Abstract

A nonlinear stochastic differential-difference control system with delay of neutral type is considered. Sufficient conditions for the exponential stability are derived by using Lyapunov-Krasovskii functionals of quadratic form with exponential factors. Upper bound estimates for the exponential rate of decay are derived.

1. Introduction

The theory and applications of functional differential equations form an important part of modern nonlinear dynamics. Such equations are natural mathematical models for various real life phenomena where the aftereffects are intrinsic features of their functioning. In recent years, functional differential equations have been used to model processes in different areas such as population dynamics and ecology, physiology and medicine, economics, and other natural sciences [13]. In many of the models the initial data and parameters are subjected to random perturbations, or the dynamical systems themselves represent stochastic processes. For this reason, stochastic functional differential equations are widely studied [4, 5].

One of the principal problems of the corresponding mathematical analysis of equations is a comprehensive study of their global dynamics and the related prediction of long-term behaviors in applied models. Of course, the problem of stability of a particular solution plays a significant role. Therefore, the study of stability of linear equations is the first natural and important step in the analysis of more complex nonlinear systems.

When applying the mathematical theory to real-world problems a mere statement of the stability in the system is hardly sufficient. In addition to stability as such, it is of significant importance to obtain constructive and verifiable estimates of the rate of convergence of solutions in time. One of the principal tools used in the related studies is the second Lyapunov method [68]. For functional differential equations, this method has been developing in two main directions in recent years. The first one is the method of finite Lyapunov functions with the additional assumption of Razumikhin type [9, 10]. The second one is the method of Lyapunov-Krasovskii functionals [11, 12]. For stochastic functional differential equations, some aspects of these two lines of research have been developed, for example, in [11, 1319] and [11, 18, 2025], respectively. In the present paper, by using the method of Lyapunov-Krasovskii functionals, we derive sufficient conditions for stability together with the rate of convergence to zero of solutions for a class of linear stochastic functional differential equation of a neutral type.

2. Preliminaries

In solving control problems for linear systems, very often, a scalar function needs to be found such that the system is asymptotically stable. Frequently, such a function depends on a scalar argument which is a linear combination of phase coordinates and its graph lies in the first and the third quadrants of the plane. An investigation of the asymptotic stability of systems with a control function that is, an investigation of systems with a function satisfying , for and a is called an analysis of the absolute stability of control systems [26]. One of the fundamental methods (called a frequency method) was developed by Gelig et al. (see, e.g., the book [27]). Another basic method is the method of Lyapunov’s functions and Lyapunov-Krasovskii functionals. Very often, the appropriate Lyapunov functions and Lyapunov-Krasovskii functionals are constructed as quadratic forms with integral terms containing a given nonlinearity [28, 29]. An overview of the present state can be found, for example, in [30, 31]. Problems of absolute stability of stochastic equations are treated, for example, in [11, 14, 15, 24].

3. Main Results

Consider the following control system of stochastic differential-difference equations of a neutral type where is an -dimensional column vector, , , , , and are real constant matrices, , , and are constant vectors, is a continuous function, is a constant delay, and is a standard scalar Wiener process with An -measurable random process is called a solution of (3.1) if it satisfies, with a probability one, the following integral equation and the initial conditions where are continuous functions. Here and in the remaining part of the paper, we will assume that the initial functions and are continuous random processes. Under those assumptions, a solution to the initial value problem (3.1), (3.5) exists and is unique for all up to its stochastic equivalent solution on the space [4].

We will use the following norms of matrices and vectors where is the largest eigenvalue of the given symmetric matrix (similarly, the symbol denotes the smallest eigenvalue of the given symmetric matrix), and is a positive parameter.

Throughout this paper, we assume that the function satisfies the inequality if where is a positive constant.

For the reader's convenience, we recall that the zero solution of (3.1) is called stable in the square mean if, for every , there exists a such that every solution of (3.1) satisfies provided that the initial conditions (3.5) are such that and . If the zero solution is stable in the square mean and, moreover, then it is called asymptotically stable in the square mean.

Definition 3.1. If there exist positive constants , , and such that the inequality holds on , then the zero solution of (3.1) is called exponentially -integrally stable in the square mean.

In this paper, we prove the exponential -integral stability in the square mean of the differential-difference equation with constant delay (3.1). We employ the method of stochastic Lyapunov-Krasovskii functionals. In [11, 18, 22, 24] the Lyapunov-Krasovskii functional is chosen to be of the form where constants and are such that the total stochastic differential of the functional along solutions is negative definite.

In the present paper, we consider the Lyapunov-Krasovskii functional in the following form: where constants , and positive definite symmetric matrices , are to be restricted later on. This allows us not only to derive sufficient conditions for the stability of the zero solution but also to obtain coefficient estimates of the rate of the exponential decay of solutions.

We set

Then, by using introduced norms, the functional (3.11) yields two-sided estimates where .

We will use an auxiliary -dimensional matrix: where where is a parameter.

Now we establish our main result on the exponential -integral stability of a trivial solution in the square mean of system (3.1) when .

Theorem 3.2. Let . Let there exist positive constants , , and positive definite symmetric matrices , such that the matrix is positively definite as well. Then the zero solution of the system (3.1) is exponentially -integrally stable in the square mean on . Moreover, every solution of (3.1) satisfies the inequality for all where

Proof. We will apply the method of Lyapunov-Krasovskii functionals using functional (3.11). Using the Itô formula, we compute the stochastic differential of (3.11) as follows Taking the mathematical expectation we obtain (we use properties (3.3)) Utilizing the matrix defined by (3.14), the last expression can be rewritten in the following vector matrix form We will show next that solutions of (3.1) decay exponentially by calculating the corresponding exponential rate.
The full derivative of the mathematical expectation for the Lyapunov-Krasovskii functional (3.11) satisfies In the following we will use inequalities being a consequence of (3.13). Let us derive conditions for the coefficients of (3.1) and parameters of the Lyapunov-Krasovskii functional (3.11) such that the following inequality: holds. We use a sequence of the following calculations supposing that either inequality holds, or the opposite inequality is valid.
(1) Let inequality (3.24) holds. Rewrite the right-hand part of inequality (3.22) in the form and substitute the latter into inequality (3.21). This results in or, equivalently, The inequality always holds. Because inequality (3.24) is valid, a differential inequality will be true as well.
(2) Let inequality (3.25) hold. We rewrite the right-hand side of inequality (3.22) in the form and substitute the latter again into inequality (3.21). This results in or in Because inequality (3.25) is valid, a differential inequality will be valid as well.
Analysing inequalities (3.30) and (3.34) we conclude that (3.23) always holds. Solving inequality (3.23) we obtain Now we derive estimates of the rate of the exponential decay of solutions. We use inequalities (3.22), (3.35). It is easy to see that Now, inequality (3.16) is a simple consequence of the latter chain of inequalities.

4. A Scalar Case

As an example, we will apply Theorem 3.2 to a scalar control stochastic differential-difference equation of a neutral type where , , , , , , , , , and are real constants, is a constant delay, and is a standard scalar Wiener process satisfying (3.3). An -measurable random process is called a solution of (4.1) if it satisfies, with a probability one, the following integral equation: The Lyapunov-Krasovskii functional reduces to where we assume and . The matrix reduces to (for simplicity we set ) and has entries where is a parameter. Therefore, the above calculation yields the following result.

Theorem 4.1. Let . Assume that positive constants , , , and are such that the matrix is positive definite. Then the zero solution of (4.1) is exponentially -integrally stable in the square mean on . Moreover, every solution satisfies the following convergence estimate: for all where

Acknowledgments

The second author was supported by Grant P201/10/1032 of Czech Grant Agency, and by Project FEKT/FSI-S-11-1-1159. The third author was supported by Grant P201/11/0768 of Czech Grant Agency, and by the Council of Czech Government MSM 0021630503. The fourth author was supported by Project M/34-2008 of Ukrainian Ministry of Education, Ukraine and by Grant P201/10/1032 of Czech Grant Agency.