Example: dental hygienist

The Unscented Kalman Filter for Nonlinear Estimation

Second-order ,therelationshipbetweentheKalmanFilter(K F)andRecursiveLeastSquares(RLS)isgivenin [3].TheuseoftheEKFfortrainingneuralnetwo rkshasbeendevelopedbySinghalandWu[9]andP uskoriousandFeldkamp[8].DualEstimationAs pecialcaseofmachinelearningariseswhenthe inputisunobserved, ,weagainconsideradiscrete-timenonlineard ynamicsystem,(6)(7) ,weintroducetheUnscentedKalmanFilter(UKF ) ,inSection4, ,are-cursiveestimationforcanbeexpressedi ntheform(see[6]),predictionofpredictiono f(8)Thisrecursionprovidestheoptimalminim ummean-squarederror(MMSE)estimateforassu mingthepriorestimateandcurrentobservatio nareGaussianRandomVari-ables(GRV).

inal noisytime-series hasa 3dBSNR). Thesuperior perfor-manceof the UKF is clearly visible. 200 210 220 230 240 250 260 270 280 290 300 −5 0 5 k x(k) Estimation of Mackey−Glass time series : EKF clean noisy EKF 200 210 220 230 240 250 260 270 280 290 300 −5 0 5 k x(k) Estimation of Mackey−Glass time series : UKF clean noisy UKF

Fullscreen Download

Tags:

  Perfor

Information

Domain:

Source:

Link to this page:

Please notify us if you found a problem with this document:

Other abuse

Advertisement

Transcription of The Unscented Kalman Filter for Nonlinear Estimation

1 Second-order ,therelationshipbetweentheKalmanFilter(K F)andRecursiveLeastSquares(RLS)isgivenin [3].TheuseoftheEKFfortrainingneuralnetwo rkshasbeendevelopedbySinghalandWu[9]andP uskoriousandFeldkamp[8].DualEstimationAs pecialcaseofmachinelearningariseswhenthe inputisunobserved, ,weagainconsideradiscrete-timenonlineard ynamicsystem,(6)(7) ,weintroducetheUnscentedKalmanFilter(UKF ) ,inSection4, ,are-cursiveestimationforcanbeexpressedi ntheform(see[6]),predictionofpredictiono f(8)Thisrecursionprovidestheoptimalminim ummean-squarederror(MMSE)estimateforassu mingthepriorestimateandcurrentobservatio nareGaussianRandomVari-ables(GRV).

2 (9)(10)(11)wheretheoptimalpredictionofis writtenas,andcorrespondstotheexpectation ofanonlinearfunctionoftherandomvariables and(similarinterpretationfortheoptimalpr ediction).Theoptimalgaintermisexpresseda safunctionofposteriorcovariancematrices( with). , ,however,theEKFapproxi-matestheoptimalte rmsas:(12)(13)(14)wherepredictionsareapp roximatedassimplythefunctionofthepriorme anvalueforestimates(noexpectationtaken)1 Thecovariancearedeterminedbylinearizingt hedynamicequations(), ,intheEKFthestatedistributionisapproxima tedbyaGRVwhichisthenpropagatedanalytical lythroughthe first-order [6] ,theEKFcanbeviewedasproviding first-order ,however,canintro-ducelargeerrorsinthetr ueposteriormeanandcovarianceofthetransfo rmed(Gaussian)randomvariable, flaws , ,andwhenpropagatedthroughthetruenon-line arsystem,capturestheposteriormeanandcova rianceaccuratelytothe3rdorder(Taylorse-r iesexpansion)

3 ,1 Thenoisemeansaredenotedbyand, second-order versionsoftheEKFexist, (UT)isamethodforcal-culatingthestatistic sofarandomvariablewhichundergoesanonline artransformation[5].Considerpropagatinga ran-domvariable(dimension)throughanonlin earfunction,. ,weformamatrixofsigmavectors(withcorresp ondingweights),accord-ingtothefollowing: (15) ( ,1e-3).isasecondaryscalingparameterwhich isusuallysetto0,andisusedtoincorporatepr iorknowledgeofthedistributionof(forGauss iandistributions,isoptimal). ,(16)andthemeanandcovarianceforareapprox imatedus-ingaweightedsamplemeanandcovari anceoftheposteriorsigmapoints,(17)(18)No tethatthismethoddifferssubstantiallyfrom general sam-pling methods( ,Monte-Carlomethodssuchasparticlefilters [1])whichrequireordersofmagnitudemoresam plepointsinanattempttopropagateanaccurat e(possiblynon-Gaussian) ,approximationsareaccuratetoatleastthese cond-order,withtheaccuracyofthirdandhigh erordermomentsdeterminedbythechoiceofand (See[4]foradetaileddiscussionoftheUT).

4 AsimpleexampleisshowninFigure1fora2-dime nsionalsystem:theleftplotshowsthetruemea nandcovariancepropagationusingMonte-Carl osampling;thecenterplotsActual (sampling)Linearized (EKF)UTsigma pointstrue meanUT mean and covarianceweighted sample meanmeanUT covariancecovariancetrue covariancetransformedsigma pointsFigure1 )actual,b)first-orderlinearization(EKF), c) ;therightplotsshowtheperformanceoftheUT( noteonly5sigmapointsarerequired). (UKF)isastraightfor-wardextensionoftheUT totherecursiveestimationinEqua-tion8,whe rethestateRVisredefinedastheconcatenatio noftheoriginalstateandnoisevariables.

5 TheUTsigmapointselectionscheme(Equation1 5)isap-pliedtothisnewaugmentedstateRVtoc alculatethecorre-spondingsigmamatrix,. , ,andhasbeenappliedinnonlinearcontrolappl ica-tionsrequiringfull-statefeedback[5]. Intheseapplications,thedynamicmodelrepre sentsaphysicallybasedparamet-ricmodel, ,weextendtheuseoftheUKFtoabroaderclassof nonlinearestimationproblems, , , :For,Calculatesigmapoints:Timeupdate:Mea surementupdateequations:where,,,=composi tescalingparameter,=dimensionofaugmented state,=processnoisecov.,=measurementnois ecov.,= :UnscentedKalmanFilter(UKF) (19)wherethemodel(parameterizedbyw) , :..(20)Intheestimationproblem,thenoisy-t imeseriesistheonlyobservedinputtoeithert heEKForUKFalgorithms(bothutilizetheknown neuralnetworkmodel).

6 (theorig-inalnoisytime-serieshasa3dBSNR) . 505kx(k) Estimation of Mackey Glass time series : EKFcleannoisyEKF 200210220230240250260270280290300 505kx(k) Estimation of Mackey Glass time series : UKFcleannoisyUKF MSEE stimation Error : EKF vs UKF on Mackey GlassEKFUKFF igure2 (seeEquation7).Asexpressedearlier, [2]waspresentedin[13].Asinthepriorstate- estimationexample, [11], ,thestate-spacerepresentationfortheweigh tsisgivenby(21)(22) ,thecurrentestimateoftheweightsisusedint hesignal- Filter , , , [7],thesignal-stateandweightvectorsareco ncatenatedintoasingle,jointstatevector:. Estimationisdonerecursivelybywrit-ingthe state-spaceequationsforthejointstateas:( 23)(24) , (SNR3dB).

7 Thesecondtimeseries(alsochaotic)comesfro manautoregressiveneuralnetworkwithrandom weightsdrivenbyGaussianprocessnoiseandal so3isusuallysettoasmallconstantwhichcanb erelatedtothetime-constantforRLSweightde cay[3].Foradatalengthof1000, (SNR3dB). , MSEC haotic AR neural networkDual UKF Dual EKF Joint UKFJ oint MSEM ackey Glass chaotic time seriesDual EKF Dual UKF Joint EKFJ oint UKF200210220230240250260270280290300 505kx(k) Estimation of Mackey Glass time series : Dual UKFclean noisy Dual UKFF igure3 , (isthenumberofweights).Theadvantageofthe UKFovertheEKFinthiscaseisalsonotasobviou s, ,aspointedoutearlier, , ,however, , (averagedover100experimentswithdifferent initialweights) , 210 1 Mackay Robot Arm : Learning curvesepochmean MSEUKF EKF 05101520253010 1100 Ikeda chaotic time series.

8 Learning curvesepochmean MSEUKF EKF Figure4 )Mackay-Robot-Arm,2-12-2 MLP,b)Ikedatimeseries, ,includingstate- Estimation ,duales-timati on, ,exten-sionstobatchlearningandnon-MSEcos tfunctions,aswellasapplicationtootherneu ralandnon-neural( ,paramet-ric) ,wearealsoexploringtheuseoftheUKFasameth odtoimproveParticleFilters[10],aswellasa nextensionoftheUKFitselfthatavoidsthelin earupdateassumptionbyusingadirectBayesia nupdate[12]. [1] , , , , ,UniversityofCambridge,Nov1998.[2] , , ,B39:1 38,1977.[3] ,Inc,3edition,1996.[4] ,February2000.[5] ,SimulationandControls.,1997.[6] ,Inc.,NewYork,1986.

9 [7] ,pages197 200,1990.[8] ,volume1,pages771 777,1991.[9] ,pages133 140,SanMateo,CA, [10] , , , , ,UniversityofCambridge, [11] ,1997.[12] ,CSLU,OregonGraduateInstituteofSciencean dTech-nology, ( ).[13] , , , , uller,editors,AdvancesinNeuralInformatio nProcessingSystems12,pages666 ,2000.

Show more

Documents from same domain

Visualizing Data using t-SNE - groups.seas.harvard.edu

Visualizing Data using t-SNE - groups.seas.harvard.edu

groups.seas.harvard.edu

VISUALIZING DATA USING T-SNE 2. Stochastic Neighbor Embedding Stochastic Neighbor Embedding (SNE) starts by converting the high-dimensional Euclidean dis-tances between datapoints into conditional probabilities that represent similarities.1 The similarity of datapoint xj to datapoint xi is the conditional probability, pjji, that xi would pick xj as its neighbor

  Using, Harvard

Related documents

Samsung Magician 7

Samsung Magician 7

www.samsung.com

Perfor mance Bench mark RAPID Perfor mance Optimi zation (TRIM) Diagno stic Scan PSID Revert Encrypte d Drive Over provisio ning Secure Erase Security Setting 840 PRO √ √ √ 840 √ √ √ √ √ √ 840 EVO √ √ √ √ √ √ √ √ 750 EVO √ √ 850 PRO √ √ √ √ √ √ √ √

  Perfor mance, Perfor, Mance

A COR's Guide to Statements of Work, Performance Work ...

A COR's Guide to Statements of Work, Performance Work ...

www.navsup.navy.mil

Jan 09, 2011 · tractors to develop the PWS (i.e., their solutions), perfor-mance metrics and a measurement plan, and a quality assurance plan, all of which should be evaluated before contract award. The SOO provides basic, top-level objectives of an acquisition and is provided in the request for proposals . in lieu of a government-written SOW or PWS. It provides

  Performance, Perfor mance, Perfor, Mance

Entropy and Information Theory - Stanford EE

Entropy and Information Theory - Stanford EE

ee.stanford.edu

simple special case of memoryless processes to characterize the optimal perfor- mance theoretically achievable when communicating information sources over constrained random media called channels.

  Information, Theory, Perfor mance, Perfor, Mance, Information theory

Self-Regulated Learning - LINCS

Self-Regulated Learning - LINCS

lincs.ed.gov

confused with a mental ability or an academic perfor-mance skill. Instead, self-regulation is a self-directive process and set of behaviors whereby learners trans-form their mental abilities into skills (Zimmerman, Bonnor, & Kovach, 2002) and habits through a deve-lopmental process (Butler, 1995, 1998, 2002) that

  Perfor mance, Perfor, Mance

Chapter 12: Network Management

Chapter 12: Network Management

www.egr.msu.edu

(ii) fault management, (iii) accounting management, (iv) security management, and (v) perfor-mance management. The ISO classi cation has gained broad acceptance for both standardized and proprietary network management systems. A description of each management function is provided in the following subsections. 2.1 Con guration Management

  Perfor mance, Perfor, Mance

NAVMC 2795 USMC User's Guide to Counseling

NAVMC 2795 USMC User's Guide to Counseling

www.marines.mil

counseling focuses on improving and maintaining future perfor-mance. Counseling should be a regular, continuing process. Counseling may be conducted formally or informally. It may be

  Perfor mance, Perfor, Mance

Chapter N 6 - docs.legis.wisconsin.gov

Chapter N 6 - docs.legis.wisconsin.gov

docs.legis.wisconsin.gov

intervention and evaluation. This standard is met through perfor-mance of each of the following steps of the nursing process: (a) Assessment. Assessment is the systematic and continual collection and analysis of data about the health status of a patient culminating in the formulation of a nursing diagnosis. (b) Planning.

  Wisconsin, Perfor mance, Perfor, Mance, Giles, Docs

Performance Management and Appraisal

Performance Management and Appraisal

www.sagepub.com

management is the process of identifying, measuring, managing, and developing the perfor - mance of the human resources in an organization. Basically we are trying to figure out how well employees perform and then to ultimately improve that performance level. When used

  Performance, Perfor mance, Perfor, Mance

Related search queries

Perfor mance, Performance, Perfor-mance, Information Theory, Perfor- mance, Information, Perfor - mance