Responsive image
博碩士論文 etd-0815117-141903 詳細資訊
Title page for etd-0815117-141903
論文名稱
Title
一種以分散式計算提升計算效率之強韌性分析方法
A Computationally Efficient Method for Robustness Analysis based on Distributed Computation
系所名稱
Department
畢業學年期
Year, semester
語文別
Language
學位類別
Degree
頁數
Number of pages
65
研究生
Author
指導教授
Advisor
召集委員
Convenor
口試委員
Advisory Committee
口試日期
Date of Exam
2017-07-24
繳交日期
Date of Submission
2017-09-16
關鍵字
Keywords
強韌穩定性、分散式運算、KYP引理、二次積分限制、(凸型)最佳化
integral quadratic constraint, robust stability, Distributed computing, convex feasibility problem, KYP lemma
統計
Statistics
本論文已被瀏覽 5674 次,被下載 24
The thesis/dissertation has been browsed 5674 times, has been downloaded 24 times.
中文摘要
許多檢測系統強韌穩定性的方法,乃是將強韌穩定性檢測問題,約化為一(凸型)最佳化的可行性求解問題。在這類方法中,以「二次積分限制」分析方法最具代表性;許多強韌穩定性判別法則,都可視為「二次積分限制」分析方法的應用。在分析具許多子系統的大型網路系統的情況下,應用「二次積分限制方法」所衍生的可行性求解問題會有許多變數需要求解,限制條件的維度也會很大。求解這類問題需要巨大的計算資源;若以單一電腦來求解這類問題,運算時間將會很長且記憶體有可能會不足。

本研究延續文獻[1]的方法,提出一分解「二次積分限制方法」所衍生的大型可行性求解問題成諸多「小維度」問題,且能將「小維度」問題分給多台電腦並行運算求解的方法。吾人所提一之分解方法與[1]相同,但改進了求解「小維度」問題的做法。我們應用KYP引理,並利用文獻[17]中提議的演算法,來改良數值解的精準度及分散式運算的收斂速率。比較文獻[1],吾人所提議的方法在「小維度」問題的狀態變數較少時,在運算效率上具有相當大的優勢。隨機產生的數值實驗結果也印證了這個趨勢。
Abstract
A common and important feature amongst the "modern" robustness analysis approaches is that the analysis problem is transformed into a (convex)
feasibility problem, and the existence of a feasible solution implies that the system under consideration does possess the robustness property
we desire. The Integral Quadratic Constraint (IQC) analysis is a well-known representative of such approaches; as a matter of fact, many robustness criteria can be obtained via applying IQC analysis. Applying the IQC analysis to analyze a large scale complex network which has many
subsystems would usually result in a feasibility problem with many (hundreds of thousands) decision variables as well as many high-dimensional constraints. Solving such problems usually require
gigantic computational resources. In particular, solving such problems in a centralized fashion using a single computer would usually take a
very long time and most likely run out of memory space before finding a solution.

In this thesis, we follow the methodology proposed in [1] and propose a
methodology and algorithm for decomposing a large scale (in terms of the
number of decision variables and the state space dimension) IQC problem
into many small scale problems, and distributing these small problems to
many computers to solve them in a distributed fashion. Our methodology
for decomposing the large scale problem is the same as that proposed in
[1]; however, our algorithm for solving the small scale problems are
different from [1]. An improvement for finding more accurate feasible
solutions in a fast-convergent fashion is proposed, by applying KYP
lemma and the algorithms proposed in [17]. Compared to [1], our approach
has a decisive advantage in the case where the small scale problems have
a small number of decision variables with a small dimension state-space
system. In such cases, our approach is computationally much more
efficient. The results of randomly generated numerical experiments
confirm this fact.
目次 Table of Contents
[緒論+1]
[文獻回顧 +1]
[強韌控制與二次積分限制分析理論 +1]
[分散地分析相互連接的不確定系統 +2]
[凸面可行性問題與分散的演算法 +3]
[研究動機、目的與貢獻 +4]
[符號說明 +5]
[預備知識 +8]
[IQC 定理 +8]
[Kalman-Yakubovich-Popov 引理 +9]
[圖論 +10]
[問題的描述與分解 +17]
[問題的描述 +17]
[為什麼要分解問題? +17]
[問題的分解 +18]
[Sina Khoshfetrat Pakazad的方法 +23]
[我們的方法 +31]
[零散成組的可行性問題與分散演算法 +37]
[凸面可行性問題 +37]
[誤差邊界與有界線性一致性+37]
[乘法空間形式+38]
[凸面最小化形式+39]
[近似操作子與近似分裂法+39]
[向前向後分裂法+40]
[分散解法+42]
[近似分裂法與分散應用+42]
[局部收斂測試+44]
[收斂速率+46]
[可行問題+46]
[不可行問題+47]
[數值結果+48]
[隨機例子的資料+48]
[隨機例子中積分不等式的乘數+48]
[隨機例子中的轉移函數矩陣的結構+48]
[隨機例子中的轉移函數矩陣的結構+48]
[可行的問題+50]
[記憶體不足+50]
[結論與未來展望+51]
[結論+51]
[未來展望+51]
[參考文獻+52]
參考文獻 References
[1] S. K.Pakazad,M.S.Andersen,A.Hansson,andA.Rantzer,“Decomposition
and projectionmethodsfordistributedrobustnessanalysisofinterconnected
uncertain systems"thisresearchissupportedbytheswedishdepartmentofed-
ucation withintheelliitresearchprogram.” IFACProceedingsVolumes, vol.46,
no. 13,pp.194–199,2013.

[2] A. MegretskiandA.Rantzer,“Systemanalysisviaintegralquadraticcon-
straints,” IEEE TransactionsonAutomaticControl, vol.42,no.6,pp.819–
830, 1997.

[3]林俊良,強韌控制系統-分析與設計.國立編譯館,1997.

[4] V. A.Yakubovich,“Frequencyconditionsforabsolutestabilityofcontrolsys-
tems withhysteresisnonlinearities,” Dokl. Akad.NaukSSSR, vol.149,no.2,
pp. 288–291,1963.

[5] C. SchererandS.Weiland,“Linearmatrixinequalitiesincontrol,” Lecture
Notes, DutchInstituteforSystemsandControl,Delft,TheNetherlands, vol.3,
2000.

[6] U. Jönsson,C.-Y.Kao,andH.Fujioka,“Apopovcriterionfornetworkedsys-
tems,” Systems &controlletters, vol.56,no.9,pp.603–610,2007.

[7] C.-Y. Kao,U.Jönsson,andH.Fujioka,“Characterizationofrobuststability
of aclassofinterconnectedsystems,” Automatica, vol.45,no.1,pp.217–224,
2009.

[8] U. T.JonssonandC.-Y.Kao,“Ascalablerobuststabilitycriterionforsystems
with heterogeneouslticomponents,” IEEE TransactionsonAutomaticControl,
vol.55,no.10,pp.2219–2234,2010

[9] K.-K. K.Kim,“Ontherobustnessofinterconnected/networkeduncertainlinear
multi-agentsystems,”2012.

[10] I. LestasandG.Vinnicombe,“Scalabledecentralizedrobuststabilitycertifi-
cates fornetworksofinterconnectedheterogeneousdynamicalsystems,” IEEE
transactionsonautomaticcontrol, vol.51,no.10,pp.1613–1625,2006.

[11] H. H.BauschkeandJ.M.Borwein,“Onprojectionalgorithmsforsolvingconvex
feasibilityproblems,” SIAM review, vol.38,no.3,pp.367–426,1996.

[12] G. T.Herman,“Imagereconstructionfromprojections,” Real-TimeImaging,
vol.1,no.1,pp.3–18,1995.

[13] H. H.BauschkeandJ.M.Borwein,“Ontheconvergenceofvonneumann’s
alternating projectionalgorithmfortwosets,” Set-ValuedAnalysis, vol.1,no.2,
pp. 185–212,1993.

[14] ——, “Dykstra￿salternatingprojectionalgorithmfortwosets,” Journal of
ApproximationTheory, vol.79,no.3,pp.418–443,1994.

[15] A. BeckandM.Teboulle,“Convergencerateanalysisanderrorboundsfor
projectionalgorithmsinconvexfeasibilityproblems,” Optimization Methods
and Software, vol.18,no.4,pp.377–394,2003.

[16] L. Gubin,B.Polyak,andE.Raik,“Themethodofprojectionsforfindingthe
common pointofconvexsets,” USSR ComputationalMathematicsandMathe-
maticalPhysics, vol.7,no.6,pp.1–24,1967.

[17] S. KhoshfetratPakazad,M.S.Andersen,andA.Hansson,“Distributedsolu-
tions forlooselycoupledfeasibilityproblemsusingproximalsplittingmethods,”
Optimization MethodsandSoftware, vol.30,no.1,pp.128–161,2015.
[18] D. P.BertsekasandJ.N.Tsitsiklis, Parallelanddistributedcomputation:nu-
mericalmethods. PrenticehallEnglewoodCliffs,NJ,1989,vol.23.
[19] S. Boyd,N.Parikh,E.Chu,B.Peleato,andJ.Eckstein,“Distributedoptimiza-
tion andstatisticallearningviathealternatingdirectionmethodofmultipliers,”
FoundationsandTrends®inMachineLearning, vol.3,no.1,pp.1–122,2011.
[20] K. C.Kiwiel,C.H.Rosa,andA.Ruszczynski,“Proximaldecompositionvia
alternating linearization,” SIAM JournalonOptimization, vol.9,no.3,pp.
668–689, 1999.
53
[21] P.L.CombettesandJ.-C.Pesquet,“Aproximaldecompositionmethodfor
solving convexvariationalinverseproblems,” Inverse problems, vol.24,no.6,
p. 065014,2008.
[22] A. R.DePierro,“Fromparalleltosequentialprojectionmethodsandvice
versainconvexfeasibility:Resultsandconjectures,” Studies inComputational
Mathematics, vol.8,pp.187–201,2001.
[23] J.-C. PesquetandN.Pustelnik,“Aparallelinertialproximaloptimization
method,” Pacific JournalofOptimization, vol.8,no.2,pp.273–305,2012.
[24] G. Pierra,“Decompositionthroughformalizationinaproductspace,” MathematicalProgramming,
vol.28,no.1,pp.96–115,1984.
[25] D. Butnariu,A.N.Iusem,andR.S.Burachik,“Iterativemethodsofsolving
stochasticconvexfeasibilityproblemsandapplications,” Computational Opti-
mization andApplications, vol.15,no.3,pp.269–307,2000.
[26] Y. CensorandT.Elfving,“Amultiprojectionalgorithmusingbregmanprojec-
tions inaproductspace,” NumericalAlgorithms, vol.8,no.2,pp.221–239,
1994.
[27] A. D.P.A.IUSEM,“Aparallelprojectionmethodforfindingacommonpoint
of afamilyofconvexsets,” Informes deMatemática,SérieB-021/84IMPA,
RiodeJaneiro, 1984.
[28] B. T.Polyak,“Minimizationofunsmoothfunctionals,” USSR Computational
Mathematics andMathematicalPhysics, vol.9,no.3,pp.14–29,1969.
[29] L. T.DosSantos,“Aparallelsubgradientprojectionsmethodfortheconvexfea-
sibilityproblem,” Journal ofComputationalandAppliedMathematics, vol.18,
no. 3,pp.307–320,1987.
[30] P.Galinier,M.Habib,andC.Paul,“Chordalgraphsandtheircliquegraphs,”
in Graph-TheoreticConceptsinComputerScience. Springer,1995,pp.358–
371.

[31] J. R.BlairandB.Peyton,“Anintroductiontochordalgraphsandcliquetrees,”
in Graphtheoryandsparsematrixcomputation. Springer,1993,pp.1–29.
[32] S. K.Pakazad,A.Hansson,M.S.Andersen,andA.Rantzer,“Distributed
robustness analysisofinterconnecteduncertainsystemsusingchordaldecom-
position,” IFACProceedingsVolumes, vol.47,no.3,pp.2594–2599,2014.
[33] P.L.CombettesandJ.-C.Pesquet,“Proximalsplittingmethodsinsignalpro-
cessing,” arXiv preprintarXiv:0912.3522, 2009.
[34] A. BeckandM.Teboulle,“Afastiterativeshrinkage-thresholdingalgorithm
for linearinverseproblems,” SIAM journalonimagingsciences, vol.2,no.1,
pp. 183–202,2009.
[35] P.Tseng,“Onacceleratedproximalgradientmethodsforconvex-concaveopti-
mization.” SubmittedtoSIAMJournalonOptimization, 2008.
電子全文 Fulltext
本電子全文僅授權使用者為學術研究之目的,進行個人非營利性質之檢索、閱讀、列印。請遵守中華民國著作權法之相關規定,切勿任意重製、散佈、改作、轉貼、播送,以免觸法。
論文使用權限 Thesis access permission:自定論文開放時間 user define
開放時間 Available:
校內 Campus: 已公開 available
校外 Off-campus: 已公開 available


紙本論文 Printed copies
紙本論文的公開資訊在102學年度以後相對較為完整。如果需要查詢101學年度以前的紙本論文公開資訊,請聯繫圖資處紙本論文服務櫃台。如有不便之處敬請見諒。
開放時間 available 已公開 available

QR Code