Multi-Sensor Control for Multi-Target Tracking Using Cauchy-Schwarz Divergence
|
|
- Erin Grant
- 5 years ago
- Views:
Transcription
1 Mult-Senor Control for Mult-Target Tracng Ung Cauchy-Schwarz Dvergence Meng Jang, We Y and Lngjang Kong School of Electronc Engneerng Unverty of Electronc Scence and Technology of Chna Emal:{uoy, lngjang.ong}@gmal.com arxv: v1 [c.sy] 28 Mar 216 Abtract The paper addree the problem of mult-enor control for mult-target tracng va labelled random fnte et (RFS) n the enor networ ytem. Baed on an nformaton theoretc dvergence meaure, namely Cauchy-Schwarz (CS) dvergence whch admt a cloed form oluton for GLMB dente, we propoe two novel mult-enor control approache n the framewor of generalzed Covarance Interecton (GCI). The frt jont decon mang (JDM) method optmal and can acheve overall good performance, whle the econd ndepent decon mang (IDM) method uboptmal a a fat realzaton wth maller amount of computaton. Smulaton n challengng tuaton preented to verfy the effectvene of the two propoed approache. I. INTRODUCTION Senor networ ytem have receved tremou attenton n lat decade due to ther ucceful applcaton that range from vehcular networ to battlefeld detecton and tracng [1]. In many practcal tuaton, due to communcaton and computatonal contrant, t requred that lmted amount of enor tae rght acton. In uch cae, the problem of enor control to fnd a member of the command et that can reult n bet meaurement for flterng purpoe [2]. In general, enor control compre two underlyng component, a mult-target flterng proce n conjuncton wth an optmal decon-mang method. Mult-target flterng ha been recently nvetgated n a more prncpled way due to the pont proce theory or fnte et tattc (FISST) baed mult-target tracng methodology [3]. Among thee random fnte et (RFS) baed method, the promng generalzed labeled mult-bernoull (GLMB) flter [4], [5], or mply the Vo-Vo flter, poee ome ueful analytcal properte [6] and a cloed form oluton to the Baye mult-target flter, can not only produce trajectore formally but alo outperform the probablty hypothe denty (PD) flter [7], cardnalzed PD (CPD) flter [8] and mult-bernoull (MB) flter [9]. Another mportant component of enor control oluton a decon-mang proce, whch motly reort to optmzaton of an objectve functon and generally fall nto two categore. The frt one ta-baed approach, enor control method are degned wth a drect focu on the expected performance and the objectve functon formulated a a cot functon, example of uch cot functon nclude etmated target cardnalty varance [1], [11], poteror expected error of cardnalty and tate (PEECS) [12], [13] and optmal ub-pattern agnment (OSPA) dtance [14]. The ta-baed approach ueful n ome tuaton epecally where the objectve functon can be formulated n the form of a ngle crteron, but there a challengng problem n the cae of multple competng objectve. To olve or avod th problem, the econd one nformaton-baed approach whch trve to quantfy the nformaton content of the mult-target dtrbuton, am at obtanng uperor overall performance acro multple ta objectve and the objectve functon formulated a a reward functon. The mot common choce of reward functon are baed on ome nformaton theoretc dvergence meaure uch a KullbacCLebler (KL) dvergence [15], [16] and more generally the Rény dvergence [17] [19]. owever, a major lmtaton of utlzng KL or Rény dvergence ther gnfcant computatonal cot, and hence mot of the tme, one ha to reort to numercal ntegraton method uch a Monte Carlo (MC) method to derve analytcally reult. An alternatve nformaton dvergence meaure the Cauchy-Schwarz (CS) dvergence. Ung th meaure, oang et al provded tractable formulaton between the probablty dente of two Poon pont procee [2], later, Beard et al exted the reult to two GLMB dente [6], [21] and preented an analytc expreon, whch opened the door to enor control cheme wth GLMB Model baed on nformaton-baed approach. The CS control wth GLMB model account for target trajectore n a prncpled manner, whch not poble ung other tracng method. When the urvellance area very large or target move n complex movement, one enor wth lmted enng range (LSR) not competent to the ta of mult-target tracng, enor networ ytem and ubequent multple enor control are neceary. Inpred by the good performance acheved by enor control wth GLMB model baed on CS dvergence, where Beard et al only condered ngle enor, n th paper, we addre the problem of mult-enor control for mult-target tracng ung CS dvergence va labelled random fnte et (RFS). To be pecfc, we ue Vo-Vo flter to enure local tracng performance, and Generalzed Covarance Interecton (GCI) fuon [22] [24] to maxmze nformaton content of the mult-target dtrbuton. The ey contrbuton of th paper are two tractable approache of mult-enor control, the one optmal wth a lttle complex calculaton and the other uboptmal a a fat realzaton. Smulaton reult verfy both propoed approache can perform well n complex tuaton.
2 2 II. BACKGROUND Th ecton provde bacground materal on labelled mult-target flterng, GCI fuon and Cauchy-Schwarz dvergence whch are neceary for the reult of th paper. For further detal, we refer the reader to [4], [6], [23], [24]. A. Notaton In th paper, we adhere to the conventon that ngle-target tate are denoted by the mall letter, e.g.,x, x whle multtarget tate are denoted by captal letter, e.g.,x, X. Symbol for labeled tate and ther dtrbuton/tattc (ngle-target or mult-target) are bolded to dtnguh them from unlabeled one, e.g.,x,x, π, etc. To be more pecfc, the labeled ngle target tate x contructed by augmentng a tate x X wth a label l L. Obervaton generated by ngle-target tate are denoted by the mall letter, e.g., z, and the multtarget obervaton are denoted by the captal letter, e.g., Z. Addtonally, blacboard bold letter repreent pace, e.g., the tate pace repreented by X, the label pace by L, and the obervaton pace by Z. The collecton of all fnte et of X denoted by F(X). Moreover, n order to upport arbtrary argument le et, vector and nteger, the generalzed Kronecer delta functon gven by { 1, f X = Y δ Y (X) (1), otherwe and δx denote the et ntegral [3] defned by 1 f(x)δx= f({x 1,,x n })dx 1 dx n (2) n! B. GLMB RFS n= An mportant labeled RFS the GLMB RFS [4], whch a cla of tractable model for on-lne Bayean nference [3] that allevate the lmtaton of the Poon model. Under the tandard mult-object model, the GLMB a conjugate pror that alo cloed under the Chapman-Kolmogorov equaton. Let L : X L X be the projecton L((x,l)) = l, and (X) = δ X ( L(X) ) denote the dtnct label ndcator. A GLMB an RFS on X L dtrbuted accordng to π(x) = (X) c Cw (c) (L(X))[p (c) ] X (3) where C a dcrete ndex et. The weght w (c) (L) and the patal dtrbuton p (c) atfy the normalzaton condton w (c) (L) = 1 L L c C p (c) (x,l)dx = 1 Further, a δ-glmb RFS [4], [5] wth tate pace X and (dcrete) label pace L a pecal cae of a GLMB RFS wth C = F(L) Ξ w (c) (L) = w (I,ξ) δ I (L) p (c) = p (I,ξ) = p (ξ) where Ξ a dcrete pace, ξ are realzaton of Ξ, and I denote a et of trac label. In target tracng applcaton, the dcrete pace Ξ typcally repreent the htory of trac to meaurement aocaton. A δ-glmb RFS thu a pecal cae of a GLMB RFS but wth a partcular tructure on the ndex pace whch are naturally n target tracng applcaton. The δ-glmb RFS ha denty π(x) = (X) w (I,ξ) δ I (L(X))[p (ξ) ] X (4) (I,ξ) F(L) Ξ C. Cauchy-Schwarz Dvergence Compared wth Kullbac-Lebler dvergence or Rény dvergence, whch are mot commonly ued meaure of nformaton gan, CS dvergence [6], [21] ha a mathematcal form whch more amenable to cloed form oluton. Ung the relatonhp between probablty denty and belef denty, the CS dvergence between two RFS, wth repectve belef dente φ and ϕ, gven by D CS (φ,ϕ) = ln K X φ(x)ϕ(x)δx K X φ 2 (X)δX K X ϕ 2 (X)δX (5) where K the unt of hyper-volume n X. In partcular, Cauchy-Schwarz dvergence ha a cloed form for GLMB dente, n the cae where the ndvdual target dente are Gauan mxture. For two GLMB wth belef dente φ(x) = (X) c C ψ(x) = (X) d C w (c) φ (L(X))[p(c) φ ]X (6) w (d) ψ (L(X))[p(d) ψ ]X (7) the Cauchy-Schwarz dvergence between φ and ψ gven by where D CS (φ,ψ) = ln ζ(φ,ψ) = L L [K c Cd D ζ(φ, ψ) ζ(φ,φ)ζ(ψ,ψ) (8) w (c) φ (L)w(d) ψ (L) (9) p (c) φ (x, )p(d) ψ (x, )dx]l Cloed form of the analytcal expreon ung CS dvergence combne GLMB dente and nformaton theoretc dvergence meaure hence lead to a more effcent mplementaton of enor control. D. Dtrbuted Fuon In the context of enor networ ytem wth LRS, where each enor ha a fnte feld of vew (FoV), dtrbuted fuon neceary to mae the bet ue of local dtrbuton nformaton n order to olve the hadowng effect. The GCI wa propoed by Mahler [22] pecfcally to ext FISST to enor networ ytem, whch capable to fue both Gauan and non-gauan formed mult-target dtrbuton from dfferent enor wth completely unnown correlaton.
3 3 Baed on GCI, wth the aumpton that all the enor node hare the ame label pace for the brth proce, Fantacc et al propoed the GCI fuon wth labeled et flter by ue the content label. The reult nclude conenu margnalzedδ- GLMB (CMδ-GLMB) and conenu LMB (CLMB) tracng flter [23]. 1) CMδ-GLMB : Suppoe that each enor = 1,...,N provded wth an Mδ-GLMB denty π of the form π = (X) L F(L) δ L (L(X))w (L) [ ] X (1) where N the total enor number and fuon weght ω (,1), N =1 ω = 1, then the fued dtrbuton gven a follow: π = (X) δ L (L(X))w (L) [ ] X (11) where w (L) = = =1 F L=1 L F(L) w (L) =1 N ( =1 w (F) [ N =1 [ N dx ] ( ) L ω (x, ) dx =1 ( ] F p (F) (x, ) dx 2) CLMB : Suppoe that each enor = 1,...,N provded wth a LMB denty π of the form{(r (l), )} l L, where N the total enor number and fuon weght ω (,1), N =1 ω = 1, then the fued dtrbuton of the form where r (l) = = =1 π = {(r (l), )} l L (12) N =1 1 r (l) =1 N ( =1 ( r (l) (x) dx N + dx =1 ( r (l) (x) dx Conenu algorthm can fue n a fully dtrbuted and calable way the nformaton collected from the multple heterogeneou and geographcally dpered enor, and therefore have a gnfcant mpact on the etmaton performance of the tracng ytem. III. MULTI-SENSOR CONTROL USING CS DIVERGENCE In mot target tracng cenaro, the enor may perform varou acton that can maxmze the tracng obervablty, and can therefore nfluence the etmaton performance of the tracng ytem. Typcally, uch acton may nclude changng the poton, alterng the enor operatng parameter, orentaton or moton of the enor platform and o on, whch n turn affect the enor ablty to detect and trac target. In the context of enor networ ytem, where there are more than one enor watng to be deployed, the allowable control acton may ncreae exponentally and hence the control of mult-enor a hgh-dmenonal optmzaton problem. Therefore, mang control decon by manual nterventon or ome determntc control polcy whch provde no guarantee of optmalty, not a good choce. Compared wth ngle enor control, there are ome challengng problem n mult-enor control uch a aforementoned hgh-dmenonal optmzaton problem and nformaton fuon problem nduced by the meaurement collected from the multple enor. In th ecton, we ee tractable oluton for mult-enor control for mult-target tracng wth GLMB model. A. Problem Formulaton In enor networ ytem, one or more enor are the drect output of the decon-mang component of the control oluton, a uch, the focu ha tradtonally been placed on mprovng the decon-mang component. owever, the mult-target tracng component alo play a gnfcant role n the overall performance of the cheme n term of accuracy and robutne. Inpred by the veratle GLMB model whch offer good trade-off between tractablty and fdelty, n flterng tage, we ue the Vo-Vo flter [4], [5] a local enor and GCI fuon to fue the nformaton collected from the multple enor n order to acheve overall uperor performance, the procedure decrbed a follow: 1) At tme tep, wth meaurement Z = {z1,,z 2,,...,z m, } where the ubcrpt denote current tme and upercrpt denote equence number of enor, each enor node = 1,...,N locally perform predcton and update ung Vo-Vo flter, the detal can be found n [5]. 2) Implement the GCI fuon wth local poteror dtrbuton π to derve the fued dtrbuton π, the upercrpt denote fued dtrbuton. Note that one need to convert δ- GLMB poteror dtrbuton to Mδ-GLMB\LMB dtrbuton for conenu fuon method ung (11) or (12). 3) After fuon, an etmate of the object et ˆX obtaned from the cardnalty probablty ma functon and the locaton PDF ung MAP technque. A peudo-code of flterng tage gven n Algorthm 1. In control trategy, we adhere to the conventon that formulatng the enor control problem a a Partally Oberved Marov Decon Proce (POMDP) ung FISST [25] and defnng the followng notaton: π ( Z 1: ) the poteror denty for enor at tme, C the control acton pace for enor and hence the N multple enor control acton pace C = C 1 C N, the length of control horzon, the π ( Z 1: ) predcted denty at tme baed on nown meaurement from tme 1 to tme, Z 1: (c 1,...,c N ) the collecton of meaurement for
4 4 Algorthm 1: Flterng Procedure INPUT: π 1{X Z 1: 1},Z OUTPUT: π {X Z 1:}, π for = 1 : N do local predcton local update π {X Z 1:} GCI( π {X Z 1:}) π MAP( π ) ˆX enor that would be oberved from tme 1 up to wth executed control acton (c 1,...,c N ) C at tme, note that c C a vector compoed of all poble acton what a enor can tae, uch a changng drecton of movement, velocty, power and o on. We ue CS dvergence a reward functon at the control horzon whch meaured between the predcted and poteror mult-target denty: R(c 1,...,c N ) = D CS ( π predcton, π update ) (13) then the optmal control acton decded by maxmng the expected value of the reward functon R(c 1,...,c N ) over the allowable acton pace C: (ĉ 1,...,ĉ N ) = arg max (c 1,...,c N) C EAP(R(c 1,...,c N )) (14) Note that the above expected reward not avalable to analytc oluton, o we reort to Monte Carlo ntegraton, EAP(R(c 1,...,c N )) 1 M M R (j) (c 1,...,c N ) (15) j=1 where M denote the number of ample. Alo for th reaon, we prefer CS dvergence whch provde a cloed-form oluton wth GLMB model to calculate R (j) (c 1,...,c N ), can allevate the de effect nduced by the Monte Carlo technque (15). In what followng we detal the degn of predcted dtrbuton and poteror dtrbuton n (13) and preent two mult-enor control approache. B. Mult-Senor Control Strategy JOINT DECISION MAKING ALGORITM In order to mae the bet ue of enor networ and overall collected nformaton, we propoe an optmal mult-enor control approach, referred to jont decon mang (JDM) algorthm. In th method, the flterng tage performed a decrbed n Algorthm 1, the fued denty π wll be ued for mult-target ample n order to olve the hadowng effect of ngle enor wth LSR and to compute the predcted denty at the of the control horzon. The pecfc procedure are a follow: 1) Mult-target Sample: At decon tme tep, draw a et Ψ S of M mult-target ample from fued dtrbuton π, t manly degned for dervng numercal analytcal reoluton of CS reward functon. 2) Peudo-Predcton: Compute the predcted denty at the of the control horzon π, whch wll be later ued a one term of computng CS dvergence, by carryng out repeated predcton tep of Vo-Vo flter, wthout traget brth or death, for th reaon, we ue the term peudo-predcton. 3) Generate predcted deal meaurement (PIMS): For each enor = 1,...,N and each mult-target ample X (j) Ψ S, generatng PIMS Z 1: (c,x (j) ) wth current control acton c C baed on ntal predcted trajectory n ample X (j), more detal n [21], [26]. 4) Run Vo-Vo Flter Recuron: Run each Vo-Vo flter wth ntal local poteror dtrbuton π {X Z 1: } ung PIMS Z 1: (c,x (j) ) to get the peudo updated dtrbuton π {X Z 1:, Z 1: (c,x (j) )}, we wll ue the term flter to denote Vo-Vo flter recuron [5]. 5) GCI Fuon: For mult-enor, for each poble control acton combnaton (c 1,...,c N ) C, perform the GCI fuon wth peudo updated dtrbuton π {X Z 1:, Z 1: (c,x (j) )} to get the fued peudo updated dtrbuton π (c 1,...,c N,X (j) ), t wll be later ued a another term of computng CS dvergence. 6) Compute Each Reward: Compute CS reward functon for each control acton combnaton and each ample ung (8), R (j) (c 1,...,c N ) = D CS ( π, π (c 1,...,c N,X (j) )) (16) after the computaton of (16) for all ample n et Ψ S, we then compute the expected value of the reward functon R(c 1,...,c N ) = EAP(R (j) (c 1,...,c N )) 1 M R (j) (c 1,...,c N ) M j=1 (17) 7) Jont Decon Mang: Maxmze the expected value of the reward functon R(c 1,...,c N ) over the allowable acton pace C ung (14). A peudo-code of above control tage hown n Algorthm 2. Note that n the JDM algorthm, GCI fuon ha been ued both n flterng tage and CS control tage, am at maxmzng obervaton nformaton content and overall CS dvergence, to enure multple enor move n drecton where the overall performance atfyng. Moreover, n order to reduce the computaton burden of the JDM algorthm, whch manly nduced by allowable control acton combnaton wth computaton complexty O( C 1 C N ), one can reort to mportance amplng technque, more detal n [27]. INDEPENDENT DECISION MAKING ALGORITM We alo propoe another uboptmal mult-enor control approach, referred to ndepent decon mang (IDM) algorthm. In th method, the flterng tage ame but the control tage mplfed a a fat mplementaton. In partcular, the GCI fuon only performed n flterng tage and each enor mae control decon ndepently n control tage, whch enable parallel executon of the control
5 5 Algorthm 2: JDM Procedure INPUT: π, π {X Z1:},C OUTPUT: (ĉ 1,...,ĉ N) Mult-target Sample: π Ψ S = {X (1),...X (M) } Peudo-Predcton: for ter = +1 : + do π π for = 1 : N do for each c C do for each X (j) Ψ S do Generate PIMS: X (j) Z 1:(c,X (j) ) Run Vo-Vo Flter Recuron: flter( π {X Z1:}, Z 1:(c,X (j) )) π {X Z1:, Z 1:(c,X (j) )} GCI Fuon: for each (c 1,...,c N) C do for each X (j) Ψ S do GCI( π 1 {X Z1:, 1 Z 1:(c,X (j) )},..., π N {X Z1:, N Z 1:(c N N,X (j) )}) π (c 1,...,c N,X (j) ) Compute Each Reward: D CS( π, π (c 1,...,c N,X (j) )) R (j) (c 1,...,c N) EAP(R (j) (c 1,...,c N)) R(c 1,...,c N) Jont Decon Mang: arg max (R(c 1,...,c N)) (ĉ 1,...,ĉ N) (c 1,...,c N ) C tep, and therefore the computaton complexty of allowable control acton reduced to O( C C N ). A peudocode of IDM algorthm hown n Algorthm 3. Note that the fued dtrbuton π ued n mult-target ample and peudo-predcton, whch can enure obervablty n control tage o that avod mang myopc decon. A comparon between JDM algorthm and IDM algorthm wth two enor llutrated n Fg. 1. GCI { 1: } { 1: } Z 2 + 1: { 1:} { 1: } GCI Peudo-Update Z 1 + 1: Peudo-Update Peudo-Predcton Z 1 + 1: + Z 2 + 1: JDM algorthm 1 { Z1:, Z 1: } 1 X } { Z1:, Z 1: } 2 X } + + IDMalgorthm Peudo-Update Peudo-Predcton Peudo-Update GCI 1 { Z1:, Z 1:, } 1 X } 2 { Z1:, Z 1:, } 2 X } ĉ 1 ĉ 2 ( cˆ,cˆ 1 2) Fg. 1. A comparon between JDM algorthm and IDM algorthm wth two enor. Algorthm 3: IDM Procedure INPUT: π, π {X Z 1:},C OUTPUT: (ĉ 1,...,ĉ N) Mult-target Sample: π Ψ S = {X (1),...X (M) } Peudo-Predcton: for ter = +1 : + do π π for = 1 : N do for each c C do for each X (j) Ψ S do Generate PIMS: X (j) Z 1:(c,X (j) ) Run Vo-Vo Flter Recuron: flter( π {X Z 1: }, Z 1:(c,X (j) )) π {X Z 1:, Z 1:(c,X (j) )} Compute Each Reward: D CS( π, π {X Z 1:, Z 1:(c,X (j) )}) R (j) (c ) EAP(R (j) (c )) R (c ) Decon Mang on Each Senor: arg max c C (R (c )) ĉ IV. SIMULATION RESULTS AND DISCUSSION In th ecton, the two propoed mult-enor control approache are appled to the problem of mult-target tracng wth two enor wth LSR. Wth both method, local flter are Vo-Vo flter, the fuon method choen a CMδ-GLMB and fuon weght of each enor ω 1,ω 2 are both choen a.5. The nematc target tate a vector of planar poton and velocty x = [t x, ṫ x, t y, ṫ y, ] T and the ngle-target tate pace model lnear Gauan accordng to tranton denty f 1 (x x 1 ) = N(x,F x 1,Q ) wth parameter F = [ ] I2 I 2,Q 2 I = σv [ 4 4 I I I 2 2 I 2 where I n and n denote the n n dentty and zero matrce repectvely, = 1 the amplng perod, σ v = 5m/ 2 the tandard devaton of the proce noe. In the context of mult-enor control, we conder the followng enor model that the meaurement a well a the detecton probablty a functon of dtance between target and enor tate. The enor meaurement are noy vector of polar poton of the form z = [ arctan( t y, y, t x, x, ) (tx, x, ) 2 +(t y, y, ) 2 ]+w (x,u ) where u = [ x, y, ] denote enor poton. w (x,u ) N( ;,R ) the meaurement noe wth covarance R = dag(σ 2 θ,σ2 r) n whch the cale of range and bearng noe are σ r = σ +η r x u 2 and σ θ = θ +η θ x u, the parameter σ = 1m, η r = m 1, θ = π/18rad ]
6 6 Y Coordnate (m) and η θ = m 1. The probablty of target detecton n each enor ndepent and of the form P D (x,u ) = N( x u ;,σ D ) N(;,σ D ) where σ D = 1m control the rate at whch the detecton probablty drop off a the range ncreae. Moreover, the urvval probablty P S, =.98, the number of clutter report n each can Poon dtrbuted wth λ c = 25. Each clutter report ampled unformly over the whole urvellance regon. The enor platform move wth contant velocty but tae coure change at pre-pecfed decon tme. The allowable control acton for each enor C = [ 18, 15,...,,...,15,18 ], the number of ample ued to compute the expected reward M = 4, the dealed meaurement are generated over a horzon length of = 5, wth amplng perod T = 2. The tet cenaro cont of 4 target, the enor eep tll durng frt 1 and mae frt decon at 1 o the econd decon at 2, thrd decon at 3, then reman on that coure untl the of the cenaro at tme 4. The regon and trac are hown n Fg. 2. Y Coordnate (m) urvve from 9 to 35 urvve from 1 to 4 Ground Truth enor1 enor2 urvve from 1 to 4 urvve from 16 to X Coordnate (m) Fg. 2. Target trajectore condered n the mulaton experment. The tart/ pont for each trajectory denoted, repectvely, by. The ndcate ntal enor poton Ground Truth enor1 enor2 decon tme X Coordnate (m) (a) Y Coordnate (m) Ground Truth enor1 enor2 decon tme X Coordnate (m) Fg. 3. (a) Trac output from a typcal run baed on IDM algorthm. (b) Trac output from a typcal run baed on JDM algorthm. Fg. 3 (a) and (b) how a ngle run to exhbt the typcal control behavour baed on IDM algorthm and JDM algorthm, repectvely. A t can be een, both control method (b) C S Dvergence C S Dvergence enor1 Coure change (deg) enor2 Coure change (deg) (a) C S Dvergence Senor 1 ( 6,3) (b) 6 Senor Fg. 4. (a) Reward curve at the tme of the econd decon (2) baed on IDM algorthm. (b) Reward curve at the tme of the econd decon (2) baed on JDM algorthm. OSPA (m)(c=1,p=2) random acton IDM algorthm JDM algorthm Tme Step Fg. 5. Comparon of OSPA error returned by randomed control acton, IDM algorthm and JDM algorthm. The plotted reult are the average of 1 Monte Carlo run. can mae proper decon that enor move cloe to the target. To be more pecfc, we denote the control acton choen by enor 1 and enor 2 by a vector (θ 1,θ 2 ), at the frt decon tme 1, two control method mae ame decon ( 3,3 ), at the econd decon tme 2, the IDM algorthm tae ( 3, ) whle the JDM algorthm tae ( 6,3 ). Fg. 4 (a) and (b) how the CS dvergence at the econd decon (2) of IDM algorthm and JDM algorthm, repectvely. Thee reult mean that compared wth the IDM algorthm, each enor controlled by JDM algorthm not greedy to oberve all target, but rather a vew of the whole pcture to mae the amount of nformaton content of fued denty larger. Fg. 5 how the comparon of OSPA error averaged over 1 Monte Carlo run among randomed control acton, IDM algorthm and JDM algorthm. A t hown, both control method can acheve better performance than randomed control trategy and the JDM algorthm preferable. Moreover, when the tuaton more complex uch a much more target or enor, the performance dfference between JDM algorthm and IDM algorthm wll ncreae and the randomed control trategy may collape. V. CONCLUSION In th paper, we addre the problem of mult-enor control for mult-target tracng va labelled random fnte et (RFS)
7 7 n the enor networ ytem. Wth the GCI fuon, two novel mult-enor control approache ung CS dvergence are preented, referred to JDM and IDM algorthm, repectvely. Smulaton reult verfy both the control approache perform well n mult-target tracng, the IDM method ha maller amount of computaton whle the JDM method mae decon from holtc pont of vew, and hence acheve better performance. ACKNOWLEDGMENT Th wor wa upported by the Natonal Natural Scence Foundaton of Chna under Grant , the Chnee Potdoctoral Scence Foundaton under Grant 214M REFERENCES [1] P. Ögren, E. Forell, and N. E. Leonard, Cooperatve control of moble enor networ: Adaptve gradent clmbng n a dtrbuted envronment, IEEE Tran. Autom. Control, vol. 49, no. 8, pp , 24. [2] V. Krhnamurthy, Algorthm for optmal chedulng and management of hdden marov model enor, IEEE Tran. Sgnal Proce., vol. 5, no. 6, pp , 22. [3] R. P. Mahler, Stattcal multource-multtarget nformaton fuon. Artech oue, Inc., 27. [4] B.-T. Vo and B.-N. Vo, Labeled random fnte et and mult-object conjugate pror, IEEE Tran. Sgnal Proce., vol. 61, no. 13, pp , 213. [5] B.-N. Vo, B.-T. Vo, and D. Phung, Labeled random fnte et and the baye mult-target tracng flter, IEEE Tran. Sgnal Proce., vol. 62, no. 24, pp , 214. [6] M. Beard, B.-T. Vo, B.-N. Vo, and S. Arulampalam, Vod probablte and cauchy-chwarz dvergence for generalzed labeled mult-bernoull model, arxv preprnt arxv: , 215. [7] B.-N. Vo and W.-K. Ma, The gauan mxture probablty hypothe denty flter, IEEE Tran. Sgnal Proce., vol. 54, no. 11, pp , 26. [8] B.-T. Vo, B.-N. Vo, and A. Canton, Analytc mplementaton of the cardnalzed probablty hypothe denty flter, IEEE Tran. Sgnal Proce., vol. 55, no. 7, pp , 27. [9] B.-T. Vo, B.-N. Vo, and A. Canton, The cardnalty balanced multtarget mult-bernoull flter and t mplementaton, IEEE Tran. Sgnal Proce., vol. 57, no. 2, pp , 29. [1]. G. oang and B. T. Vo, Senor management for mult-target tracng va mult-bernoull flterng, Automatca, vol. 5, no. 4, pp , 214. [11] A. K. Gotar, R. oennezhad, and A. Bab-adahar, Mult-bernoull enor control for mult-target tracng, n Intellgent Senor, Senor Networ and Informaton Proceng, 213 IEEE Eghth Internatonal Conference on, pp , IEEE, 213. [12] A. K. Gotar, R. oennezhad, and A. Bab-adahar, Robut multbernoull enor electon for mult-target tracng n enor networ, IEEE Sgnal Proceng Letter, vol. 2, no. 12, pp , 213. [13] A. K. Gotar, R. oennezhad, and A. Bab-adahar, Mult-bernoull enor control va mnmzaton of expected etmaton error, IEEE Tran. Aerop. Electron. Syt., vol. 51, no. 3, pp , 215. [14] A. K. Gotar, R. oennezhad, A. Bab-adahar, and F. Pap, Opabaed enor control, n Control, Automaton and Informaton Scence (ICCAIS), 215 Internatonal Conference on, pp , IEEE, 215. [15] K. Katella, Dcrmnaton gan to optmze detecton and clafcaton, Sytem, Man and Cybernetc, Part A: Sytem and uman, IEEE Tranacton on, vol. 27, no. 1, pp , [16] J. M. Aughenbaugh and B. R. La Cour, Metrc electon for nformaton theoretc enor management, n Informaton Fuon, 28 11th Internatonal Conference on, pp. 1 8, IEEE, 28. [17] C. Kreucher, A. O. ero III, and K. Katella, A comparon of ta drven and nformaton drven enor management for target tracng, n Decon and Control, 25 and 25 European Control Conference. CDC-ECC 5. 44th IEEE Conference on, pp , IEEE, 25. [18] B. Rtc and B.-N. Vo, Senor control for mult-object tate-pace etmaton ung random fnte et, Automatca, vol. 46, no. 11, pp , 21. [19] B. Rtc, B.-N. Vo, and D. Clar, A note on the reward functon for phd flter wth enor control, IEEE Tran. Aerop. Electron. Syt., vol. 47, no. 2, pp , 211. [2]. G. oang, B.-N. Vo, B.-T. Vo, and R. Mahler, The cauchy chwarz dvergence for poon pont procee, IEEE Tran. Inf. Theory, vol. 61, no. 8, pp , 215. [21] M. Beard, B.-T. Vo, B.-N. Vo, and S. Arulampalam, Senor control for mult-target tracng ung cauchy-chwarz dvergence, n Informaton Fuon (Fuon), th Internatonal Conference on, pp , IEEE, 215. [22] R. P. Mahler, Optmal/robut dtrbuted data fuon: a unfed approach, n AeroSene 2, pp , Internatonal Socety for Optc and Photonc, 2. [23] C. Fantacc, B.-N. Vo, B.-T. Vo, G. Batttell, and L. Chc, Conenu labeled random fnte et flterng for dtrbuted mult-object tracng, arxv preprnt arxv: , 215. [24] B. Wang, W. Y, S. L, M. R. Morelande, L. Kong, and X. Yang, Dtrbuted mult-target tracng va generalzed mult-bernoull random fnte et, n Informaton Fuon (Fuon), th Internatonal Conference on, pp , IEEE, 215. [25] R. P. Mahler, Global poteror dente for enor management, n Aeropace/Defene Senng and Control, pp , Internatonal Socety for Optc and Photonc, [26] R. Mahler, Multtarget enor management of dpered moble enor, Theory and Algorthm for Cooperatve Sytem, Kluwer, Sprnger, 25. [27] M. R. Morelande, Jont data aocaton ung mportance amplng, n Informaton Fuon, 29. FUSION 9. 12th Internatonal Conference on, pp , IEEE, 29.
Additional File 1 - Detailed explanation of the expression level CPD
Addtonal Fle - Detaled explanaton of the expreon level CPD A mentoned n the man text, the man CPD for the uterng model cont of two ndvdual factor: P( level gen P( level gen P ( level gen 2 (.).. CPD factor
More informationA Kernel Particle Filter Algorithm for Joint Tracking and Classification
A Kernel Partcle Flter Algorthm for Jont Tracng and Clafcaton Yunfe Guo Donglang Peng Inttute of Informaton and Control Automaton School Hangzhou Danz Unverty Hangzhou Chna gyf@hdueducn Huaje Chen Ane
More informationSpecification -- Assumptions of the Simple Classical Linear Regression Model (CLRM) 1. Introduction
ECONOMICS 35* -- NOTE ECON 35* -- NOTE Specfcaton -- Aumpton of the Smple Clacal Lnear Regreon Model (CLRM). Introducton CLRM tand for the Clacal Lnear Regreon Model. The CLRM alo known a the tandard lnear
More informationStart Point and Trajectory Analysis for the Minimal Time System Design Algorithm
Start Pont and Trajectory Analy for the Mnmal Tme Sytem Degn Algorthm ALEXANDER ZEMLIAK, PEDRO MIRANDA Department of Phyc and Mathematc Puebla Autonomou Unverty Av San Claudo /n, Puebla, 757 MEXICO Abtract:
More informationImprovements on Waring s Problem
Improvement on Warng Problem L An-Png Bejng, PR Chna apl@nacom Abtract By a new recurve algorthm for the auxlary equaton, n th paper, we wll gve ome mprovement for Warng problem Keyword: Warng Problem,
More informationEstimation of Finite Population Total under PPS Sampling in Presence of Extra Auxiliary Information
Internatonal Journal of Stattc and Analy. ISSN 2248-9959 Volume 6, Number 1 (2016), pp. 9-16 Reearch Inda Publcaton http://www.rpublcaton.com Etmaton of Fnte Populaton Total under PPS Samplng n Preence
More informationThe multivariate Gaussian probability density function for random vector X (X 1,,X ) T. diagonal term of, denoted
Appendx Proof of heorem he multvarate Gauan probablty denty functon for random vector X (X,,X ) px exp / / x x mean and varance equal to the th dagonal term of, denoted he margnal dtrbuton of X Gauan wth
More informationTeam. Outline. Statistics and Art: Sampling, Response Error, Mixed Models, Missing Data, and Inference
Team Stattc and Art: Samplng, Repone Error, Mxed Model, Mng Data, and nference Ed Stanek Unverty of Maachuett- Amhert, USA 9/5/8 9/5/8 Outlne. Example: Doe-repone Model n Toxcology. ow to Predct Realzed
More informationChapter 11. Supplemental Text Material. The method of steepest ascent can be derived as follows. Suppose that we have fit a firstorder
S-. The Method of Steepet cent Chapter. Supplemental Text Materal The method of teepet acent can be derved a follow. Suppoe that we have ft a frtorder model y = β + β x and we wh to ue th model to determne
More informationMethod Of Fundamental Solutions For Modeling Electromagnetic Wave Scattering Problems
Internatonal Workhop on MehFree Method 003 1 Method Of Fundamental Soluton For Modelng lectromagnetc Wave Scatterng Problem Der-Lang Young (1) and Jhh-We Ruan (1) Abtract: In th paper we attempt to contruct
More informationHarmonic oscillator approximation
armonc ocllator approxmaton armonc ocllator approxmaton Euaton to be olved We are fndng a mnmum of the functon under the retrcton where W P, P,..., P, Q, Q,..., Q P, P,..., P, Q, Q,..., Q lnwgner functon
More informationConfidence intervals for the difference and the ratio of Lognormal means with bounded parameters
Songklanakarn J. Sc. Technol. 37 () 3-40 Mar.-Apr. 05 http://www.jt.pu.ac.th Orgnal Artcle Confdence nterval for the dfference and the rato of Lognormal mean wth bounded parameter Sa-aat Nwtpong* Department
More informationChapter 6 The Effect of the GPS Systematic Errors on Deformation Parameters
Chapter 6 The Effect of the GPS Sytematc Error on Deformaton Parameter 6.. General Beutler et al., (988) dd the frt comprehenve tudy on the GPS ytematc error. Baed on a geometrc approach and aumng a unform
More informationSmall signal analysis
Small gnal analy. ntroducton Let u conder the crcut hown n Fg., where the nonlnear retor decrbed by the equaton g v havng graphcal repreentaton hown n Fg.. ( G (t G v(t v Fg. Fg. a D current ource wherea
More informationImprovements on Waring s Problem
Imrovement on Warng Problem L An-Png Bejng 85, PR Chna al@nacom Abtract By a new recurve algorthm for the auxlary equaton, n th aer, we wll gve ome mrovement for Warng roblem Keyword: Warng Problem, Hardy-Lttlewood
More informationarxiv: v1 [cs.gt] 15 Jan 2019
Model and algorthm for tme-content rk-aware Markov game Wenje Huang, Pham Vet Ha and Wllam B. Hakell January 16, 2019 arxv:1901.04882v1 [c.gt] 15 Jan 2019 Abtract In th paper, we propoe a model for non-cooperatve
More informationVerification of Selected Precision Parameters of the Trimble S8 DR Plus Robotic Total Station
81 Verfcaton of Selected Precon Parameter of the Trmble S8 DR Plu Robotc Total Staton Sokol, Š., Bajtala, M. and Ježko, J. Slovak Unverty of Technology, Faculty of Cvl Engneerng, Radlnkého 11, 81368 Bratlava,
More informationScattering of two identical particles in the center-of. of-mass frame. (b)
Lecture # November 5 Scatterng of two dentcal partcle Relatvtc Quantum Mechanc: The Klen-Gordon equaton Interpretaton of the Klen-Gordon equaton The Drac equaton Drac repreentaton for the matrce α and
More informationStatistical Properties of the OLS Coefficient Estimators. 1. Introduction
ECOOMICS 35* -- OTE 4 ECO 35* -- OTE 4 Stattcal Properte of the OLS Coeffcent Etmator Introducton We derved n ote the OLS (Ordnary Leat Square etmator ˆβ j (j, of the regreon coeffcent βj (j, n the mple
More informationMULTIPLE REGRESSION ANALYSIS For the Case of Two Regressors
MULTIPLE REGRESSION ANALYSIS For the Cae of Two Regreor In the followng note, leat-quare etmaton developed for multple regreon problem wth two eplanator varable, here called regreor (uch a n the Fat Food
More informationTwo Approaches to Proving. Goldbach s Conjecture
Two Approache to Provng Goldbach Conecture By Bernard Farley Adved By Charle Parry May 3 rd 5 A Bref Introducton to Goldbach Conecture In 74 Goldbach made h mot famou contrbuton n mathematc wth the conecture
More informationOptimal inference of sameness Supporting information
Optmal nference of amene Supportng nformaton Content Decon rule of the optmal oberver.... Unequal relablte.... Equal relablte... 5 Repone probablte of the optmal oberver... 6. Equal relablte... 6. Unequal
More informationIntroduction. Modeling Data. Approach. Quality of Fit. Likelihood. Probabilistic Approach
Introducton Modelng Data Gven a et of obervaton, we wh to ft a mathematcal model Model deend on adutable arameter traght lne: m + c n Polnomal: a + a + a + L+ a n Choce of model deend uon roblem Aroach
More informationDeep Reinforcement Learning with Experience Replay Based on SARSA
Deep Renforcement Learnng wth Experence Replay Baed on SARSA Dongbn Zhao, Hatao Wang, Kun Shao and Yuanheng Zhu Key Laboratory of Management and Control for Complex Sytem Inttute of Automaton Chnee Academy
More informationThe Essential Dynamics Algorithm: Essential Results
@ MIT maachuett nttute of technology artfcal ntellgence laboratory The Eental Dynamc Algorthm: Eental Reult Martn C. Martn AI Memo 003-014 May 003 003 maachuett nttute of technology, cambrdge, ma 0139
More informationSolution Methods for Time-indexed MIP Models for Chemical Production Scheduling
Ian Davd Lockhart Bogle and Mchael Farweather (Edtor), Proceedng of the 22nd European Sympoum on Computer Aded Proce Engneerng, 17-2 June 212, London. 212 Elever B.V. All rght reerved. Soluton Method for
More informationRoot Locus Techniques
Root Locu Technque ELEC 32 Cloed-Loop Control The control nput u t ynthezed baed on the a pror knowledge of the ytem plant, the reference nput r t, and the error gnal, e t The control ytem meaure the output,
More information2.3 Least-Square regressions
.3 Leat-Square regreon Eample.10 How do chldren grow? The pattern of growth vare from chld to chld, o we can bet undertandng the general pattern b followng the average heght of a number of chldren. Here
More informationOPTIMAL COMPUTING BUDGET ALLOCATION FOR MULTI-OBJECTIVE SIMULATION MODELS. David Goldsman
Proceedng of the 004 Wnter Smulaton Conference R.G. Ingall, M. D. Roett, J. S. Smth, and B. A. Peter, ed. OPTIMAL COMPUTING BUDGET ALLOCATION FOR MULTI-OBJECTIVE SIMULATION MODELS Loo Hay Lee Ek Peng Chew
More informationIntroduction to Interfacial Segregation. Xiaozhe Zhang 10/02/2015
Introducton to Interfacal Segregaton Xaozhe Zhang 10/02/2015 Interfacal egregaton Segregaton n materal refer to the enrchment of a materal conttuent at a free urface or an nternal nterface of a materal.
More informationBatch RL Via Least Squares Policy Iteration
Batch RL Va Leat Square Polcy Iteraton Alan Fern * Baed n part on lde by Ronald Parr Overvew Motvaton LSPI Dervaton from LSTD Expermental reult Onlne veru Batch RL Onlne RL: ntegrate data collecton and
More informationGeneralized Linear Methods
Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set
More informationTracking with Kalman Filter
Trackng wth Kalman Flter Scott T. Acton Vrgna Image and Vdeo Analyss (VIVA), Charles L. Brown Department of Electrcal and Computer Engneerng Department of Bomedcal Engneerng Unversty of Vrgna, Charlottesvlle,
More informationAP Statistics Ch 3 Examining Relationships
Introducton To tud relatonhp between varable, we mut meaure the varable on the ame group of ndvdual. If we thnk a varable ma eplan or even caue change n another varable, then the eplanator varable and
More informationExtended Prigogine Theorem: Method for Universal Characterization of Complex System Evolution
Extended Prgogne Theorem: Method for Unveral Characterzaton of Complex Sytem Evoluton Sergey amenhchkov* Mocow State Unverty of M.V. Lomonoov, Phycal department, Rua, Mocow, Lennke Gory, 1/, 119991 Publhed
More informationDiscrete Simultaneous Perturbation Stochastic Approximation on Loss Function with Noisy Measurements
0 Amercan Control Conference on O'Farrell Street San Francco CA USA June 9 - July 0 0 Dcrete Smultaneou Perturbaton Stochatc Approxmaton on Lo Functon wth Noy Meaurement Q Wang and Jame C Spall Abtract
More informationJoint Source Coding and Higher-Dimension Modulation
Jont Codng and Hgher-Dmenon Modulaton Tze C. Wong and Huck M. Kwon Electrcal Engneerng and Computer Scence Wchta State Unvert, Wchta, Kana 676, USA {tcwong; huck.kwon}@wchta.edu Abtract Th paper propoe
More informationA Quadratic Constraint Total Least-squares Algorithm for Hyperbolic Location
I. J. Communcaton, Network and Sytem Scence, 8,, 5-6 Publhed Onlne May 8 n ScRe (http://www.srpublhng.org/journal/jcn/). A Quadratc Contrant otal Leat-quare Algorthm for Hyperbolc Locaton Ka YANG, Janpng
More informationMODELLING OF TRANSIENT HEAT TRANSPORT IN TWO-LAYERED CRYSTALLINE SOLID FILMS USING THE INTERVAL LATTICE BOLTZMANN METHOD
Journal o Appled Mathematc and Computatonal Mechanc 7, 6(4), 57-65 www.amcm.pcz.pl p-issn 99-9965 DOI:.75/jamcm.7.4.6 e-issn 353-588 MODELLING OF TRANSIENT HEAT TRANSPORT IN TWO-LAYERED CRYSTALLINE SOLID
More informationParametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010
Parametrc fractonal mputaton for mssng data analyss Jae Kwang Km Survey Workng Group Semnar March 29, 2010 1 Outlne Introducton Proposed method Fractonal mputaton Approxmaton Varance estmaton Multple mputaton
More informationQuick Visit to Bernoulli Land
Although we have een the Bernoull equaton and een t derved before, th next note how t dervaton for an uncopreble & nvcd flow. The dervaton follow that of Kuethe &Chow ot cloely (I lke t better than Anderon).
More informationKEY POINTS FOR NUMERICAL SIMULATION OF INCLINATION OF BUILDINGS ON LIQUEFIABLE SOIL LAYERS
KY POINTS FOR NUMRICAL SIMULATION OF INCLINATION OF BUILDINGS ON LIQUFIABL SOIL LAYRS Jn Xu 1, Xaomng Yuan, Jany Zhang 3,Fanchao Meng 1 1 Student, Dept. of Geotechncal ngneerng, Inttute of ngneerng Mechanc,
More informationAdaptive Centering with Random Effects in Studies of Time-Varying Treatments. by Stephen W. Raudenbush University of Chicago.
Adaptve Centerng wth Random Effect n Stde of Tme-Varyng Treatment by Stephen W. Radenbh Unverty of Chcago Abtract Of wdepread nteret n ocal cence are obervatonal tde n whch entte (peron chool tate contre
More informationLogistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI
Logstc Regresson CAP 561: achne Learnng Instructor: Guo-Jun QI Bayes Classfer: A Generatve model odel the posteror dstrbuton P(Y X) Estmate class-condtonal dstrbuton P(X Y) for each Y Estmate pror dstrbuton
More informationMarkov Chain Monte Carlo Lecture 6
where (x 1,..., x N ) X N, N s called the populaton sze, f(x) f (x) for at least one {1, 2,..., N}, and those dfferent from f(x) are called the tral dstrbutons n terms of mportance samplng. Dfferent ways
More informationAPPROXIMATE FUZZY REASONING BASED ON INTERPOLATION IN THE VAGUE ENVIRONMENT OF THE FUZZY RULEBASE AS A PRACTICAL ALTERNATIVE OF THE CLASSICAL CRI
Kovác, Sz., Kóczy, L.T.: Approxmate Fuzzy Reaonng Baed on Interpolaton n the Vague Envronment of the Fuzzy Rulebae a a Practcal Alternatve of the Clacal CRI, Proceedng of the 7 th Internatonal Fuzzy Sytem
More informationFusion of Possible Biased Local Estimates in Sensor Network Based on Sensor Selection
Fuon of Poble Baed Local Etmate n Senor etwork Baed on Senor Selecton Hongyan hu Shuo Chen Chongzhao Han Yan Ln Int.of Integrated Automaton MOE KLIS Lab School of Electronc & Informaton Engneerng, X an
More informationOutline. Communication. Bellman Ford Algorithm. Bellman Ford Example. Bellman Ford Shortest Path [1]
DYNAMIC SHORTEST PATH SEARCH AND SYNCHRONIZED TASK SWITCHING Jay Wagenpfel, Adran Trachte 2 Outlne Shortest Communcaton Path Searchng Bellmann Ford algorthm Algorthm for dynamc case Modfcatons to our algorthm
More information2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification
E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton
More informationA New Inverse Reliability Analysis Method Using MPP-Based Dimension Reduction Method (DRM)
roceedng of the ASME 007 Internatonal Degn Engneerng Techncal Conference & Computer and Informaton n Engneerng Conference IDETC/CIE 007 September 4-7, 007, La Vega, eada, USA DETC007-35098 A ew Inere Relablty
More informationVariable Structure Control ~ Basics
Varable Structure Control ~ Bac Harry G. Kwatny Department of Mechancal Engneerng & Mechanc Drexel Unverty Outlne A prelmnary example VS ytem, ldng mode, reachng Bac of dcontnuou ytem Example: underea
More informationLinear Approximation with Regularization and Moving Least Squares
Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...
More informationA Robust Method for Calculating the Correlation Coefficient
A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal
More informationDesign of Recursive Digital Filters IIR
Degn of Recurve Dgtal Flter IIR The outut from a recurve dgtal flter deend on one or more revou outut value, a well a on nut t nvolve feedbac. A recurve flter ha an nfnte mule reone (IIR). The mulve reone
More informationAn Improved multiple fractal algorithm
Advanced Scence and Technology Letters Vol.31 (MulGraB 213), pp.184-188 http://dx.do.org/1.1427/astl.213.31.41 An Improved multple fractal algorthm Yun Ln, Xaochu Xu, Jnfeng Pang College of Informaton
More information728. Mechanical and electrical elements in reduction of vibrations
78. Mechancal and electrcal element n reducton of vbraton Katarzyna BIAŁAS The Slean Unverty of Technology, Faculty of Mechancal Engneerng Inttute of Engneerng Procee Automaton and Integrated Manufacturng
More informationA Novel Approach for Testing Stability of 1-D Recursive Digital Filters Based on Lagrange Multipliers
Amercan Journal of Appled Scence 5 (5: 49-495, 8 ISSN 546-939 8 Scence Publcaton A Novel Approach for Tetng Stablty of -D Recurve Dgtal Flter Baed on Lagrange ultpler KRSanth, NGangatharan and Ponnavakko
More informationFor now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.
Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson
More informationOn the SO 2 Problem in Thermal Power Plants. 2.Two-steps chemical absorption modeling
Internatonal Journal of Engneerng Reearch ISSN:39-689)(onlne),347-53(prnt) Volume No4, Iue No, pp : 557-56 Oct 5 On the SO Problem n Thermal Power Plant Two-tep chemcal aborpton modelng hr Boyadjev, P
More informationForesighted Resource Reciprocation Strategies in P2P Networks
Foreghted Reource Recprocaton Stratege n PP Networ Hyunggon Par and Mhaela van der Schaar Electrcal Engneerng Department Unverty of Calforna Lo Angele (UCLA) Emal: {hgpar mhaela@ee.ucla.edu Abtract We
More informationDistributed Control for the Parallel DC Linked Modular Shunt Active Power Filters under Distorted Utility Voltage Condition
Dtrbted Control for the Parallel DC Lnked Modlar Shnt Actve Power Flter nder Dtorted Utlty Voltage Condton Reearch Stdent: Adl Salman Spervor: Dr. Malabka Ba School of Electrcal and Electronc Engneerng
More informationA PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS
HCMC Unversty of Pedagogy Thong Nguyen Huu et al. A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS Thong Nguyen Huu and Hao Tran Van Department of mathematcs-nformaton,
More informationIterative Methods for Searching Optimal Classifier Combination Function
htt://www.cub.buffalo.edu Iteratve Method for Searchng Otmal Clafer Combnaton Functon Sergey Tulyakov Chaohong Wu Venu Govndaraju Unverty at Buffalo Identfcaton ytem: Alce Bob htt://www.cub.buffalo.edu
More informationDEADLOCK INDEX ANALYSIS OF MULTI-LEVEL QUEUE SCHEDULING IN OPERATING SYSTEM USING DATA MODEL APPROACH
GESJ: Computer Scence and Telecommuncaton 2 No.(29 ISSN 2-232 DEADLOCK INDEX ANALYSIS OF MULTI-LEVEL QUEUE SCHEDULING IN OPERATING SYSTEM USING DATA MODEL APPROACH D. Shukla, Shweta Ojha 2 Deptt. of Mathematc
More informationWeighted Least-Squares Solutions for Energy-Based Collaborative Source Localization Using Acoustic Array
IJCSS Internatonal Journal of Computer Scence and etwork Securty, VOL.7 o., January 7 59 Weghted Leat-Square Soluton for nergy-baed Collaboratve Source Localzaton Ung Acoutc Array Kebo Deng and Zhong Lu
More informationThis is a repository copy of An iterative orthogonal forward regression algorithm.
Th a repotory copy of An teratve orthogonal forward regreon algorthm. Whte Roe Reearch Onlne URL for th paper: http://eprnt.whteroe.ac.uk/0735/ Veron: Accepted Veron Artcle: Guo, Y., Guo, L. Z., Bllng,
More informationThis appendix presents the derivations and proofs omitted from the main text.
Onlne Appendx A Appendx: Omtted Dervaton and Proof Th appendx preent the dervaton and proof omtted from the man text A Omtted dervaton n Secton Mot of the analy provded n the man text Here, we formally
More informationWind - Induced Vibration Control of Long - Span Bridges by Multiple Tuned Mass Dampers
Tamkang Journal of Scence and Engneerng, Vol. 3, o., pp. -3 (000) Wnd - Induced Vbraton Control of Long - Span Brdge by Multple Tuned Ma Damper Yuh-Y Ln, Ch-Mng Cheng and Davd Sun Department of Cvl Engneerng
More informationBatch Reinforcement Learning
Batch Renforcement Learnng Alan Fern * Baed n part on lde by Ronald Parr Overvew What batch renforcement learnng? Leat Square Polcy Iteraton Ftted Q-teraton Batch DQN Onlne veru Batch RL Onlne RL: ntegrate
More informationECE559VV Project Report
ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate
More informationLecture Notes on Linear Regression
Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume
More informationA NUMERICAL MODELING OF MAGNETIC FIELD PERTURBATED BY THE PRESENCE OF SCHIP S HULL
A NUMERCAL MODELNG OF MAGNETC FELD PERTURBATED BY THE PRESENCE OF SCHP S HULL M. Dennah* Z. Abd** * Laboratory Electromagnetc Sytem EMP BP b Ben-Aknoun 606 Alger Algera ** Electronc nttute USTHB Alger
More informationResonant FCS Predictive Control of Power Converter in Stationary Reference Frame
Preprnt of the 9th World Congre The Internatonal Federaton of Automatc Control Cape Town, South Afrca. Augut -9, Reonant FCS Predctve Control of Power Converter n Statonary Reference Frame Lupng Wang K
More informationDiscrete MRF Inference of Marginal Densities for Non-uniformly Discretized Variable Space
2013 IEEE Conference on Computer Von and Pattern Recognton Dcrete MRF Inference of Margnal Dente for Non-unformly Dcretzed Varable Space Maak Sato Takayuk Okatan Kochro Deguch Tohoku Unverty, Japan {mato,
More informationMACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression
11 MACHINE APPLIED MACHINE LEARNING LEARNING MACHINE LEARNING Gaussan Mture Regresson 22 MACHINE APPLIED MACHINE LEARNING LEARNING Bref summary of last week s lecture 33 MACHINE APPLIED MACHINE LEARNING
More informationModule 5. Cables and Arches. Version 2 CE IIT, Kharagpur
odule 5 Cable and Arche Veron CE IIT, Kharagpur Leon 33 Two-nged Arch Veron CE IIT, Kharagpur Intructonal Objectve: After readng th chapter the tudent wll be able to 1. Compute horzontal reacton n two-hnged
More informationCovariance Intersection Fusion Kalman Estimator for the Two-sensor Time-delayed System
Covarance Interecton Fuon Kalman Etmator for the Two-enor Tme-elaye Sytem Jnfang Lu Department of Automaton Helongang Unverty Harbn Chna englhlu@yahoo.com.cn Zl Deng Yuan Gao Department of Automaton Helongang
More informationModeling of Wave Behavior of Substrate Noise Coupling for Mixed-Signal IC Design
Modelng of Wave Behavor of Subtrate Noe Couplng for Mxed-Sgnal IC Degn Georgo Veron, Y-Chang Lu, and Robert W. Dutton Center for Integrated Sytem, Stanford Unverty, Stanford, CA 9435 yorgo@gloworm.tanford.edu
More informationPower law and dimension of the maximum value for belief distribution with the max Deng entropy
Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng
More informationGREY PREDICTIVE PROCESS CONTROL CHARTS
The 4th Internatonal Conference on Qualty Relablty Augut 9-th, 2005 Bejng, Chna GREY PREDICTIVE PROCESS CONTROL CHARTS RENKUAN GUO, TIM DUNNE Department of Stattcal Scence, Unverty of Cape Town, Prvate
More informationPROBABILITY-CONSISTENT SCENARIO EARTHQUAKE AND ITS APPLICATION IN ESTIMATION OF GROUND MOTIONS
PROBABILITY-COSISTET SCEARIO EARTHQUAKE AD ITS APPLICATIO I ESTIATIO OF GROUD OTIOS Q-feng LUO SUARY Th paper preent a new defnton of probablty-content cenaro earthquae PCSE and an evaluaton method of
More informationA New Virtual Indexing Method for Measuring Host Connection Degrees
A New Vrtual Indexng Method for Meaurng ot Connecton Degree Pnghu Wang, Xaohong Guan,, Webo Gong 3, and Don Towley 4 SKLMS Lab and MOE KLINNS Lab, X an Jaotong Unverty, X an, Chna Department of Automaton
More informationNo! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Survey Results. Class 15. Is the following possible?
Survey Reult Chapter 5-6 (where we are gong) % of Student 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Hour Spent on ChE 273 1-2 3-4 5-6 7-8 9-10 11+ Hour/Week 2008 2009 2010 2011 2012 2013 2014 2015 2017 F17
More informationPrimer on High-Order Moment Estimators
Prmer on Hgh-Order Moment Estmators Ton M. Whted July 2007 The Errors-n-Varables Model We wll start wth the classcal EIV for one msmeasured regressor. The general case s n Erckson and Whted Econometrc
More informationCHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE
CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng
More informationBias-corrected nonparametric correlograms for geostatistical radar-raingauge combination
ERAD 00 - THE SIXTH EUROPEA COFERECE O RADAR I METEOROLOGY AD HYDROLOGY Ba-corrected nonparametrc correlogram for geotattcal radar-rangauge combnaton Renhard Schemann, Rebekka Erdn, Marco Wll, Chrtoph
More informationComputation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models
Computaton of Hgher Order Moments from Two Multnomal Overdsperson Lkelhood Models BY J. T. NEWCOMER, N. K. NEERCHAL Department of Mathematcs and Statstcs, Unversty of Maryland, Baltmore County, Baltmore,
More informationFinite Mixture Models and Expectation Maximization. Most slides are from: Dr. Mario Figueiredo, Dr. Anil Jain and Dr. Rong Jin
Fnte Mxture Models and Expectaton Maxmzaton Most sldes are from: Dr. Maro Fgueredo, Dr. Anl Jan and Dr. Rong Jn Recall: The Supervsed Learnng Problem Gven a set of n samples X {(x, y )},,,n Chapter 3 of
More informationPhysics 111. CQ1: springs. con t. Aristocrat at a fixed angle. Wednesday, 8-9 pm in NSC 118/119 Sunday, 6:30-8 pm in CCLIR 468.
c Announcement day, ober 8, 004 Ch 8: Ch 10: Work done by orce at an angle Power Rotatonal Knematc angular dplacement angular velocty angular acceleraton Wedneday, 8-9 pm n NSC 118/119 Sunday, 6:30-8 pm
More informationParameter Estimation for Dynamic System using Unscented Kalman filter
Parameter Estmaton for Dynamc System usng Unscented Kalman flter Jhoon Seung 1,a, Amr Atya F. 2,b, Alexander G.Parlos 3,c, and Klto Chong 1,4,d* 1 Dvson of Electroncs Engneerng, Chonbuk Natonal Unversty,
More informationResource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud
Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal
More informationThe Study of Teaching-learning-based Optimization Algorithm
Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute
More informationJoint Energy-Efficient Cooperative Spectrum Sensing and Power Allocation in Cognitive Machine-to-Machine Communications
Jont Energy-Effcent Cooperatve Spectrum Senng and Power Allocaton n Cogntve Machne-to-Machne Communcaton Ha Ngoc Pham, Yan Zhang, Tor Skee, Paal E. Engeltad, Frank Elaen Department of Informatc, Unverty
More informationAPPLICATIONS OF RELIABILITY ANALYSIS TO POWER ELECTRONICS SYSTEMS
APPLICATIONS OF RELIABILITY ANALYSIS TO POWER ELECTRONICS SYSTEMS Chanan Sngh, Fellow IEEE Praad Enjet, Fellow IEEE Department o Electrcal Engneerng Texa A&M Unverty College Staton, Texa USA Joydeep Mtra,
More informationQuantifying Uncertainty
Partcle Flters Quantfyng Uncertanty Sa Ravela M. I. T Last Updated: Sprng 2013 1 Quantfyng Uncertanty Partcle Flters Partcle Flters Appled to Sequental flterng problems Can also be appled to smoothng problems
More informationA LINEAR PROGRAM TO COMPARE MULTIPLE GROSS CREDIT LOSS FORECASTS. Dr. Derald E. Wentzien, Wesley College, (302) ,
A LINEAR PROGRAM TO COMPARE MULTIPLE GROSS CREDIT LOSS FORECASTS Dr. Derald E. Wentzen, Wesley College, (302) 736-2574, wentzde@wesley.edu ABSTRACT A lnear programmng model s developed and used to compare
More informationNot at Steady State! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Yes! Class 15. Is the following possible?
Chapter 5-6 (where we are gong) Ideal gae and lqud (today) Dente Partal preure Non-deal gae (next tme) Eqn. of tate Reduced preure and temperature Compreblty chart (z) Vapor-lqud ytem (Ch. 6) Vapor preure
More information4DVAR, according to the name, is a four-dimensional variational method.
4D-Varatonal Data Assmlaton (4D-Var) 4DVAR, accordng to the name, s a four-dmensonal varatonal method. 4D-Var s actually a drect generalzaton of 3D-Var to handle observatons that are dstrbuted n tme. The
More informationValid Inequalities Based on Demand Propagation for Chemical Production Scheduling MIP Models
Vald Inequalte Baed on Demand ropagaton for Chemcal roducton Schedulng MI Model Sara Velez, Arul Sundaramoorthy, And Chrto Maravela 1 Department of Chemcal and Bologcal Engneerng Unverty of Wconn Madon
More informationAlpha Risk of Taguchi Method with L 18 Array for NTB Type QCH by Simulation
Proceedng of the World Congre on Engneerng 00 Vol II WCE 00, July -, 00, London, U.K. Alpha Rk of Taguch Method wth L Array for NTB Type QCH by Smulaton A. Al-Refae and M.H. L Abtract Taguch method a wdely
More information