Advances in Mathematical Physics

Volume 2017 (2017), Article ID 4538230, 11 pages

https://doi.org/10.1155/2017/4538230

## Asymptotic Stability and Asymptotic Synchronization of Memristive Regulatory-Type Networks

Hubei Normal University, Hubei 435002, China

Correspondence should be addressed to Jin-E Zhang

Received 31 October 2016; Accepted 10 January 2017; Published 26 January 2017

Academic Editor: Xin Yu

Copyright © 2017 Jin-E Zhang. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

Memristive regulatory-type networks are recently emerging as a potential successor to traditional complementary resistive switch models. Qualitative analysis is useful in designing and synthesizing memristive regulatory-type networks. In this paper, we propose several succinct criteria to ensure global asymptotic stability and global asymptotic synchronization for a general class of memristive regulatory-type networks. The experimental simulations also show the performance of theoretical results.

#### 1. Introduction

Using memristive devices as synapses is a focus in memristive networks. To extract the benefits of high-efficiency memristive memory, various memristive networks have been reported to date [1–18]. Unlike conventional two-terminal devices, memristive networks exhibit pinched memristor hysteresis loop characteristics, making them particularly suitable for linear-drift devices [10]. Moreover, as the modular compact model for memristors, memristive regulatory-type networks are further broadened to memristive systems that exhibit the phenomenon of closed-form sneak paths, which enable nanoscale geometries with short access latencies. A memristive regulatory-type network contains multiply-threshold synapses, which has been heralded as a new paradigm in large-scale circuits. Compared with some memristive systems, a memristive regulatory-type network has the following advantages: it is more biomimetic in behaviors with simple system structure; it simplifies the structure and complication of circuits and is easy to realize. With these coveted properties, memristive regulatory-type networks have the potential of realizations in module-based nanoscale neuromorphic computing systems.

The underlying physics mechanism of memristor models is extremely complex. In order to explore the characteristics and applications of memristive networks, several attempts in [1–4, 6–18] have been made, using nonlinear system theory, to develop behavioral models of memristors. An ideal dynamic property is a critical requirement for the development and validation of memristive networks. Evolutional characteristics of memristive networks are an interesting and prosperous research area. However, deploying nonlinear analysis technology in memristive networks is challenging because a memristive network is basically a switched network cluster [13, 15]. Such switched network cluster thus possesses the synaptic action, in which the synaptic weight can be incrementally ameliorated by adjusting the charge or flux through it. There are two major obstacles to analyze and control the memristive networks, namely, high complexity and switched hybridity [12, 13]. On the other hand, dynamical analysis for memristive networks can explain carrier dynamics and associated transients. Once the electronic properties of memristive networks are revealed, then the circuit models can be implemented based upon the underlying dynamic nature. By tweaking physical structures and bias conditions, system designer can optimize the circuit performance, and then, numerous potential applications of the memristors have been exploited, such as neuromorphic, digital, and quantum computation.

In spite of having significant progress in the area of nonlinear control systems [19–35], memristive regulatory-type networks constituting switched network cluster have received less attention. It has been reasoned that much like neuroevolutionary systems, memristive regulatory-type networks could be responsible for different neuromorphic architectures [36, 37]. To this end, we focus on the evolution of memristive regulatory-type networks. In this paper, we study global asymptotic stability and global asymptotic synchronization of a class of memristive regulatory-type networks. Based on -matrix theory, we develop less conservative global asymptotic stability results and global asymptotic synchronization results for memristive regulatory-type networks. Such theoretical analysis can significantly help understand and identify system performance, especially in neuromorphic computing era where stability or synchronization is crucial. In fact, dynamic analysis of memristive regulatory-type networks can provide an overview for optimizing the circuit device and enhancing circuit performances.

The rest of this paper is organized as follows. Section 2 introduces model description and preliminaries. Section 3 gives main results. Section 4 discusses two numerical examples to demonstrate the effectiveness of theoretical results. Finally, Section 5 concludes the paper with some remarks.

#### 2. Model Description and Preliminaries

Consider a general class of memristive regulatory-type networks described by the following delay differential equations: for , ,where and represent the concentration variations of memristive messenger gene and affiliated organic compound , respectively, and denote the degradation rates of memristive messenger gene and affiliated organic compound , respectively, represents the translating rate, nonlinear function is bounded and , and ( is a constant) denote the regulating delay and the translating delay, respectively, and represents regulatory relationship of the network, which is defined aswhere and are constants.

The initial conditions of system (1) are assumed to bewhere and are both continuous functions defined on .

In addition, we also assume that the nonlinear function satisfies the Lipschitz condition with the Lipschitz constant ; that is,

In this paper, solutions of all the systems considered in the following are in Filippov’s sense. denotes closure of the convex hull of set . denotes closure of the convex hull generated by real numbers and . Let , , and , for , .

When considering memristive regulatory-type networks (1), throughout this paper, let us define the set-valued maps as follows:

Obviously, for , ,

By the theory of differential inclusions, from (1), then for , ,

A solution , in the sense of Filippov of system (1) with initial conditions , , and , , , is absolutely continuous on any compact interval of , and

*Definition 1. *The constant vectors and are called an equilibrium point of system (1), if for , ,

*Definition 2. *The equilibrium point of system (1) is said to be globally asymptotically stable if it is locally stable in sense of Lyapunov and globally attractive.

According to Lyapunov direct method, from Definition 2, as we know, if there exists an appropriate Lyapunov function which is positive definite and radially unbounded, such that the time-derivative of along the trajectory of system (1) is negative definite, then the equilibrium point of system (1) is globally asymptotically stable.

*Definition 3. *For drive system , , response system , , and , define the synchronization error signal , ; then the error dynamics can be expressed by the following form:and we say that the response system can be globally asymptotically synchronized with the drive system if the zero solution of error system is globally asymptotically stable.

#### 3. Main Results

In this section, we will first give two lemmas, which play important role in the analysis and synthesis of memristive regulatory-type networks (1).

Lemma 4. *In system (1) at least one equilibrium point exists: ; .*

Lemma 5. *For system (1), we havewhere and are defined as those in (5).*

Using standard arguments as Lemmas and in [15], Lemmas 4 and 5 of this paper can be proved, respectively.

##### 3.1. Global Asymptotic Stability

According to Lemma 4, memristive regulatory-type networks (1) have the equilibrium points and ; we shift the equilibrium points and to the origin by the translation and in the differential inclusion (7), which results inwhere

According to Lemma 5,

Theorem 6. *The equilibrium points and of system (1) is globally asymptotically stable, if the following matrix is a nonsingular -matrix, where , , , and .*

*Proof. *Since matrix is a nonsingular -matrix, by the -matrix theory, it follows that is a nonsingular -matrix. Based on the fact that is a nonsingular -matrix, then there exists an -dimensional vector such that ; that is,Choose and then we get Consider the following positive definite and radially unbounded Lyapunov function:Calculating the upper right Dini derivative of along the trajectory of system (12) yieldsBy Lyapunov global asymptotic stability theory, we can conclude system (12) is globally asymptotically stable. Thus, the equilibrium points and of system (1) are globally asymptotically stable. The proof is completed.

Next we extend Theorem 6 to other possible cases.

Corollary 7. *The equilibrium points and of system (1) are globally asymptotically stable, if*

*Proof. *Select the -dimensional unit vector as in the proof of Theorem 6, from (22); it follows that (17) hold. Therefore, the conclusion of Corollary 7 is obvious.

Corollary 8. *When , the equilibrium points and of system (1) are globally asymptotically stable, if the matrix is a nonsingular -matrix, where , , , and .*

*Proof. *The proof is a direct result of Theorem 6.

##### 3.2. Global Asymptotic Synchronization

Let (1) be the drive memristive regulatory-type networks. The response memristive regulatory-type networks are described by the following: for , ,where , , , , denote the appropriate control inputs that will be designed in order to obtain a certain control objective.

Next, the linear feedback scheme is used to achieve synchronization between drive memristive regulatory-type networks (1) and response memristive regulatory-type networks (24); that is, the controllers , , , , are designed as follows:where , denote the control gains.

Letfor , . Then by drive memristive regulatory-type networks (1), response memristive regulatory-type networks (24), and the controllers (25), the error system can be described byfor , .

To apply the theories of set-valued maps and differential inclusions, (27) is equivalent to

According to Lemma 5,

Theorem 9. *The zero solution of system (28) is globally asymptotically stable; that is, the response system (24) can be globally asymptotically synchronized with the drive system (1), if the following matrix is a nonsingular -matrix, where , , , and .*

*Proof. *Since matrix is a nonsingular -matrix, by the -matrix theory, it follows that is a nonsingular -matrix. Based on the fact that is a nonsingular -matrix, then there exists an -dimensional vector such that ; that is,Choose and then we get Consider the following positive definite and radially unbounded Lyapunov function:Calculating the upper right Dini derivative of along the trajectory of system (28) yieldsBy Lyapunov global asymptotic stability theory, we can conclude that system (28) is globally asymptotically stable. Thus, the response system (24) can be globally asymptotically synchronized with the drive system (1). The proof is completed.

Next we extend Theorem 9 to other possible cases.

Corollary 10. *The zero solution of system (28) is globally asymptotically stable; that is, the response system (24) can be globally asymptotically synchronized with the drive system (1), if*

*Proof. *Select the -dimensional unit vector as in the proof of Theorem 9, from (37), it follows that (32) hold. Therefore, the conclusion of Corollary 10 is obvious.

Corollary 11. *When , the zero solution of system (28) is globally asymptotically stable; that is, the response system (24) can be globally asymptotically synchronized with the drive system (1), if the matrix is a nonsingular -matrix, where , , , and .*

*Proof. *The proof is a direct result of Theorem 9.

*Remark 12. *Theorem 9 and Corollaries 10 and 11 show the feasibility of linear feedback scheme for designing a perfect control in memristive regulatory-type networks, and the sufficient conditions only depend on some system parameters, which are easy to be checked.

*Remark 13. *Compared with many other control strategies, linear feedback scheme is more suitable for implementation in memristive regulatory-type networks. For one thing, transient states are quite prevalent in memristive regulatory-type networks; that is, state-dependent jump abruptly spikes up or down with uncertainty. For another thing, linear feedback scheme itself is relatively cheaper and simpler to operate. It is more reasonable and implementable for linear feedback scheme only carried out at finite gain and bandwidth.

*Remark 14. *The asymptotic synchronization strategy contains more general synchronization behaviors. Through the node cluster, asymptotic synchronization in each group can achieve complete synchronization.

#### 4. Illustrative Examples

In this section, we discuss two numerical examples to illustrate the theoretical results.

*Example 1. *Consider the following memristive regulatory-type networks:where , , ,Obviously, we can calculate that and then and the eigenvalues of matrix are , , , , and ; thus the matrix is a nonsingular -matrix. According to Theorem 6, system (39) is globally asymptotically stable.

The simulation results of system (39) with some initial values are depicted in Figures 1 and 2. Based on the dynamical evolutions in Figures 1 and 2, we can see that the experimental results agree with the theory very well.