Research Article | Open Access

Nawel Aoun, Marwen Kermani, Anis Sakly, "Simple Algebraic Criteria for Asymptotic Stability of a Class of TS Fuzzy Time-Delay Switched Systems", *Mathematical Problems in Engineering*, vol. 2018, Article ID 1265359, 13 pages, 2018. https://doi.org/10.1155/2018/1265359

# Simple Algebraic Criteria for Asymptotic Stability of a Class of TS Fuzzy Time-Delay Switched Systems

**Academic Editor:**Stefan Balint

#### Abstract

This paper investigates the asymptotic stability of a class of TS fuzzy switched systems when an arbitrary switching strategy is adopted. The proposed method, applied to neutral and retarded-type systems, is based on the vector-norms approach. The idea consists in defining a common comparison system to all the fuzzy models. If this comparison system can be described by a state matrix that fulfills the properties of the opposite of an M-matrix, then we can conclude on the asymptotic stability of the initial system via simple algebraic delay-independent conditions.

#### 1. Introduction

Switched systems are defined as an important class of hybrid dynamical systems that consists of several continuous-time or discrete-time subsystems. Each subsystem is activated during a certain time interval according to a specific signal that governs the switching between all the subsystems [1, 2]. In the literature, this signal is sometimes referred to as switching rule or sequence and may be function of time, which is the case in this work, or of the state variables or any other external input.

Switched systems are widely used in many industrial fields as they can properly model many real-world systems which exhibit switching features such as mechanical systems, chemical processes, communication networks, switching power converters, robotic systems, air traffic and aircraft, etc.

Besides and during the past decades, fuzzy theory about modeling and control and more particularly about Takagi-Sugeno (TS) fuzzy models has been recognized as a powerful approximation tool when dealing with complex and ill-determined systems [3–6].

This model-based approach consists in modeling the dynamic of a complex system as a convex sum of the dynamics of several subsystems where each subsystem results from the linearization of the initial nonlinear system around a point of the state space.

Obviously, we live in an era where technology witnesses a rapid evolution. Technological advances lead inevitably to an increasing complexity of systems and thus, new tools and concepts need to be developed in order to solve efficiently the problem of analysis/synthesis of such systems. Multiple nonlinear systems, switched nonlinear systems, and nonholonomic systems for instance are part of these complicated real-world systems whose study requirements have given rise to the new concept of TS fuzzy switched systems [7–9]. By definition, a switched fuzzy system is a switched system whose all subsystems are fuzzy subsystems.

On the other hand, time delay is an inherent feature of many physical processes and it is well known that this phenomenon may cause the deterioration of the system’s performances and can eventually lead to its instability [10]. Time-delay, whether constant or variable, single or multiple, is unavoidable in many engineering applications including rolling mills, pneumatic and hydraulic systems, aircraft and robotic systems, etc.

Time-delay systems can be classified into two types. The first type is called neutral-type systems in which the time-delay argument appears in the derivative of the state variables. They arise mainly in heat exchangers, distributed networks containing lossless transmission lines, etc. The second type, probably the most known, is referred to as retarded-type systems.

The problem of stability and stabilization of switched systems in general (this is also the case for TS fuzzy switched systems) is quite difficult. The problem lies in the fact that the asymptotic stability of each individual system does not necessarily imply that of the whole system. Indeed, the nature of the switching signal, i.e., the instants of switching between the different constituent subsystems, plays an important role in the stability of the overall system [11].

In this context, three strategies are generally adopted. The first one is based on searching sufficient conditions guaranteeing the asymptotic stability of such systems under an arbitrary switching law [12, 13]. This kind of switching, which is the subject of this paper, is also called random or unconstrained switching. An important result in this area states that it suffices to find a common Lyapunov function for all the subsystems to conclude to the asymptotic stability of the overall system. In the case of time-delay switched systems, a Lyapunov-Krasovskii functional is generally searched yielding delay-dependent or delay-independent stability criteria [14].

The second strategy aims at stabilizing the set of modes or subsystems by restricting the class of admissible switching sequences to those in which the interval between any two consecutive switching instants is no smaller than a number called dwell time [15].

The last strategy consists in constructing a stabilizing switching law in order to stabilize a set of initially unstable subsystems.

Undoubtedly, Lyapunov theory has been at the origin of most of the results on stability of TS fuzzy switched time-delay systems under arbitrary switching strategy. The reported results are thus expressed in terms of Linear Matrix Inequalities (LMIs) [16]. In particular, in the case of TS fuzzy neutral or retarded time-delay switched systems, finding a common Lyapunov-Krasovskii functional for all the fuzzy models is a hard task. In fact, the difficulty arises from the computational complexity over LMIs or even from the problem of inexistence of such a functional.

In this paper, we propose to study the stability of TS fuzzy switched time-delay systems under arbitrary switching via the vector-norms approach as an alternative to the Lyapunov method [17–19]. The idea consists in constructing a common pseudo-overvaluing/comparison system that is described by the arrow form matrix. This matrix description permits exploiting the properties of the M-matrices and consequently the application of the Kotelyanski lemma [20]. In the case of nonlinear systems, we recall that this special form is suitable for the application of the Borne and Gentina practical stability criterion since it isolates the nonlinear elements in only one row or column [21].

Stability of this common comparison system permits to conclude to that of the original system by means of simple algebraic and practical conditions. The vector-norms approach, applied to switched time-delay systems, has already been introduced in [22–27]. For instance, in [26], the authors have derived stability conditions for nonlinear retarded-type switched time-delay systems under arbitrary switching. Moreover, delay-dependent stability criteria are established in [27] for switched retarded-type systems modeled by Takagi-Sugeno fuzzy models. However, to the best of the authors’ knowledge, the same approach has not been applied yet to switched systems with neutral-type time delay.

The remainder of this paper is organized as follows. Section 2 gives the description as well as the main results related to the neutral type of the considered class of systems. In Section 3, the retarded type is investigated followed by an extension of the results to the case of systems with multiple delays. Section 4 illustrates the results through two examples. Finally, some concluding remarks are given in Section 5.

*Notations*. The notations used throughout this paper are fairly standard. denotes the -dimensional Euclidean space, is the identity matrix with appropriate dimensions, and denotes Euclidean vector norm. is defined as the scalar product of vectors and . In addition, matrices are assumed to be compatible for algebraic operations, if their dimensions are not explicitly stated. and stand for the transpose and the inverse of matrix , respectively. We denote with if and if and .

#### 2. Neutral-Type TS Fuzzy Switched Time-Delay Systems Described by Functional Differential Equations

##### 2.1. Problem Statement

In this section, we are interested in the autonomous switched systems of neutral type that are described by the following functional differential equation:where , and are nonlinear coefficients for each and , , is a constant time delay, is a continuously differentiable vector-valued initial function, is the number of subsystems switching among each other, and is an exogenous function indicating the active subsystem of index at instant and defined byA change of variable of the form allows system (1) to be represented in the state space byTherefore, the neutral-type time-delay switched system is described by the following state space representation:or equivalently bywhere is the state vector, , , and are matrices of appropriate dimensions, and is the switching signal assumed to be available in real time.

Matrices , , and , are given byandIn addition, each subsystem is described by a set of IF-THEN fuzzy rules where each rule is related to a region of the state space in which the subsystem could be approximated by a local linear model.

Thus, the* l*th rule associated with the* i*th subsystem is given bywhere are the fuzzy sets, are the premise variables, , , and are matrices of appropriate dimensions, and and are the number of fuzzy rules and premise variables, respectively.

For each fuzzy rule , we attribute a weighting factor which depends on the degree of membership of to the fuzzy sets (denoted by ) as well as on the choice of the conjunctive operator “and”. The latter, usually considered as the product operator, is computed as follows:The final output of the switched system is then inferred aswhere is the firing strength of membership function . It is assumed that and . are thus the normalized weighting functions satisfying the properties of convex sums such that and .

Matrices , , and are given byandRelatively to the regular vector norm , system (10) admits the following comparison system:

##### 2.2. Preliminaries

In this subsection, we recall some of the definitions and remarks that will be useful throughout the paper.

**Kotelyanski Lemma** (see [20]). The real parts of the eigenvalues of matrix , with nonnegative off-diagonal elements, are less than a number if and only if those of matrix ; are positive, with the identity matrix.

*Definition 1 (see [21]). *The matrix is called an M-matrix, if the following conditions are met:(i),and,(ii)The principal minors of are all positive(iii)For any positive real vector , the algebraic equations have a positive solution .

*Remark 2. * is the opposite of an M-matrix if is an M-matrix.

*Remark 3. *A continuous-time system characterized by is stable if is the opposite of an M-matrix. In this case, the main minors of are sign-alternate (the first is negative) and the Kotelyanski lemma permits to conclude to the stability of the system characterized by .

*Definition 4. *A dynamic system is said to be a comparison system to another dynamic system according to a specific property (for instance, the stability of the origin), if the verification of the property for system implies the same property for system .

*Definition 5 (see [19, 21]). *If is a pseudo-overvaluing matrix of the system with respect to the vector norm , the following inequality is satisfied:where denotes the right hand derivative operator.

Consequently, the stability of the comparison system with initial conditions implies the same property for the initial system.

##### 2.3. Stability Analysis

In the rest of the paper, for any set of matrices and , we denote by the maximum value of such thatand by the following matrix:

Theorem 6. *System (10) is asymptotically stable under an arbitrary switching law , if the matrix is the opposite of an M-matrix, withand and .*

*Proof. *Let be a strictly positive vector () and choose a common radially unbounded candidate Lyapunov functional for all the fuzzy model as follows:withandIt is clear that .

The right derivative of along the trajectory of system (10) is expressed as follows:wherewithThen,Also, we haveandNotice that and . This permits the simplification of (24) through the following expressions:Similarly,Finally, we obtainwhere , , and is the minimal pseudo-overvaluing matrix for subsystem given byand is a common pseudo-overvaluing matrix already defined in (19).

On the other hand, assume that is the opposite of an M-matrix and according to the M-matrices properties, we can find a vector , () such that . Therefore,This completes the proof of Theorem 6.

Now, to apply Theorem 6, a system’s description by a special matrix form, called the arrow form matrix, can be considered useful. Indeed, a change of base of the form of system (10) leads to the new state space representation:where matrices , , and and are in the arrow form and determined as follows:and is the corresponding passage matrix given byThe elements of matrices , and are computed bywhereas the elements of matrices and , and are defined, respectively, byandNote that are distinct constant parameters that are chosen arbitrarily and that , , and denote the respective instantaneous characteristic polynomials of matrices , , and such thatandThe computation of givesNotice that can be simplified using the fact that . Hence, . Therefore, for and , it is possible to define for all the fuzzy models the following comparison system:such that the arrow from pseudo-overvaluing matrices is given bywithAt this level, we can state the following theorem.

Theorem 7. *System (10) is asymptotically stable under an arbitrary switching law , if there exist , such that the following inequality is satisfied:*

*Proof. *The common pseudo-overvaluing matrix has the form:Since , it suffices to check thatto conclude to the asymptotic stability of system (10).

Knowing that the determinant of the arrow form matrix is computed as follows:and by denoting , we can writeHence, we deduce relation (49).

This achieves the proof of Theorem 7.

*Remark 8. *To study the asymptotic stability of each individual fuzzy model separately, it is obvious that it suffices to verify for each and thatwhere

*Remark 9. *Relation (54) can be widely simplified if some further conditions are met. We give, for instance, two situations in which .

*Situation 10. *If the following inequalities hold:andcondition (54) is reduced to

*Situation 11. *If in addition to (56) and (57), we haveandthen, the asymptotic stability condition becomes

*Proof. *See the Appendix.

*Remark 12. *Relations (60) and (63) are considered useful when the common comparison system is identical to one of the fuzzy models.

#### 3. Retarded-Type TS Fuzzy Switched Systems

##### 3.1. Stability Analysis

In the special case , it is obvious that system (10) becomes

Theorem 13. *System (64) is asymptotically stable under an arbitrary switching rule , if the matrix is the opposite of an -matrix, with*

*Proof. *It suffices to choose the following radially unbounded Lyapunov functional:In this case, all the results of Section 2 are extended to retarded-type TS fuzzy switched systems by replacing by 0.

##### 3.2. Generalization of the Result to the Case of Retarded-Type Systems with Multiple Delay

Consider the class of TS fuzzy switched time-delay systems that are represented bywhere , , , and are matrices of appropriate dimensions, is a continuous vector-valued function specifying the initial state of the system, denotes the multiple delay, and and are already defined in (2) and (11), respectively.

Delay-independent sufficient stability conditions generalized to systems with multiple delays are given in the following theorem.

Theorem 14. *System (67) is asymptotically stable under an arbitrary switching rule , , if matrix is the opposite of an -matrix, with*

*Proof. *In this case, choosing a candidate radially unbounded Lyapunov functional of the formand following the same steps as in the proof of Theorem 6 yields the common comparison system given by .

##### 3.3. Stabilization of TS Fuzzy Switched Systems via Delayed PDC Control

Consider the TS fuzzy switched time-delay system given as above:where matrices and are in the companion form as in (12) and the input matrix is common for all the fuzzy models.

The* l*th fuzzy rule of the PDC controller stabilizing the fuzzy model of index is given bywhere and is the local feedback gain vector. We recall that and are, respectively, the premise variables and the fuzzy sets.

This leads towhich can be written in the case as follows:Denoting , and , the problem is then reduced to determining admissible values of parameters ensuring the asymptotic stability of system (73).

#### 4. Numerical Examples

*Example 15. *Consider a switched system composed of two subsystems where each one is represented by two fuzzy models as follows:whereandA transformation of matrices , , and under the arrow form , and () with gives