吳付科 張維海
摘要
從所應(yīng)用的主要方法出發(fā),回顧了隨機(jī)連續(xù)系統(tǒng)的各種穩(wěn)定性理論結(jié)果,并探討了這些穩(wěn)定性之間的關(guān)系.關(guān)鍵詞隨機(jī)系統(tǒng);隨機(jī)微分方程;幾乎處處穩(wěn)定性;矩穩(wěn)定性;依概率穩(wěn)定性;分布穩(wěn)定性;隨機(jī)鎮(zhèn)定
中圖分類號(hào)P393
文獻(xiàn)標(biāo)志碼A
收稿日期20170416
資助項(xiàng)目國(guó)家自然科學(xué)基金(61473125);國(guó)家自然科學(xué)優(yōu)秀青年基金(11422110)
作者簡(jiǎn)介
吳付科,男,博士,教授,2011年入選教育部新世紀(jì)優(yōu)秀人才支持計(jì)劃,2014年獲得基金委優(yōu)秀青年基金資助,主要從事隨機(jī)微分方程以及相關(guān)領(lǐng)域的研究.wufuke@hust.edu.cn
1前言及穩(wěn)定性介紹
隨機(jī)現(xiàn)象廣泛存在于生物、金融、通信及控制等領(lǐng)域,是影響系統(tǒng)性質(zhì)的重要因素.當(dāng)一個(gè)系統(tǒng)受到隨機(jī)波動(dòng)的干擾時(shí),結(jié)果將變得更加多樣和復(fù)雜.比如:在生物系統(tǒng)中,隨機(jī)因素往往是生物多樣性的關(guān)鍵因素[12],同時(shí),適當(dāng)?shù)碾S機(jī)因素也能誘導(dǎo)系統(tǒng)產(chǎn)生新的穩(wěn)定狀態(tài)[34]或?qū)е路N群產(chǎn)生穩(wěn)定分布[5],另一方面,過強(qiáng)的隨機(jī)沖擊也能導(dǎo)致種群滅絕[67].由此可見,隨機(jī)因素的引入使系統(tǒng)產(chǎn)生了豐富的研究課題,研究隨機(jī)因素對(duì)系統(tǒng)的影響對(duì)于揭示系統(tǒng)運(yùn)行的機(jī)制具有重要意義.由于這些隨機(jī)系統(tǒng)往往需要用隨機(jī)微分方程描述,因此從數(shù)學(xué)的角度來討論隨機(jī)微分方程的性質(zhì)及其應(yīng)用就變得至關(guān)重要.
自從It引進(jìn)隨機(jī)積分以來的半個(gè)多世紀(jì)里,隨機(jī)微分方程獲得了迅速的發(fā)展,當(dāng)Lyapunov 方法被引入隨機(jī)微分方程之后,隨機(jī)微分方程的穩(wěn)定性理論獲得了快速發(fā)展,在Arnold[8]、Friedman[9]、Khasminskii[10]、Kushner[1113]和毛學(xué)榮教授[1416]及其他學(xué)者的努力下,隨機(jī)穩(wěn)定性理論及其應(yīng)用已經(jīng)形成了一個(gè)龐大的理論體系,在自動(dòng)控制、生物化學(xué)反應(yīng)、通信和制造領(lǐng)域具有重要的應(yīng)用價(jià)值.本文的主要目的是從所利用的方法出發(fā),回顧近年來隨機(jī)微分方程穩(wěn)定性理論的發(fā)展、研究的方法和一些應(yīng)該注意的研究課題.
本文利用如下記號(hào):|·|表示n維歐式空間Rn的范數(shù),如果A 是向量或者矩陣,則A′表示其轉(zhuǎn)置,如果A 是矩陣,其跡范數(shù)表示為A′A,R+=[0,∞).
(Ω,F(xiàn),{Ft}t≥0,P)表示一個(gè)完備的概率空間,{Ft}t≥0是一個(gè)滿足通常條件(即遞增、右連續(xù)且包含所有的零概率集)的σ代數(shù)流.w(t)是定義于這個(gè)概率空間上的m維Brown運(yùn)動(dòng),不失一般性,假定{Ft}t≥0就是w(t)生成的自然流,即Ft=σ(w(s):0≤s≤t).用Lp(Ω,F(xiàn),P)表示隨機(jī)變量x的集合滿足E|x|p<∞.
本文從研究隨機(jī)穩(wěn)定性的常用方法出發(fā),回顧方程(1)的各種穩(wěn)定性結(jié)果.為了使結(jié)果更加聚焦,本文不考慮帶有控制項(xiàng)和Markov切換項(xiàng)的問題,雖然這些問題也同樣具有豐富的成果和重要的意義.又因?yàn)楣P者的知識(shí)范圍所限,對(duì)于后面三種穩(wěn)定性相對(duì)較為熟悉一些,因此本文主要考慮p階矩穩(wěn)定性、幾乎處處穩(wěn)定性和依分布穩(wěn)定性.但是在穩(wěn)定性之間的關(guān)系討論時(shí),也討論了依概率穩(wěn)定性與其他三種穩(wěn)定性的關(guān)系.因?yàn)槊糠N穩(wěn)定性都有海量的文獻(xiàn),也有很多的綜述文章,比如文獻(xiàn)[17]等,所以本文在回顧這些穩(wěn)定性結(jié)果的時(shí)候,主要從所利用的方法出發(fā),討論同類的方法在當(dāng)前文獻(xiàn)中的應(yīng)用.
2幾乎處處穩(wěn)定性
幾乎處處穩(wěn)定性也就是軌道穩(wěn)定性,刻畫隨機(jī)微分方程解的軌道的漸近性質(zhì),主要的方法是基于It公式基礎(chǔ)上的Lyanpunov函數(shù)方法,運(yùn)用的技術(shù)主要是指數(shù)鞅不等式、大數(shù)定理或半鞅收斂定理等,或者在一定的條件下通過p階矩穩(wěn)定性得到.關(guān)于通過矩穩(wěn)定性得出幾乎處處穩(wěn)定性的問題,將在后面在穩(wěn)定性之間的關(guān)系中描述,此處重點(diǎn)回顧指數(shù)鞅不等式、大數(shù)定理和半鞅收斂定理的技術(shù)在幾乎處處穩(wěn)定性研究中的應(yīng)用.首先回顧如下基于指數(shù)鞅不等式的結(jié)果(參考文獻(xiàn)[16]):
定理1假設(shè)存在一個(gè)函數(shù)V∈C2,1(Rn×R+;R+)和常數(shù)p>0,c1>0,c2∈R,c3>0,使得對(duì)任意的x≠0和t≥0,
2)借助于LF,G,可以建立隨機(jī)系統(tǒng)精確能觀性、精確能檢測(cè)性的PBH判據(jù),從而將線性系統(tǒng)理論中關(guān)于完全能觀性、完全能檢測(cè)性的PBH判據(jù)推廣到隨機(jī)系統(tǒng)[5051,58];
3)借助于微分同胚變換,將一個(gè)非線性的隨機(jī)時(shí)不變系統(tǒng)轉(zhuǎn)化為一個(gè)線性隨機(jī)系統(tǒng)[59],然后借助于LF,G,同樣可以討論非線性隨機(jī)時(shí)不變系統(tǒng)的區(qū)域穩(wěn)定性問題,這是一個(gè)值得探索的方向;
4) 若隨機(jī)系統(tǒng)中帶有控制變量u,則可以考慮隨機(jī)系統(tǒng)的極點(diǎn)型配置問題.文獻(xiàn)[5051]中提出了一些未解決的問題,值得探索.
4依分布穩(wěn)定性
隨機(jī)過程的分布穩(wěn)定性本質(zhì)上說明隨機(jī)過程的統(tǒng)計(jì)特性(比如隨機(jī)過程的期望、方差和矩等)不隨時(shí)間的變化而改變.如果隨機(jī)過程是遍歷的,則穩(wěn)定分布就可以看做這個(gè)隨機(jī)過程的極限分布.本文主要回顧兩類研究分布穩(wěn)定性的方法,第一類方法由Khasminskii基于Markov過程的常返性所建立的理論(參考文獻(xiàn) [23]第四章),對(duì)于方程(1),決定它的解過程的Markov性及其常返性,主要體現(xiàn)為如下假設(shè):
定理13在假設(shè)2成立的條件下,如果λ1>λ2,方程(1)的解過程x(t)存在一個(gè)唯一的穩(wěn)定分布(不變測(cè)度)μ,并且這個(gè)不變測(cè)度是指數(shù)混合的(exponentially mixing).
對(duì)于基于第一類方法的分布穩(wěn)定性,毛學(xué)榮教授[5]利用其建立了隨機(jī)LotkaVolterra種群方程的穩(wěn)定分布的存在唯一性,劉紅等[60]考慮了帶有切換的LotkaVolterra利他種群系統(tǒng)的遍歷性和正常返性的問題,關(guān)于隨機(jī)種群系統(tǒng)的不變測(cè)度更進(jìn)一步的討論可以參考文獻(xiàn)[61].
對(duì)于基于第二類方法的分布穩(wěn)定性,張希承教授建立了非Lipschitz條件下的不變測(cè)度的存在性和指數(shù)遍歷性結(jié)果[62],席福寶教授[63]討論了狀態(tài)依賴的切換擴(kuò)散過程的Feller性和指數(shù)遍歷性問題,王健教授討論了Levy過程驅(qū)動(dòng)的OrnsteinUhlenbeck過程的不變測(cè)度的存在性和指數(shù)遍歷性問題[64].
由于以上兩種方法都基于隨機(jī)過程的Markov性,延遲系統(tǒng)的解不滿足Markov性,因此建立隨機(jī)延遲系統(tǒng)的不變測(cè)度和遍歷性很長(zhǎng)時(shí)間沒有進(jìn)展,在Mohammed考慮隨機(jī)泛函微分方程中解映射的適應(yīng)性、Markov性等基礎(chǔ)上[65],鮑建海等[6069]通過一系列文章建立各種隨機(jī)延遲和泛函微分方程并討論了隨機(jī)偏微分方程中解映射的不變測(cè)度的存在性和遍歷性問題.文獻(xiàn)[70]建立了無窮延遲的隨機(jī)泛函微分方程的解映射的不變測(cè)度和遍歷性結(jié)果.
5各種穩(wěn)定性之間的關(guān)系
由于隨機(jī)性的引入,導(dǎo)致了隨機(jī)序列的收斂性更加多樣,各種穩(wěn)定性之間既互有聯(lián)系,又有強(qiáng)弱的不同,因此有了更加豐富的結(jié)果.根據(jù)經(jīng)典的概率論和隨機(jī)過程知識(shí),以上各種穩(wěn)定性之間關(guān)系如下(參考文獻(xiàn)[71]):
定理14以上四種穩(wěn)定性關(guān)系如下:
1)幾乎處處穩(wěn)定性依概率穩(wěn)定性;
2)p階矩穩(wěn)定性依概率穩(wěn)定性;
3)依概率穩(wěn)定性依分布穩(wěn)定性;
4)依概率穩(wěn)定性存在一個(gè)子序列有幾乎處處穩(wěn)定性.
在某些條件下,上面有的關(guān)系是可逆的,比如:如果一個(gè)隨機(jī)變量依分布收斂到一個(gè)常數(shù)(退化分布),那么這個(gè)收斂性也是依概率收斂的;如果存在一個(gè)隨機(jī)變量y∈Lp(Ω,F(xiàn),P)并且對(duì)任意的t≥0,隨機(jī)過程|x(t)|≤y,則此時(shí)幾乎處處穩(wěn)定性是最強(qiáng)的穩(wěn)定性.根據(jù)控制收斂定理,幾乎處處穩(wěn)定性可以得出p階矩穩(wěn)定性,而且進(jìn)一步,依分布穩(wěn)定性也可以得出p階矩穩(wěn)定性.
在隨機(jī)系統(tǒng)中,有一個(gè)值得注意的結(jié)果是如果對(duì)方程(1)的系數(shù)施加一定的條件,則從這個(gè)方程的平凡解的p階矩穩(wěn)定性可以得出幾乎處處穩(wěn)定性(參考文獻(xiàn)[16]),結(jié)果可以描述如下:
6結(jié)束語
由于篇幅和筆者的知識(shí)范圍所限,本文仍然有較多的重要結(jié)果沒有涉及.對(duì)于延遲系統(tǒng),Yorke的方法和隨機(jī)Razumikhin定理是研究延遲系統(tǒng)的一個(gè)重要的穩(wěn)定性原理,毛學(xué)榮教授[7374]建立了隨機(jī)形式的Razumikhin定理,據(jù)此得出了p階矩穩(wěn)定性,但是如何直接利用Razuminskii定理的思想建立幾乎處處穩(wěn)定性仍然是一個(gè)沒有解決的問題.
參考文獻(xiàn)
References
[1]Ozbudak E M,Thattai M,Kurtser I,et al.Regulation of noise in the expression of a single gene[J].Nature genetics,2002,31:6973
[2]Thattai M,Van Oudenaarden A.Stochastic gene expression in fluctuating environments[J].Genetics,2004,167:523530
[3]Kepler T B,Elston T C.Stochasticity in transcriptional regulation:origins,consequences,and mathematical representations[J].Biophysical Journal,2001,81:31163136
[4]Turcotte M,GarciaOjalvo J,Süel G M.A genetic timer through noiseinduced stabilization of an unstable state[J].Proceedings of the National Academy of Sciences,2008,105:1573215737
[5]Mao X.Stationary distribution of stochastic population systems[J].Systems & Control Letters,2011,60(6):398405
[6]Mao X.Delay population dynamics and environmental noise[J].Stochastics and Dynamics,2005,5(2):149162
[7]Wu F,Yin G.Environmental noises produce suppression and extinction in stochastic LotkaVolterra model with infinite delay[J].Journal of Mathematical Analysis and Applications,2012,396:772785
[8]Arnold L.Stochastic differential equations:Theory and applications[M].New York:Wiley,1972
[9]Friedman A.Stochastic differential equations and applications[M].New York:Academic Press,1975
[10]Khasminskii R Z.Stochastic stability of differential equations[M].Berlin:SpringerVerlag,2012
[11]Kushner H J.Stochastic stability and control[M].New York:Academic Press,1967
[12]Kushner H J.Approximation and weak convergence methods for random processes,with applications to stochastic systems theory[M].Cambridge:MIT Press,1984
[13]Kushner H J.Weak convergence methods and singularly perturbed stochastic control and filtering problems[M].Boston:Birkhuser,1990
[14]Mao X.Stability of stochastic differential equations with respect to semimartingales[J].Longman Scientific and Technical,1991,35(2):267277
[15]Mao X.Exponential stability of stochastic differential equations[M].New York:Marcel Dekker,1994
[16]Mao X.Stochatic differential equations and applications[M].Chichester:Horwood Publishing,1997
[17]Teel A R,Subbaraman A,Sferlazza A.Stability analysis for stochastic hybrid systems:A survey[J]Automatica,2014,50(10):24352456
[18]Zhang H,Wu Z,Xia Y.Exponential stability of stochastic systems with hysteresis switching[J].Automatica,2014,50(2):599606
[19]Mao X,Song Q,Yang D.A note on exponential almost sure stability of stochastic differential equation[J].Bulletin of the Korean Mathematical Society,2014,51(1):221227
[20]Caraballo T,Hammami M A,Mchiri L.Practical asymptotic stability of nonlinear stochastic evolution equation[J].Stochastic Analysis and Applications,2014,32(1):7787
[21]Barbata A,Zasadzinski M,Ali H S,et al.Exponential observer for a class of onesided Lipschitz stochastic nonlinear systems[J].IEEE Transactions on Automatic Control,2015,60(1):259264
[22]Zong X,Wu F,Yin G,et al.Almost sure and pthmoment stability and stabilization of regimeswitching jump diffusion systems[J].SIAM Journal on Control and Optimization,2014,52(4):25952622
[23]Hale J K,Lunel S M V.Introduction to functional differential equations[M].New York:SpringerVerlag,1993
[24]LaSalle J P.Stability theory of ordinary differential equations[J].Journal of Differential Equations,1968,4:5765
[25]Mao X.Stochastic versions of the LaSalle theorem[J].Journal of Differential Equations,1999,153:175195
[26]Mao X.Some contributions to stochastic asymptotic stability and boundedness via multiple Lyapunov functions[J].Journal of Mathematical Analysis and Applications,2001,260:325340
[27]Gihman I I,Skorokhod A V.The theory of stochastic processes I[M].Heideberg:SpringerVerlag,2004
[28]Liptser R Sh,Shiryayev A N.Theory of martingales[M].Dordrecht:Kluwer Academic Publishers,1989
[29]Caraballo T,GarridoAtienza M J,Real J.Stochastic stabilization of differential systems with general decay rate[J].System & Control Letters,2003,48:397406
[30]Wu F,Hu S,Huang C.Robustness of general decay stability of nonlinear neutral stochastic functional differential equations with infinite delay[J].System & Control Letters,2010,59:195202
[31]Basin M,Rodkina A.On delaydependent stability for a class of nonlinear stochastic systems with multiple state delays[J].Nonlinear Analysis:Theory,Methods & Applications,2008,68(8):21472157
[32]Rodkina A,Basin M.On delaydependent stability for a class of nonlinear stochastic delaydifferential equations[J].Mathematics of Control,Signals and Systems,2006,18(2):187197
[33]Mao X.Attraction,stability and boundedness for stochastic differential delay equations[J].Nonlinear Analysis,2001,47:47954806
[34]Shen Y,Luo Q,Mao X.The improved LaSalletype theorems for stochastic functional differential equations[J].Journal of Mathematical Analysis and Applications,2006,318(1):134154
[35]Wu F,Hu S.Attraction,stability and robustness for stochastic functional differential equations with infinite delay[J].Automatica,2011,47:22242232
[36]Blythe S,Mao X,Liao X.Stability of stochastic delay neural networks[J].Journal of Franklin Institute,2001,338:481495
[37]Yuan C,Mao X.Attraction and stochastic asymptotic stability and boundedness of stochastic functional differential equations with respect to semimartingales[J].Stochastic Analysis and Applications,2006,24:11691184
[38]Huang C,Cao J.Almost sure exponential stability of stochastic cellular neural networks with unbounded distributed delays[J].Neurocomputing,2009,72(13):33523356
[39]Li M,Deng F.Almost sure stability with general decay rate of neutral stochastic delayed hybrid systems with Lévy noise[J].Nonlinear Analysis(Hybrid Systems),2017,24:171185
[40]Mao X.Stability and stabilisation of stochastic differential delay equations[J].IET Control Theory & Applications,2007,1(6):15511566
[41]Zhao X,Deng F.A new type of stability theorem for stochastic systems with application to stochastic stabilization[J].IEEE Transactions on Automatic Control,2016,61(1):240245
[42]Liu W,Mao X.Almost sure stability of the EulerMaruyama method with random variable stepsize for stochastic differential equations[M].Numerical Algorithms,2016:120
[43]Wu F,Mao X,Szpruch L.Almost sure exponential stability of numerical solutions for stochastic delay differential equations[J].Numerische Mathematik,2010,115:681697
[44]Zhang W,Lin X,Chen B.LaSalletype theorem and its applications to infinite horizon optimal control of discretetime nonlinear stochastic systems[J].IEEE Transactions on Automatic Control,2017,62:250272
[45]Appleby J A D,Cheng J,Rodkina A.Characterisation of the asymptotic behaviour of scalar linear differential equations with respect to a fading stochastic perturbation[J].Discrete & Continuous Dynamical Systems,2011,3(3):7990
[46]Chen W,Guan Z,Lu X.Delaydependent exponential stability of uncertain stochastic systems with multiple delays:An LMI approach[J].Systems & Control Letters,2005,54(6):547555
[47]Xu S,Lam J,Mao X,et al.A new LMI condition for delaydependent robust stability of stochastic timedelay systems[J].Asian Journal of Control,2005,7(4):419423
[48]Liao X,Chen G,SanchezE N.Delaydependent exponential stability analysis of delayed neural networks:an LMI approach[J].Neural networks,2002,15(7):855866
[49]Da Prato G,Zabczyk J.Stochastic equations in infinite dimensions,encyclopedia of mathematics and its applications[M].Cambridge:Cambridge University Press,1992
[50]Zhang W,Chen B S.On stabilization and exact observability of stochastic systems with their applications[J].Automatica,2004,40(1):8794
[51]Zhang W,Zhang H,Chen B S.Generalized Lyapunov equation approach to statedependent stochastic stabilization/detectability criterion[J].IEEE Trans Automat Control,2008,53(7):16301642
[52]Hou T,Zhang W,Ma H.Essential instability and essential destabilisation of linear stochastic systems[J].IET Control Theory and Applications,2011,5(2):334340
[53]黃琳.穩(wěn)定性理論[M].北京:北京大學(xué)出版社,1992
HUANG Lin.Stability theories[M].Beijing:Peking University Press,1992
[54]Zhang W,Xie L.Interval stability and stabilization of linear stochastic systems[J].IEEE Trans Automat Control,2009,54(4):810815
[55]Chilali M,Gahinet P,Apkarian P.Robust pole placement in LMI regions[J].IEEE Trans Automat Control,1999,44(12):22572270
[56]Gutman S,Jury E I.A general theory for matrix root clustering in subregions of the complex plan[J].IEEE Trans Automat Control,1981,26(4):853863
[57]Zhang W,Chen B S.Hrepresentation and applications to
generalized Lyapunov equations and linear stochastic systems[J].IEEE Trans Automat Control,2012,57(12):30093022
[58]Hou T,Ma H,Zhang W.Spectral tests for observability and detectability of periodic Markov jump systems with nonhomogeneous Markov chain[J].Automatica,2016,63(1):175181
[59]Pan Z.Differential geometric condition for feedback complete linearization of stochastic nonlinear system[J].Automatica,2001,37:145149
[60]Liu H,Li X,Yang Q.The ergodicity property and positive recurrence of a multigroup LotakVolterra mutualistic system with regime switching[J].System & Control Letters,2013,62:805810
[61]Nguyen D H,Yin G.Coexistence and exclusion of stochastic competitive LotkaVolterra models[J].Journal of Differential Equations,2017,262(3):11921225
[62]Zhang X.Exponential ergodicity of nonLipschitz stochastic differential equations[J].Proceeding of the American Mathematical Society,2009,137(1):329337
[63]Xi F.Feller property and exponential ergodicity of diffusion processes with statedependent switching[J].Science in China Series A:Mathematics,2008,51(3):329342
[64]Wang J.On the exponential ergodicity of Lévydriven OrnsteinUhlenbeck processes[J].Journal of Applied Probability,2012,49(4):9901004
[65]Mohammed S E A.Stochastic functional differential equations[J].Pitman Advanced Pub Program,1984,18(2):6364
[66]Bao J,Wang F Y,Yuan C.Bismut formulae and applications for functional SPDEs[J].Bulletin des Sciences Mathèmatiques,2013,137:509522
[67]Bao J,Yin G,Yuan C.Ergodicity for functional stochastic differential equations and applications[J].Nonlinear Analysis,2014,98:6682
[68]Bao J,Yin G,Yuan C,et al.Exponential ergodicity for retarded stochastic differential equations[J].Applied Analysis,2014,93:23302349
[69]Bao J,Yin G,Yuan C.Stationary distributions for retarded stochastic differential equations without dissipativity[J].Stochastics,2017,89(2):530549
[70]Wu F,Yin G,Mei H.Stochastic functional differential equations with infinite delay:Existence and uniqueness of solutions,solution maps,Markov properties,and ergodicity[J].Journal of Differential Equations,2017,262(3):12261252
[71]Shiryaev A N.Probability[M].New York:SpringerVerlag,1996
[72]Karatzas I,Shreve S E.Brownian motion and stochastic calculus[M].New York:SpringerVerlag,2012
[73]Mao X.Razumikhintype theorems on exponential stability of stochastic functionaldifferential equations[J].Stochastic Processes and Their Applications,1996,65(2):233250
[74]Mao X.Razumikhintype theorems on exponential stability of neutral stochastic functionaldifferential equations[J].SIAM Journal on Mathematical Analysis,1997,28(2):389401
Abstract
From aspects of the research methods,this paper reviews various classes of stability results of continuous stochastic systems,and discusses the relationship among these stabilities under different conditions.
Key wordsstochastic systems;stochastic differential equations;almost sure stability;moment stability;stability in probability;stationary distribution;stochastic stabilization