The pth mean practical stability problem is studied for a general
class of Itô-type stochastic differential equations over both finite and infinite time horizons.
Instead of the comparison principle, a function η(t) which is nonnegative, nondecreasing,
and differentiable is cooperated with the Lyapunov-like functions to analyze the practical
stability. By using this technique, the difficulty in finding an auxiliary deterministic stable
system is avoided. Then, some sufficient conditions are established that guarantee the pth
moment practical stability of the considered equations. Moreover, the practical stability
is compared with traditional Lyapunov stability; some differences between them are given.
Finally, the results derived in this paper are demonstrated by an illustrative example.
1. Introduction
Lyapunov stability is one of the most important conceptions of stability and has been widely applied to many fields involving nearly all aspects of reality. As we all know, however, the Lyapunov stability is usually employed to study the steady-state property over an infinite horizon and cannot cope with the transient behavior of the trajectory. Therefore, even a stable system in the sense of Lyapunov cannot be applied in the practice since the trajectory exhibits undesirable transient behaviors such as exceeding certain boundary imposed on the trajectory. Moreover, for a Lyapunov stable system, the domain of the desired attractor may be too small to control the initial perturbation in it, which also limits the uses of the Lyapunov stability. On the other hand, for an unstable system in the sense of Lyapunov, it is often the case that its trajectory oscillates sufficiently near by the desired state, which is absolutely acceptable in the practical engineering. As such, we are more interested in the transient behavior over a finite or infinite horizon rather than the steady-state property over an infinite horizon. For this purpose, a new notion of stability, that is, the practical stability has first been proposed in [1], where it has been shown that the Lyapunov stability may not assure the practical stability and vice versa. Subsequently, the theory on the practical stability has been developed in [2–4].
Up to now, the practical stability problem has been well investigated for deterministic differential equations and many desirable results have been achieved. For example, in [5], a concept of finite time stability, as one special case of practical stability proposed in [6], has been introduced to examine the behavior of systems contained within prespecified bounds during a fixed time interval. The practical stability with respect to a set rather than the particular state x=0 has been extended. In [7, 8], some results on the practical stability have been obtained for discontinuous systems and some differences between the practical stability and the Lyapunov stability have been given. In [9], by using the method of Lyapunov function and Dini derivative, some sufficient conditions have been derived for various types of practical stability. In [10], a new definition of generalized practical stability is introduced. By making use of Lyapunov-like functions, some sufficient conditions are established.
With respect to the stochastic differential systems, we just mention the following representative works. The practical stability in the pth mean has been proposed for discontinuous systems in [11]. In [12], by using the Lyapunov-like functions and the comparison principle, a unified approach is developed to deal with the problems of both the pth mean Lyapunov stability and the pth mean practical stability for the delayed stochastic systems. In [13], some criteria of practical stability in probability have been established in terms of deterministic auxiliary systems with initial conditions. The results obtained in [11, 13] have been further extended to a class of large-scale Itô-type stochastic systems in [14], where the initial conditions of the resulting auxiliary systems are random. In all papers mentioned above, the practical stability of the stochastic systems is determined through testing one corresponding auxiliary deterministic system, whereas, in [15], the sufficient conditions for practical stability in the mean square for a class of stochastic dynamical systems are established by using an integrable function and Lyapunov-like functions instead of the comparison principle.
In this paper, we are concerned with the problem of the practical stability in the pth mean for a general class of Itô-type stochastic differential equations over both finite and infinite time intervals. By using Lyapunov-like functions and a nonnegative, nondecreasing, and differentiable function η(t), some criteria are established to ensure the pth mean practical stability for the considered stochastic system. This technique avoids the difficulty in finding an auxiliary deterministic stable system. Moreover, the practical stability is compared with traditional Lyapunov stability and some differences between them are presented. Finally, an illustrative example is provided to demonstrate the results derived in this paper.
Notation.
Rn denotes the n-dimensional Euclidean space. T0 denotes the interval [t0,T), where t0,T∈R+ (in this paper, T can be finite or infinite). M[t0,T) represents the family of nonnegative, nondecreasing, and differentiable functions on [t0,T). C1,2(T0×Rn,R+) represents the family of all real-valued functions V(t,x(t)) defined on T0×Rn which are continuously twice differentiable in x(t)∈Rn and once differentiable in t∈R+. Let (Ω,ℱ,P) be a complete probability space. For a random variable ξ, E∥ξ∥p means the pth mean of ξ. The followings are the other notions in this paper:
S0(t)={x(t)∈Rn:E‖x(t)‖p<λ},S(t)={x(t)∈Rn:E‖x(t)‖p≤A},VMS0(t)=sup{EV(t,x(t)):x(t)∈S0(t)},VmŜ(t)=inf{EV(t,x(t)):x(t)∈∂S(t)},
where λ,A(λ<A) are given.
2. Preliminaries and Definitions
Consider the stochastic system described by the following n-dimensional stochastic differential equation:dx(t)=f(t,x(t))dt+g(t,x(t))dB(t)ont∈T0,x(t0)=x0,
where dx(t) is the stochastic increment in the sense of Itô and B(t) is an m-dimensional Brownian motion. f(t,x(t)) and g(t,x(t)) are n×1 and n×m matrix functions, respectively. And x(t0)=x0 is the initial value. Then, we let x(t)=x(t;t0,x0) be any solution process of (2.1) with the initial value x(t0)=x0. Furthermore, we assume that (2.1) satisfies the theorem of the existence and uniqueness of solutions [16] as follows.
(Lipschitz condition) for all x(t),y(t)∈Rn,and t∈T0,
‖f(t,x(t))-f(t,y(t))‖2⋁‖g(t,x(t))-g(t,y(t))‖2≤K‖x(t)-y(t)‖2.
(Linear growth condition) for all x(t),y(t)∈Rn, and t∈T0,
‖f(t,x(t))‖2⋁‖g(t,x(t))‖2≤K*(1+‖x(t)‖2),
where K and K* are two positive constants.
Note that S0(t) and S(t) satisfy the conditions
S0(t)⊂S(t),∂S0(t)∩∂S(t)=∅.
By using Itô formula, The derivative of the Lyapunov-like function V(t,x(t))∈C1,2(T0×Rn,R+) with respect to t along the solution x(t) of (2.1) is given by
dV(t,x(t))=LV(t,x(t))dt+Vx(t,x(t))g(t,x(t))dB(t),
whereLV(t,x(t))=Vt(t,x(t))+Vx(t,x(t))f(t,x(t))+12trace[gT(t,x(t))Vxx(t,x(t))g(t,x(t))].
Now, we give the definitions on the practical stability in the pth mean for (2.1).
Definition 2.1.
System (2.1) is said to be practically stable in the pth mean (PSM) with respect to (λ,A), 0<λ<A; if there exist (λ,A), then one has E∥x0∥p<λ; one implies that
E‖x(t;t0,x0)‖p<A∀t∈T0.
Remark 2.2.
In Definition 2.1, for T0=[t0,T), if the T is finite time, then the system (2.1) is called finite time practically stable, which is one special case of practical stability.
Noticing the notations of S0(t) and S(t) above, we can see that S0(t0) is a subset of the initial-state set when the initial time is t0, and S(t) is a subset of the state space at time t. Therefore, it is easy to see that E∥x0∥p<λ; one implies that x(t0)∈S0(t0), E∥x(t;t0,x0)∥p<A, and x(t;t0,x0)∈intS(t). Thus, we give the following definition which is equal to Definition 2.1.
Definition 2.3.
System (2.1) is said to be PSM with respect to (λ,A), 0<λ<A, if, for given S0(t),S(t) with S0(t)⊂S(t) and ∂S0(t)∩∂S(t)=∅, one has x(t0)∈S0(t0) then it is implied that
x(t;t0,x0)∈intS(t)∀t∈T0.
Remark 2.4.
If the conditions of Definition 2.3 are satisfied, then the system (2.1) is also said to be practically stable in the pth mean with respect to (S0(t0),S(t)).
In Section 3, the criteria for practical stability in the pth mean will be established for (2.1).
3. Practical Stability Criteria
In this section, the practical stability in the pth mean will be investigated in detail, and some stability criteria will be derived for (2.1) by using a Lyapunov-like function and a nonnegative, nondecreasing, and differentiable function η(t).
Theorem 3.1.
If the following conditions are met:
S0(t)⊂S(t) and ∂S0(t)∩∂S(t)=∅for all t∈T0,
there exists a function V(t,x(t))∈C1,2(T0×Rn,R+), which is satisfying the following conditions:
ELV(t,x(t))≤0t∈T0,x(t)∈S(t),
VMS0(t0)<VmŜ(t)∀t∈T0,
then (2.1) is PSM with respect to (λ,A).
Proof.
For all x0∈S0(t0), let x(t)=x(t;t0,x0) be a solution of (2.1) with the initial value x0. For contradiction, we assume that there exists a first time t1∈T0 such that E∥x(t)∥p<A for t0≤t<t1 and E∥x(t1)∥p=A.
By the notations of VMS0(t), VmŜ(t), and (2)-(b), we have
VMS0(t0)<VmŜ(t1)≤EV(t1,x(t1)).
Noticing the V(t,x(t)) and (2.5), (2.6), it can be obtained that
V(t1,x(t1))-V(t0,x(t0))=∫t0t1LV(s,x(s))ds+∫t0t1Vx(s,x(s))g(s,x(s))dB(s).
By the assumption (2)-(a) and taking the expected value on the both sides of (3.4), we have
E[V(t1,x(t1))-V(t0,x(t0))]=E[∫t0t1LV(s,x(s))ds]=∫t0t1ELV(s,x(s))ds≤0
because
E[∫t0t1Vx(s,x(s))g(s,x(s))dB(s)]=0,
then we have
VMS0(t0)<VmŜ(t1)≤EV(t1,x(t1))≤EV(t0,x(t0))≤VMS0(t0).
This is a contradiction, so the proof is complete.
Remark 3.2.
In Theorem 3.1, if the T0 is a finite time interval, then (2.1) is practically stable on finite time. Furthermore, it should be pointed out that the condition ELV(t,x(t))≤0 is necessary to guarantee the pth moment stability for (2.1) in the sense of Lyapunov. However, it would be too strict for the pth mean practical stability of (2.1). In the following theorem, this condition is replaced by
ELV(t,x(t))≤dη(t)dt,
where η(t)∈M[t0,T).
Theorem 3.3.
If the following conditions are met:
S0(t)⊂S(t) and ∂S0(t)∩∂S(t)=∅ for all t∈T0,
there exists a function η(t)∈M[t0,T), which satisfies the following conditions:
ELV(t,x(t))≤dη(t)dtt∈T0,x(t)∈S(t),
η(t0)=VMS0(t0),
η(t)<VmŜ(t)∀t∈T0,
then (2.1) is PSM with respect to (λ,A).
Proof.
Let x(t) be a solution of (2.1) with the initial value x0∈S0(t0). For contradiction, we assume that the result is not true, which means that there exists a first time t1∈T0 such that E∥x(t)∥p<A for t0≤t<t1 and E∥x(t1)∥p=A.
Noticing the notation of VmŜ(t), we have
VmŜ(t1)≤EV(t1,x(t1)).
By using (2.5), (2.6), it follows that
V(t1,x(t1))-V(t0,x(t0))=∫t0t1LV(s,x(s))ds+∫t0t1Vx(s,x(s))g(s,x(s))dB(s).
Taking the expectation on the both sides of (3.13), considering
E[∫t0t1Vx(s,x(s))g(s,x(s))dB(s)]=0
and the assumption (2)-(a), we obtain
EV(t1,x(t1))=EV(t0,x(t0))+∫t0t1ELV(s,x(s))ds≤EV(t0,x(t0))+∫t0t1dη(s)=EV(t0,x(t0))+η(t1)-η(t0)=η(t1)-[VMS0(t0)-EV(t0,x(t0))]≤η(t1).
Then, it follows from (3.12) and (3.15) that
VmŜ(t1)≤EV(t1,x(t1))≤η(t1)
which contradicts with the condition (2)-(c) of Theorem 3.3, and, hence, the proof is complete.
Remark 3.4.
In Theorem 3.3, we mainly use the function η(t) and the Lyapunov-like functions but not the comparison principle in [11–14] to achieve the result, which avoids the difficulty in finding an auxiliary deterministic stable system. Here, we assume that
ELV(t,x(t))≤dη(t)dt
holds on x(t)∈S(t). Next, the condition x(t)∈S(t) will be replaced by a weaker one, that is, x(t)∈S(t)/S0(t).
Theorem 3.5.
If the following conditions are met:
S0(t)⊂S(t) and ∂S0(t)∩∂S(t)=∅ for all t∈T0,
there exists a function η(t)∈M[t0,T), which satisfies the following conditions:
ELV(t,x(t))≤dη(t)dtt∈T0,x(t)∈S(t)S0(t),
η(t)=VMS0(t),x(t)∈S0(t),
η(t)<VmŜ(t)∀t∈T0,
then (2.1) is PSM with respect to (λ,A).
Proof.
Let x(t) be a solution of (2.1) with the initial value x0∈S0(t0). For contradiction, we assume that there exists a first time t2∈T0 such that E∥x(t)∥p<A for t0≤t<t2 and E∥x(t2)∥p=A. Due to the continuity of E∥x(t)∥p and the connectivity of S(t),S0(t), there exists such a time t1 and E∥x(t1)∥p=λ holds for the last time before the time t2. So, we get that x(t)∈S(t)/S0(t) when t∈[t1,t2].
Noticing the (2.5), (2.6), it can be obtained that, when t∈[t1,t2],
V(t2,x(t2))-V(t1,x(t1))=∫t1t2LV(s,x(s))ds+∫t1t2Vx(s,x(s))g(s,x(s))dB(s).
By virtue of
E[∫t1t2Vx(s,x(s))g(s,x(s))dB(s)]=0,
we take the conditional expectation of (3.21) conditioning on the initial value x(t0)=x0; it can be seen from condition (2)-(a) that
E[V(t2,x(t2))-V(t1,x(t1))∣x(t0)=x0]=E[∫t1t2LV(s,x(s))ds∣x(t0)=x0]=∫t1t2ELV(s,x(s))ds≤∫t1t2dη(s)=η(t2)-η(t1).
Taking the expectation on the both sides of (3.23) and using the assumption (2)-(b), we obtain
EV(t2,x(t2))=EV(t1,x(t1))+η(t2)-η(t1)=η(t2)-[η(t1)-EV(t1,x(t1))]≤η(t2)-[η(t1)-VMS0(t1)]=η(t2).
Then,
VmŜ(t2)≤EV(t2,x(t2))≤η(t2).
Noticing the assumption (2)-(c), this is a contradiction, Then, the proof is complete.
In the theorems above, some sufficient conditions that guarantee the pth mean practical stability are derived for (2.1). It is worth mentioning that the establishment of the practical stability criteria here avoids introducing other auxiliary stable systems, which make it convenient to determine whether an Itô-type stochastic differential system is the pth mean practically stable. In Section 4, an example will be employed to demonstrate the obtained results.
4. Example
In this section, one numerical example is given to demonstrate the result in Theorem 3.3. The results obtained in Theorems 3.1 and 3.5 can be verified in the same way.
Example 4.1.
Consider the one-dimensional stochastic differential equation as follow:
dx(t)=x(t)sin(t)dt+dB(t)ont∈[t0,T),x(t0)=x0,
where B(t) is a one-dimensional Brownian motion.
Let K=K*=1; it is obvious that (4.1) satisfies both the Lipschitz condition and the Linear growth condition, so the existence and uniqueness of the solution x(t) of (4.1) is guaranteed.
Now, we investigate the practical stability in the 1st mean for (4.1) with respect to λ=1 and A=2. One assumes that the initial value x(t0) satisfies the conditions E|x(t0)|<1 and E|x(t;t0,x0)|<2 for t∈T0. Then, we approximate the value of t0 and T.
We define a Lyapunov-like function asV(t,x(t))=|x(t)|.
Due to the fact that V(t,x(t)) is a positive-definite function, one can easily get V(t,x(t))>0 when x(t)≠0.
So, when x(t)≠0, it is obvious that Vt(t,x(t))=0,Vx(t,x(t))={1,x>0,-1,x<0,Vxx(t,x(t))=0.
By using the Itô formula, we calculate the derivative of the Lyapunov-like function V(t,x(t)) along the solution x(t) of (4.1), and noticing the (2.6), we have LV(t,x(t))=Vt(t,x(t))+Vx(t,x(t))f(t,x(t))+12trace[gT(t,x(t))Vxx(t,x(t))g(t,x(t))]=0±x(t)sin(t)+0≤|x(t)|.
Taking the expectation on both sides of (4.4), one obtainsE[LV(t,x(t))]≤E|x(t)|<A=2,
so we defineη(t)=2t.
From (4.4)–(4.6), it can be easily verified that the condition (2)-(a) of Theorem 3.3 is satisfied. Then, by the condition (2)-(b) of Theorem 3.3, we have η(t0)=VMS0(t0)=sup{E|x(t0)|;x(t0)∈S0(t0)}=λ=1
and hence, it can be obtained from (4.6) that t0=12.
On the other hand, from the condition (2)-(c) of Theorem 3.3, we have η(t)<VmŜ(t)=inf{E|x(t)|:x(t)∈∂S(t)}=A=2.
So, we have t<1.
Now, we have the fact that t0=1/2 and T=1. According to Theorem 3.3, (4.1) is practically stable in the 1st mean with respect to λ=1 and A=2 on t∈[1/2,1). In the simulation, we take 50 initial values satisfying E|x(1/2)|<1. For every initial value, the 1st mean orbit and the maximum of E|x(t)| for t∈[1/2,1) are computed numerically. The simulation result is depicted in Figure 1.
Illustration of the practical stability in the 1st mean.
5. Conclusion
This paper mainly establishes the sufficient conditions of practical stability in the pth mean for the Itô-type stochastic differential equation over finite or infinite time interval. By using Lyapunov-like functions and a nonnegative, nondecreasing, and differentiable function η(t) instead of the comparison principle, the difficulty in finding an auxiliary deterministic stable system is avoided. Moreover, this paper indicates that the practical stability can be examined over finite or infinite time interval and it can be used to depict the transient behavior of the trajectory.
For further studies, we can extend practical stability in the pth mean to uniformly practical stability and strict practical stability in the pth mean by the same methods in this paper. And, we can also consider other techniques to establish the sufficient conditions for the practical stability in probability and the almost sure practical stability instead of the comparison principle. Other future research topics include the investigation on the filtering and control problems for uncertain nonlinear stochastic systems; see, for example, [17–26].
Acknowledgment
This paper is supported by the National Natural Science Foundation of China (No. 60974030) and the Science and Technology Project of Education Department in Fujian Province, China (No. JA11211).
LaSalleJ. P.LefschetzS.1961New York, NY, USAAcademic PressBernfeldS. R.LakshmikanthamV.Practical stability and Lyapunov functions198032460761310.2748/tmj/1178229544601930ZBL0438.34040MartynyukA. A.Methods and problems of practical stability of motion theory1984221946MartynyukA. A.SunZ. Q.2003Beijing, ChinaScience PressWeissL.InfanteE. F.On the stability of systems defined over a finite time interval1965544448017942710.1073/pnas.54.1.44ZBL0134.30702WeissL.InfanteE. F.Finite time stability under perturbing forces and on product spaces19671254590209589ZBL0168.33903HeJ. X.Practical stability of discontinuous systems with external perturbations198334385398744509ZBL0535.34039HeJ. X.Practical stability of discontinuous systems198555560LiaoX. X.1988Huazhong Normal University PressZhaiG.MichelA. N.Generalized practical stability analysis of discontinuous dynamical systemsProceedings of the 42nd IEEE Conference on Decision and ControlDecember 2003166316682-s2.0-1542290107FengZ. S.LiuY. Q.GuoF. W.Criteria for practical stability in the pth mean of nonlinear stochastic systems1992492-3251260116216510.1016/0096-3003(92)90028-YZBL0755.93080FengZ. S.Lyapunov stability and practical stability of nonlinear delay stochastic systems: a unified approachProceedings of the 32nd Conference on Decision and ControlDecember 19938658702-s2.0-0027727363TsoiA. H.ZhangB.Practical stabilities of Itô type nonlinear stochastic differential systems and related control problems1997611071241434621ZBL0874.60051SathananthanS.SuthaharanS.Practical stability criteria for large-scale nonlinear stochastic systems by decomposition and aggregation2001822272481824806ZBL0988.93086MichelA. N.HouL.Finite-time and practical stability of a class of stochastic dynamical systemsProceedings of the 47th IEEE Conference on Decision and ControlDecember 2008345234562-s2.0-6294920295410.1109/CDC.2008.4738705MaoX.20082ndEngland, UKInternational Publishers in Science and Technologyxviii+4222380366DongH.WangZ.HoD. W. C.GaoH.Robust H∞ filtering for Markovian
jump systems with randomly occurring nonlinearities and sensor saturation: the
finite-horizon case201159730483057TangY.WangZ.WongW. K.KurthsJ.FangJ.Multiobjective synchroniza-
tion of coupled chaotic systems2011212025114WangZ.LamJ.MaL.BoY.GuoZ.Variance-constrained dissipative observer-based control for a class of nonlinear stochastic systems with degraded measurements20113772645658276916410.1016/j.jmaa.2010.11.038ZBL1214.93104ShenB.WangZ.HungY. S.ChesiG.Distributed H∞ filtering for polynomial
nonlinear stochastic systems in sensor networks201158519711979LiangJ.WangZ.LiuX.Distributed state estimation for discrete-time sensor
networks with randomly varying nonlinearities and missing measurements2011223486496HeX.WangZ.JiY.ZhouD.Robust fault detection for networked systems
with distributed sensors2011471166177ShenB.WangZ.LiuX.Bounded H∞ synchronization and state estimation
for discrete time-varying stochastic complex networks over a finite-horizon2011221145157ShenB.WangZ.ShuH.WeiG.H∞ filtering for uncertain time-varying sys-
tems with multiple randomly occurred nonlinearities and successive packet dropouts2011211416931709WangZ.HoD. W. C.DongH.GaoH.Robust H∞ finite-horizon control for a class of stochastic nonlinear time-varying systems subject to sensor and actuator saturations201055717161722267583810.1109/TAC.2010.2047033WangZ.LiuY.LiuX.Exponential stabilization of a class of stochastic system with Markovian jump parameters and mode-dependent mixed time-delays201055716561662267582710.1109/TAC.2010.2046114