 - Inroducion o Compuer Neworks, Winer 218 Soluions for ssignmen 2 Quesion 1 (2 Poins): Go-ack n RQ In his quesion, we review how Go-ack n RQ can be

More information

Energy Storage Benchmark Problems

Energy Storage Benchmark Problems Energy Sorage Benchmark Problems Daniel F. Salas 1,3, Warren B. Powell 2,3 1 Deparmen of Chemical & Biological Engineering 2 Deparmen of Operaions Research & Financial Engineering 3 Princeon Laboraory

More information

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon 3..3 INRODUCION O DYNAMIC OPIMIZAION: DISCREE IME PROBLEMS A. he Hamilonian and Firs-Order Condiions in a Finie ime Horizon Define a new funcion, he Hamilonian funcion, H. H he change in he oal value of

More information

Comments on Window-Constrained Scheduling

Comments on Window-Constrained Scheduling Commens on Window-Consrained Scheduling Richard Wes Member, IEEE and Yuing Zhang Absrac This shor repor clarifies he behavior of DWCS wih respec o Theorem 3 in our previously published paper [1], and describes

More information

arxiv: v1 [math.oc] 11 Sep 2017

arxiv: v1 [math.oc] 11 Sep 2017 Online Learning in Weakly Coupled Markov Decision Processes: A Convergence ime Sudy Xiaohan Wei, Hao Yu and Michael J. Neely arxiv:709.03465v [mah.oc] Sep 07. Inroducion Absrac: We consider muliple parallel

More information

Energy-Aware Wireless Scheduling with Near Optimal Backlog and Convergence Time Tradeoffs

Energy-Aware Wireless Scheduling with Near Optimal Backlog and Convergence Time Tradeoffs IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 24, NO. 4, AUG. 2016, PP. 2223-2236 1 Energy-Aware Wireless Scheduling wih Near Opimal Backlog and Convergence Time Tradeoffs Michael J. Neely Universiy of Souhern

More information

Decentralized Stochastic Control with Partial History Sharing: A Common Information Approach

Decentralized Stochastic Control with Partial History Sharing: A Common Information Approach 1 Decenralized Sochasic Conrol wih Parial Hisory Sharing: A Common Informaion Approach Ashuosh Nayyar, Adiya Mahajan and Demoshenis Tenekezis arxiv:1209.1695v1 [cs.sy] 8 Sep 2012 Absrac A general model

More information

1 Review of Zero-Sum Games

1 Review of Zero-Sum Games COS 5: heoreical Machine Learning Lecurer: Rob Schapire Lecure #23 Scribe: Eugene Brevdo April 30, 2008 Review of Zero-Sum Games Las ime we inroduced a mahemaical model for wo player zero-sum games. Any

More information

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle Chaper 2 Newonian Mechanics Single Paricle In his Chaper we will review wha Newon s laws of mechanics ell us abou he moion of a single paricle. Newon s laws are only valid in suiable reference frames,

More information

L07. KALMAN FILTERING FOR NON-LINEAR SYSTEMS. NA568 Mobile Robotics: Methods & Algorithms

L07. KALMAN FILTERING FOR NON-LINEAR SYSTEMS. NA568 Mobile Robotics: Methods & Algorithms L07. KALMAN FILTERING FOR NON-LINEAR SYSTEMS NA568 Mobile Roboics: Mehods & Algorihms Today s Topic Quick review on (Linear) Kalman Filer Kalman Filering for Non-Linear Sysems Exended Kalman Filer (EKF)

More information

10. State Space Methods

10. State Space Methods . Sae Space Mehods. Inroducion Sae space modelling was briefly inroduced in chaper. Here more coverage is provided of sae space mehods before some of heir uses in conrol sysem design are covered in he

More information

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 2, May 2013, pp. 209 227 ISSN 0364-765X (prin) ISSN 1526-5471 (online) hp://dx.doi.org/10.1287/moor.1120.0562 2013 INFORMS On Boundedness of Q-Learning Ieraes

More information

Lecture 2 October ε-approximation of 2-player zero-sum games

Lecture 2 October ε-approximation of 2-player zero-sum games Opimizaion II Winer 009/10 Lecurer: Khaled Elbassioni Lecure Ocober 19 1 ε-approximaion of -player zero-sum games In his lecure we give a randomized ficiious play algorihm for obaining an approximae soluion

More information

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles Diebold, Chaper 7 Francis X. Diebold, Elemens of Forecasing, 4h Ediion (Mason, Ohio: Cengage Learning, 006). Chaper 7. Characerizing Cycles Afer compleing his reading you should be able o: Define covariance

More information

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Simulaion-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Week Descripion Reading Maerial 2 Compuer Simulaion of Dynamic Models Finie Difference, coninuous saes, discree ime Simple Mehods Euler Trapezoid

More information

20. Applications of the Genetic-Drift Model

20. Applications of the Genetic-Drift Model 0. Applicaions of he Geneic-Drif Model 1) Deermining he probabiliy of forming any paricular combinaion of genoypes in he nex generaion: Example: If he parenal allele frequencies are p 0 = 0.35 and q 0

More information

Optimal Power Cost Management Using Stored Energy in Data Centers

Optimal Power Cost Management Using Stored Energy in Data Centers Opimal Power Cos Managemen Using Sored Energy in Daa Ceners Rahul Urgaonkar, Bhuvan Urgaonkar, Michael J. Neely, Anand Sivasubramaniam Dep. of EE, Dep. of CSE Universiy of Souhern California, The Pennsylvania

More information

Stability and Bifurcation in a Neural Network Model with Two Delays

Stability and Bifurcation in a Neural Network Model with Two Delays Inernaional Mahemaical Forum, Vol. 6, 11, no. 35, 175-1731 Sabiliy and Bifurcaion in a Neural Nework Model wih Two Delays GuangPing Hu and XiaoLing Li School of Mahemaics and Physics, Nanjing Universiy

More information

Optimality Conditions for Unconstrained Problems

Optimality Conditions for Unconstrained Problems 62 CHAPTER 6 Opimaliy Condiions for Unconsrained Problems 1 Unconsrained Opimizaion 11 Exisence Consider he problem of minimizing he funcion f : R n R where f is coninuous on all of R n : P min f(x) x

More information

Georey E. Hinton. University oftoronto. Technical Report CRG-TR February 22, Abstract

Georey E. Hinton. University oftoronto.   Technical Report CRG-TR February 22, Abstract Parameer Esimaion for Linear Dynamical Sysems Zoubin Ghahramani Georey E. Hinon Deparmen of Compuer Science Universiy oftorono 6 King's College Road Torono, Canada M5S A4 Email: zoubin@cs.orono.edu Technical

More information

2.160 System Identification, Estimation, and Learning. Lecture Notes No. 8. March 6, 2006

2.160 System Identification, Estimation, and Learning. Lecture Notes No. 8. March 6, 2006 2.160 Sysem Idenificaion, Esimaion, and Learning Lecure Noes No. 8 March 6, 2006 4.9 Eended Kalman Filer In many pracical problems, he process dynamics are nonlinear. w Process Dynamics v y u Model (Linearized)

More information

Pade and Laguerre Approximations Applied. to the Active Queue Management Model. of Internet Protocol

Pade and Laguerre Approximations Applied. to the Active Queue Management Model. of Internet Protocol Applied Mahemaical Sciences, Vol. 7, 013, no. 16, 663-673 HIKARI Ld, www.m-hikari.com hp://dx.doi.org/10.1988/ams.013.39499 Pade and Laguerre Approximaions Applied o he Acive Queue Managemen Model of Inerne

More information

The Optimality of Two Prices: Maximizing Revenue in a Stochastic Network

The Optimality of Two Prices: Maximizing Revenue in a Stochastic Network PROC. OF 45TH ANNUAL ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING (INVITED PAPER), SEPT. 2007 1 The Opimaliy of Two Prices: Maximizing Revenue in a Sochasic Newor Longbo Huang, Michael

More information

Online Appendix to Solution Methods for Models with Rare Disasters

Online Appendix to Solution Methods for Models with Rare Disasters Online Appendix o Soluion Mehods for Models wih Rare Disasers Jesús Fernández-Villaverde and Oren Levinal In his Online Appendix, we presen he Euler condiions of he model, we develop he pricing Calvo block,

More information

Zürich. ETH Master Course: L Autonomous Mobile Robots Localization II

Zürich. ETH Master Course: L Autonomous Mobile Robots Localization II Roland Siegwar Margaria Chli Paul Furgale Marco Huer Marin Rufli Davide Scaramuzza ETH Maser Course: 151-0854-00L Auonomous Mobile Robos Localizaion II ACT and SEE For all do, (predicion updae / ACT),

More information

GMM - Generalized Method of Moments

GMM - Generalized Method of Moments GMM - Generalized Mehod of Momens Conens GMM esimaion, shor inroducion 2 GMM inuiion: Maching momens 2 3 General overview of GMM esimaion. 3 3. Weighing marix...........................................

More information

Sliding Mode Extremum Seeking Control for Linear Quadratic Dynamic Game

Sliding Mode Extremum Seeking Control for Linear Quadratic Dynamic Game Sliding Mode Exremum Seeking Conrol for Linear Quadraic Dynamic Game Yaodong Pan and Ümi Özgüner ITS Research Group, AIST Tsukuba Eas Namiki --, Tsukuba-shi,Ibaraki-ken 5-856, Japan e-mail: pan.yaodong@ais.go.jp

More information

MATH 5720: Gradient Methods Hung Phan, UMass Lowell October 4, 2018

MATH 5720: Gradient Methods Hung Phan, UMass Lowell October 4, 2018 MATH 5720: Gradien Mehods Hung Phan, UMass Lowell Ocober 4, 208 Descen Direcion Mehods Consider he problem min { f(x) x R n}. The general descen direcions mehod is x k+ = x k + k d k where x k is he curren

More information

PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD

PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD HAN XIAO 1. Penalized Leas Squares Lasso solves he following opimizaion problem, ˆβ lasso = arg max β R p+1 1 N y i β 0 N x ij β j β j (1.1) for some 0.

More information

Information Relaxations and Duality in Stochastic Dynamic Programs

Information Relaxations and Duality in Stochastic Dynamic Programs OPERATIONS RESEARCH Vol. 58, No. 4, Par 1 of 2, July Augus 2010, pp. 785 801 issn 0030-364X eissn 1526-5463 10 5804 0785 informs doi 10.1287/opre.1090.0796 2010 INFORMS Informaion Relaxaions and Dualiy

More information

Linear Response Theory: The connection between QFT and experiments

Linear Response Theory: The connection between QFT and experiments Phys540.nb 39 3 Linear Response Theory: The connecion beween QFT and experimens 3.1. Basic conceps and ideas Q: How do we measure he conduciviy of a meal? A: we firs inroduce a weak elecric field E, and

More information

O Q L N. Discrete-Time Stochastic Dynamic Programming. I. Notation and basic assumptions. ε t : a px1 random vector of disturbances at time t.

O Q L N. Discrete-Time Stochastic Dynamic Programming. I. Notation and basic assumptions. ε t : a px1 random vector of disturbances at time t. Econ. 5b Spring 999 C. Sims Discree-Time Sochasic Dynamic Programming 995, 996 by Chrisopher Sims. This maerial may be freely reproduced for educaional and research purposes, so long as i is no alered,

More information

RANDOM LAGRANGE MULTIPLIERS AND TRANSVERSALITY

RANDOM LAGRANGE MULTIPLIERS AND TRANSVERSALITY ECO 504 Spring 2006 Chris Sims RANDOM LAGRANGE MULTIPLIERS AND TRANSVERSALITY 1. INTRODUCTION Lagrange muliplier mehods are sandard fare in elemenary calculus courses, and hey play a cenral role in economic

More information

E β t log (C t ) + M t M t 1. = Y t + B t 1 P t. B t 0 (3) v t = P tc t M t Question 1. Find the FOC s for an optimum in the agent s problem.

E β t log (C t ) + M t M t 1. = Y t + B t 1 P t. B t 0 (3) v t = P tc t M t Question 1. Find the FOC s for an optimum in the agent s problem. Noes, M. Krause.. Problem Se 9: Exercise on FTPL Same model as in paper and lecure, only ha one-period govenmen bonds are replaced by consols, which are bonds ha pay one dollar forever. I has curren marke

More information

Notes on Kalman Filtering

Notes on Kalman Filtering Noes on Kalman Filering Brian Borchers and Rick Aser November 7, Inroducion Daa Assimilaion is he problem of merging model predicions wih acual measuremens of a sysem o produce an opimal esimae of he curren

More information

) were both constant and we brought them from under the integral.

) were both constant and we brought them from under the integral. YIELD-PER-RECRUIT (coninued The yield-per-recrui model applies o a cohor, bu we saw in he Age Disribuions lecure ha he properies of a cohor do no apply in general o a collecion of cohors, which is wha

More information

A DELAY-DEPENDENT STABILITY CRITERIA FOR T-S FUZZY SYSTEM WITH TIME-DELAYS

A DELAY-DEPENDENT STABILITY CRITERIA FOR T-S FUZZY SYSTEM WITH TIME-DELAYS A DELAY-DEPENDENT STABILITY CRITERIA FOR T-S FUZZY SYSTEM WITH TIME-DELAYS Xinping Guan ;1 Fenglei Li Cailian Chen Insiue of Elecrical Engineering, Yanshan Universiy, Qinhuangdao, 066004, China. Deparmen

More information

SZG Macro 2011 Lecture 3: Dynamic Programming. SZG macro 2011 lecture 3 1

SZG Macro 2011 Lecture 3: Dynamic Programming. SZG macro 2011 lecture 3 1 SZG Macro 2011 Lecure 3: Dynamic Programming SZG macro 2011 lecure 3 1 Background Our previous discussion of opimal consumpion over ime and of opimal capial accumulaion sugges sudying he general decision

More information

Chapter 2. First Order Scalar Equations

Chapter 2. First Order Scalar Equations Chaper. Firs Order Scalar Equaions We sar our sudy of differenial equaions in he same way he pioneers in his field did. We show paricular echniques o solve paricular ypes of firs order differenial equaions.

More information

Traffic Splitting with Network Calculus for Mesh Sensor Networks

Traffic Splitting with Network Calculus for Mesh Sensor Networks Traffic Spliing wih Nework Calculus for Mesh Sensor Neworks Huimin She, Zhonghai Lu, Axel Jansch, Li-Rong Zheng, Dian Zhou Dep. of Elecronic, Compuer and Sofware Sysems, Royal Insiue of Technology, Sweden

More information

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017 Two Popular Bayesian Esimaors: Paricle and Kalman Filers McGill COMP 765 Sep 14 h, 2017 1 1 1, dx x Bel x u x P x z P Recall: Bayes Filers,,,,,,, 1 1 1 1 u z u x P u z u x z P Bayes z = observaion u =

More information

Macroeconomic Theory Ph.D. Qualifying Examination Fall 2005 ANSWER EACH PART IN A SEPARATE BLUE BOOK. PART ONE: ANSWER IN BOOK 1 WEIGHT 1/3

Macroeconomic Theory Ph.D. Qualifying Examination Fall 2005 ANSWER EACH PART IN A SEPARATE BLUE BOOK. PART ONE: ANSWER IN BOOK 1 WEIGHT 1/3 Macroeconomic Theory Ph.D. Qualifying Examinaion Fall 2005 Comprehensive Examinaion UCLA Dep. of Economics You have 4 hours o complee he exam. There are hree pars o he exam. Answer all pars. Each par has

More information

Expert Advice for Amateurs

Expert Advice for Amateurs Exper Advice for Amaeurs Ernes K. Lai Online Appendix - Exisence of Equilibria The analysis in his secion is performed under more general payoff funcions. Wihou aking an explici form, he payoffs of he

More information

Control of computer chip semi-conductor wafer fabs

Control of computer chip semi-conductor wafer fabs Conrol of compuer chip semi-conducor wafer fabs Cos: x9 $ Reurn: years Cycle ime: 6 weeks WIP: 6, wafers, 8x6 $ Challenge: Conrol he queues a ~5 work seps Conrol of Manufacuring Sysems:! Sochasic racking

More information

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions Muli-Period Sochasic Models: Opimali of (s, S) Polic for -Convex Objecive Funcions Consider a seing similar o he N-sage newsvendor problem excep ha now here is a fixed re-ordering cos (> 0) for each (re-)order.

More information

CHAPTER 10 VALIDATION OF TEST WITH ARTIFICAL NEURAL NETWORK

CHAPTER 10 VALIDATION OF TEST WITH ARTIFICAL NEURAL NETWORK 175 CHAPTER 10 VALIDATION OF TEST WITH ARTIFICAL NEURAL NETWORK 10.1 INTRODUCTION Amongs he research work performed, he bes resuls of experimenal work are validaed wih Arificial Neural Nework. From he

More information

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H. ACE 56 Fall 005 Lecure 5: he Simple Linear Regression Model: Sampling Properies of he Leas Squares Esimaors by Professor Sco H. Irwin Required Reading: Griffihs, Hill and Judge. "Inference in he Simple

More information

Appendix 14.1 The optimal control problem and its solution using

Appendix 14.1 The optimal control problem and its solution using 1 Appendix 14.1 he opimal conrol problem and is soluion using he maximum principle NOE: Many occurrences of f, x, u, and in his file (in equaions or as whole words in ex) are purposefully in bold in order

More information

Modal identification of structures from roving input data by means of maximum likelihood estimation of the state space model

Modal identification of structures from roving input data by means of maximum likelihood estimation of the state space model Modal idenificaion of srucures from roving inpu daa by means of maximum likelihood esimaion of he sae space model J. Cara, J. Juan, E. Alarcón Absrac The usual way o perform a forced vibraion es is o fix

More information

Notes for Lecture 17-18

Notes for Lecture 17-18 U.C. Berkeley CS278: Compuaional Complexiy Handou N7-8 Professor Luca Trevisan April 3-8, 2008 Noes for Lecure 7-8 In hese wo lecures we prove he firs half of he PCP Theorem, he Amplificaion Lemma, up

More information

Block Diagram of a DCS in 411

Block Diagram of a DCS in 411 Informaion source Forma A/D From oher sources Pulse modu. Muliplex Bandpass modu. X M h: channel impulse response m i g i s i Digial inpu Digial oupu iming and synchronizaion Digial baseband/ bandpass

More information

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION Vol. XI Control of Stochastic Systems - P.R. Kumar

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION Vol. XI Control of Stochastic Systems - P.R. Kumar CONROL OF SOCHASIC SYSEMS P.R. Kumar Deparmen of Elecrical and Compuer Engineering, and Coordinaed Science Laboraory, Universiy of Illinois, Urbana-Champaign, USA. Keywords: Markov chains, ransiion probabiliies,

More information

Map Task Scheduling in MapReduce with Data Locality: Throughput and Heavy-Traffic Optimality

Map Task Scheduling in MapReduce with Data Locality: Throughput and Heavy-Traffic Optimality Map Task Scheduling in MapReduce wih Daa Localiy: Throughpu and Heavy-Traffic Opimaliy Weina Wang, Kai Zhu and Lei Ying Elecrical, Compuer and Energy Engineering Arizona Sae Universiy Tempe, Arizona 85287

More information

Energy Storage and Renewables in New Jersey: Complementary Technologies for Reducing Our Carbon Footprint

Energy Storage and Renewables in New Jersey: Complementary Technologies for Reducing Our Carbon Footprint Energy Sorage and Renewables in New Jersey: Complemenary Technologies for Reducing Our Carbon Fooprin ACEE E-filliaes workshop November 14, 2014 Warren B. Powell Daniel Seingar Harvey Cheng Greg Davies

More information

Lecture Notes 2. The Hilbert Space Approach to Time Series

Lecture Notes 2. The Hilbert Space Approach to Time Series Time Series Seven N. Durlauf Universiy of Wisconsin. Basic ideas Lecure Noes. The Hilber Space Approach o Time Series The Hilber space framework provides a very powerful language for discussing he relaionship

More information

International Journal of Computer Science Trends and Technology (IJCST) Volume 3 Issue 6, Nov-Dec 2015

International Journal of Computer Science Trends and Technology (IJCST) Volume 3 Issue 6, Nov-Dec 2015 Inernaional Journal of Compuer Science Trends and Technology (IJCST) Volume Issue 6, Nov-Dec 05 RESEARCH ARTICLE OPEN ACCESS An EPQ Model for Two-Parameer Weibully Deerioraed Iems wih Exponenial Demand

More information

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits DOI: 0.545/mjis.07.5009 Exponenial Weighed Moving Average (EWMA) Char Under The Assumpion of Moderaeness And Is 3 Conrol Limis KALPESH S TAILOR Assisan Professor, Deparmen of Saisics, M. K. Bhavnagar Universiy,

More information

Econ107 Applied Econometrics Topic 7: Multicollinearity (Studenmund, Chapter 8)

Econ107 Applied Econometrics Topic 7: Multicollinearity (Studenmund, Chapter 8) I. Definiions and Problems A. Perfec Mulicollineariy Econ7 Applied Economerics Topic 7: Mulicollineariy (Sudenmund, Chaper 8) Definiion: Perfec mulicollineariy exiss in a following K-variable regression

More information

Unit Root Time Series. Univariate random walk

Unit Root Time Series. Univariate random walk Uni Roo ime Series Univariae random walk Consider he regression y y where ~ iid N 0, he leas squares esimae of is: ˆ yy y y yy Now wha if = If y y hen le y 0 =0 so ha y j j If ~ iid N 0, hen y ~ N 0, he

More information

On Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction

On Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction On Mulicomponen Sysem Reliabiliy wih Microshocks - Microdamages Type of Componens Ineracion Jerzy K. Filus, and Lidia Z. Filus Absrac Consider a wo componen parallel sysem. The defined new sochasic dependences

More information

Lecture 20: Riccati Equations and Least Squares Feedback Control

Lecture 20: Riccati Equations and Least Squares Feedback Control 34-5 LINEAR SYSTEMS Lecure : Riccai Equaions and Leas Squares Feedback Conrol 5.6.4 Sae Feedback via Riccai Equaions A recursive approach in generaing he marix-valued funcion W ( ) equaion for i for he

More information

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin ACE 56 Fall 005 Lecure 4: Simple Linear Regression Model: Specificaion and Esimaion by Professor Sco H. Irwin Required Reading: Griffihs, Hill and Judge. "Simple Regression: Economic and Saisical Model

More information

Electrical and current self-induction

Electrical and current self-induction Elecrical and curren self-inducion F. F. Mende hp://fmnauka.narod.ru/works.hml mende_fedor@mail.ru Absrac The aricle considers he self-inducance of reacive elemens. Elecrical self-inducion To he laws of

More information

BU Macro BU Macro Fall 2008, Lecture 4

BU Macro BU Macro Fall 2008, Lecture 4 Dynamic Programming BU Macro 2008 Lecure 4 1 Ouline 1. Cerainy opimizaion problem used o illusrae: a. Resricions on exogenous variables b. Value funcion c. Policy funcion d. The Bellman equaion and an

More information

Chapter 2. Models, Censoring, and Likelihood for Failure-Time Data

Chapter 2. Models, Censoring, and Likelihood for Failure-Time Data Chaper 2 Models, Censoring, and Likelihood for Failure-Time Daa William Q. Meeker and Luis A. Escobar Iowa Sae Universiy and Louisiana Sae Universiy Copyrigh 1998-2008 W. Q. Meeker and L. A. Escobar. Based

More information

A linear algorithm for reliable predictive network control

A linear algorithm for reliable predictive network control A linear algorihm for reliable predicive nework conrol Richard Schoeffauer and Gerhard Wunder Heisenberg CIT Group, Free Universiy of Berlin; Berlin, Germany richardschoeffauer@fu-berlinde, gwunder@fu-berlinde

More information

arxiv: v1 [cs.dc] 8 Mar 2012

arxiv: v1 [cs.dc] 8 Mar 2012 Consensus on Moving Neighborhood Model of Peerson Graph arxiv:.9v [cs.dc] 8 Mar Hannah Arend and Jorgensen Jos Absrac In his paper, we sudy he consensus problem of muliple agens on a kind of famous graph,

More information

Orientation. Connections between network coding and stochastic network theory. Outline. Bruce Hajek. Multicast with lost packets

Orientation. Connections between network coding and stochastic network theory. Outline. Bruce Hajek. Multicast with lost packets Connecions beween nework coding and sochasic nework heory Bruce Hajek Orienaion On Thursday, Ralf Koeer discussed nework coding: coding wihin he nework Absrac: Randomly generaed coded informaion blocks

More information

Robust estimation based on the first- and third-moment restrictions of the power transformation model

Robust estimation based on the first- and third-moment restrictions of the power transformation model h Inernaional Congress on Modelling and Simulaion, Adelaide, Ausralia, 6 December 3 www.mssanz.org.au/modsim3 Robus esimaion based on he firs- and hird-momen resricions of he power ransformaion Nawaa,

More information

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still.

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still. Lecure - Kinemaics in One Dimension Displacemen, Velociy and Acceleraion Everyhing in he world is moving. Nohing says sill. Moion occurs a all scales of he universe, saring from he moion of elecrons in

More information

The Arcsine Distribution

The Arcsine Distribution The Arcsine Disribuion Chris H. Rycrof Ocober 6, 006 A common heme of he class has been ha he saisics of single walker are ofen very differen from hose of an ensemble of walkers. On he firs homework, we

More information

Math 334 Fall 2011 Homework 11 Solutions

Math 334 Fall 2011 Homework 11 Solutions Dec. 2, 2 Mah 334 Fall 2 Homework Soluions Basic Problem. Transform he following iniial value problem ino an iniial value problem for a sysem: u + p()u + q() u g(), u() u, u () v. () Soluion. Le v u. Then

More information

Delay Reduction via Lagrange Multipliers in Stochastic Network Optimization

Delay Reduction via Lagrange Multipliers in Stochastic Network Optimization PROC. OF 7TH INTL. SYMPOSIM ON MODELING AND OPTIMIZATION IN MOBILE, AD HOC, AND WIRELESS NETWORKS (WIOPT), JNE 009 1 Delay Reducion via Lagrange Mulipliern Sochasic Nework Opimizaion Longbo Huang, Michael

More information

Christos Papadimitriou & Luca Trevisan November 22, 2016

Christos Papadimitriou & Luca Trevisan November 22, 2016 U.C. Bereley CS170: Algorihms Handou LN-11-22 Chrisos Papadimiriou & Luca Trevisan November 22, 2016 Sreaming algorihms In his lecure and he nex one we sudy memory-efficien algorihms ha process a sream

More information

Appendix to Creating Work Breaks From Available Idleness

Appendix to Creating Work Breaks From Available Idleness Appendix o Creaing Work Breaks From Available Idleness Xu Sun and Ward Whi Deparmen of Indusrial Engineering and Operaions Research, Columbia Universiy, New York, NY, 127; {xs2235,ww24}@columbia.edu Sepember

More information

1. An introduction to dynamic optimization -- Optimal Control and Dynamic Programming AGEC

1. An introduction to dynamic optimization -- Optimal Control and Dynamic Programming AGEC This documen was generaed a :45 PM 8/8/04 Copyrigh 04 Richard T. Woodward. An inroducion o dynamic opimizaion -- Opimal Conrol and Dynamic Programming AGEC 637-04 I. Overview of opimizaion Opimizaion is

More information

This document was generated at 1:04 PM, 09/10/13 Copyright 2013 Richard T. Woodward. 4. End points and transversality conditions AGEC

This document was generated at 1:04 PM, 09/10/13 Copyright 2013 Richard T. Woodward. 4. End points and transversality conditions AGEC his documen was generaed a 1:4 PM, 9/1/13 Copyrigh 213 Richard. Woodward 4. End poins and ransversaliy condiions AGEC 637-213 F z d Recall from Lecure 3 ha a ypical opimal conrol problem is o maimize (,,

More information

Tom Heskes and Onno Zoeter. Presented by Mark Buller

Tom Heskes and Onno Zoeter. Presented by Mark Buller Tom Heskes and Onno Zoeer Presened by Mark Buller Dynamic Bayesian Neworks Direced graphical models of sochasic processes Represen hidden and observed variables wih differen dependencies Generalize Hidden

More information

Exploiting Wireless Channel State Information for Throughput Maximization

Exploiting Wireless Channel State Information for Throughput Maximization Exploiing Wireless Channel ae Informaion for Throughpu Maximizaion V Tsibonis, L Georgiadis, L Tassiulas Absrac We consider he problem of scheduling packes over a number of channels wih ime varying conneciviy

More information

The field of mathematics has made tremendous impact on the study of

The field of mathematics has made tremendous impact on the study of A Populaion Firing Rae Model of Reverberaory Aciviy in Neuronal Neworks Zofia Koscielniak Carnegie Mellon Universiy Menor: Dr. G. Bard Ermenrou Universiy of Pisburgh Inroducion: The field of mahemaics

More information

Object tracking: Using HMMs to estimate the geographical location of fish

Object tracking: Using HMMs to estimate the geographical location of fish Objec racking: Using HMMs o esimae he geographical locaion of fish 02433 - Hidden Markov Models Marin Wæver Pedersen, Henrik Madsen Course week 13 MWP, compiled June 8, 2011 Objecive: Locae fish from agging

More information