## Open Challenges on the Stability of Complex Systems: Insights of Nonlinear Phenomena with or without Delay 2020

View this Special IssueResearch Article | Open Access

Xiangru Xing, Jin-E Zhang, "Input-to-State Stabilization of a Class of Uncertain Nonlinear Systems via Observer-Based Event-Triggered Impulsive Control", *Complexity*, vol. 2020, Article ID 3951381, 19 pages, 2020. https://doi.org/10.1155/2020/3951381

# Input-to-State Stabilization of a Class of Uncertain Nonlinear Systems via Observer-Based Event-Triggered Impulsive Control

**Guest Editor:**Edgar Cristian DÃaz GonzÃ¡lez

#### Abstract

This article concerns the problem of input-to-state stabilization for a group of uncertain nonlinear systems equipped with nonabsolutely available states and exogenous disturbances. To appropriately cope with these partially measurable state variables as well as dramatically minimize controller updating burden and communication costs, an event-triggered mechanism is skillfully devised and an observer-based impulsive controller with the combination of sample control is correspondingly presented. By resorting to the iterative method and Lyapunov technology, some sufficient criteria are established to guarantee the input-to-state stability of the newly uncertain controlled system under the employed controller, in which an innovative approximation condition as to the uncertain term is proposed and the linear matrix inequality technique is utilized for restraining sophisticated parameter uncertainties. Furthermore, the Zeno behavior in the proposed event-triggered strategy is excluded. The control gains and event-triggered mechanism parameters are conjointly designed by resolving some inequalities of linear matrix. Eventually, the availability and feasibility of the achieved theoretical works are elucidated by two simulation examples.

#### 1. Introduction

Since it is originally put forward by [1, 2], input-to-state stabilization has caught widespread attention [3â€“5], attributing to its performance in characterizing dynamical systems reaction to exogenous disturbances with bounded magnitude. The property of input-to-state stabilization, crudely speaking, symbolizes that the system state will ultimately approach the origin neighborhood whose dimension is in direct proportion to the size of the system input regardless of the magnitude of the initial state. With this characteristic, a system is asymptotically stabilizable under disturbance-free condition and has the evolution of bounded state in the bounded perturbation circumstance. Indeed, input-to-state stability behavior can characterize robustness and stability on dynamic systems possessed disturbances, in which the corresponding stabilization problem has a great signality for the control issue of [5â€“7]. Input-to-state stability is incipiently introduced for continuous systems to evaluate dynamical behaviors, which is especially a fundamental conception for investigating robust dynamics on nonlinear systems influenced by noise, inputs, or interferences [8]. Afterwards, it is diffusely capitalized for stabilizing controller synthesis and stabilization analysis of diverse discontinuous systems, to name a few, switched systems [9, 10], stochastic systems [5, 11], and fuzzy systems [12, 13].

Accompanied by the prompt development of some technologies such as digital control for resource-limited models and sensors incorporated embedded microprocessors, event-triggered impulsive control strategy, more recently, has been highly valued. On the one hand, the impulsive system, composed of discrete dynamics and continuous dynamics, is an important hybrid system in which the uncontinuous behavior is a momentary state jump occurring at given moments, while the consecutive behavior is usually expressed as differential equation. Correspondingly, impulsive control is a control approach that the control signals are transmitted to a system only at certain moments. In comparison with continuous control [14, 15], it has the advantages of only discrete control which is required for deriving the desired performance, discontinuity, and stronger robustness. Consequently, the control approach is extensively applied in practice, such as ecosystem management [16, 17], satellite orbit transfer [18], secure communication [19], pharmacokinetics [20], and complex switched network [21]. Furthermore, controlling the operation of systems all the while is unnecessary or even impossible in practice. In population model [16], for example, it is merited to release predators at appropriate discrete circumstances, rather than the continuous instances for controlling the amount of a category insect. Moreover, as [22] amply demonstrated, impulsive control allows utilizing small control impulses as much as possible to stabilize a type of chaotic system. Not merely does it reduce redundant information transmission, but it increases the robustness of disturbances rejection. On the other hand, event-based control, as the name implies, is the strategy that event is triggered by some elaborate state-based or output-based event conditions to update the control input, which compared to the conventional time-triggered control is capable of avoiding unnecessary communication since a system adjusts the sampling rate adaptively according to the current situation [23, 24]. Specifically, the issue of self-learning optimal supervision on discrete systems via event-driven formulation is investigated in [24]. And the critics learning standard is improved for the design of nonlinear state-feedback control based on events [25]. Distinct from the extant achievements involving sectionally continuous or consecutive control inputs, event-triggered impulsive control is able to dramatically minimize communication load and communication cost as well as enhance robustness, which, for these reasons, is deserving increasing attention. Simultaneously, the integration of two control strategies also creates tremendous challenges in designing appropriate controller.

As yet, some (but few) significant accomplishments about the event-triggered impulsive control such as [26â€“29] have been reported. Taking [26] as an instance, the synchronization issue on multiple neural networks with disconnected switching topology and delay under this control strategy is studied. Nevertheless, systems are generally affected by some uncertain factors such as human error, random disturbance, information loss, inherent deviation, or environmental noise. The uncertainty caused by these factors is referred to as the parametric uncertainty that is perhaps foremost provenance of model uncertainty [30]. Without taking model uncertainty into account, it seems to be far-fetched and preposterous in reality for analyzing performances of various systems like estimating the property indexes on steady state. In this condition, none of the before-mentioned results are valid. Besides, in the control engineering application, when it comes to the fact that the system states may not be fully available because of implementation costs or physical restrictions, it becomes crucial and inevitable to formulate the event-triggered impulsive control strategy according to practical observer measurements. At this juncture, once the incomplete testability of states and the uncertainty of parameters are incorporated into the characterization of nonlinear systems, then these uncertainties may give rise to a totally new rule with more uncertain antecedents and results. What is exhilarating is that there is no work on the observer-based event-triggered impulsive control strategy to achieve the input-to-state property of uncertain nonlinear systems. After all, it is of more difficulty to find a feasible analytical framework compared with the nominal nonlinear systems. Moreover, in comparison with the previous methods, the robust handling for uncertain parameters during the course of system performance implementation becomes increasingly tricky as the number of uncertain parameters surges. Therefore, the theoretical challenges and technical deficiencies urge us to explore the actual performance evaluation for nonlinear systems with parametric uncertainties under observer-based control.

The abovementioned analysis motivates us to focus on issues of both input-to-state stability and event-triggered impulsive control scheme design on a type of uncertain nonlinear systems with incomplete measurable state variables and exogenous disturbances in this paper. Firstly, we establish a category of newly uncertain nonlinear systems, where the uncertainty terms are legitimately estimated by capitalizing on a creationary approximation condition of uncertainties, matrix synthesis method, and some inequalities of linear matrix. Secondly, a novel observer is constructed on the uncertain nonlinear system, in which the information between plant and observer is transmitted as impulses. In particular, the impulsive controllers are dependent upon the partial measurement output of observer and plant, which can eliminate the adverse effects of output data loss attributed to the external environment. Thirdly, an applicable observer-based event-triggered mechanism is designed and an event-triggered impulsive control strategy is correspondingly constructed, which could lessen burden of sampling and information transmission. At last, several sufficient criteria on excluding the Zeno behavior and analyzing the input-to-state stability property are developed, meanwhile, which suggest that a more extensible framework in complex dynamics can be explored through taking full advantage of a range of the employed ideas and methods.

The content of the remaining sections is summarized as follows. Section 2 puts forward the model and preparatory works for a kind of uncertain nonlinear systems. Section 3 furnishes primary research results. In addition, Section 4 corroborates the validity of the derived results by two numerical simulations. Finally, conclusion is exhibited in Section 5.

#### 2. Preliminaries and Model Description

##### 2.1. Notations

Throughout this article, , , and are separately the set of all real matrices and dimensional Euclidean space and the set of positive integers. stands for an identity matrix with matched dimensionality in matrices or matrix inequalities. 0 in matrices is a zero matrix of appropriate dimensions. Let and denote the 2-norm of matrix and the supremum of on the interval , respectively. For a matrix , , , , and represent severally its inverse, transposition, maximum eigenvalue, and minimum eigenvalue. The symbol is defined as the symmetric term in a matrix. and represent the maximum and minimum of and , respectively. and mean that are symmetric positive definite and symmetric negative definite separately. Let , , and is strictly increasing in , and , for each fixed , belongs to the function of class as regards , but , for each fixed , is strictly decreasing to 0 as .

##### 2.2. Some Preliminaries and Problem Formulation

A class of uncertain nonlinear systems incorporated exogenous disturbances is of the following form:in which , , , and are the system state, the measurement output, and measurable locally bounded exogenous disturbances, respectively; means the control input in which is the sample control input and is the Dirac delta control input; a nonlinear vector-valued function satisfies some conditions that will be provided in the sequel, and represents the right-hand derivative of . , , , , and are constant matrices, and , , , , and are the norm-bounded uncertain parameters.

Given that the incompletely procurable system states can generate the ineffectiveness of state-feedback controllers, an observer-based controller is considered in this paper, and the state observer for uncertain system (1) is constructed bywhere ; ; and are separately the estimated state and the estimated output. The control input of observer is described aswhere ; and are control gains; is the Dirac delta function, which is also called the impulsive control function. And the impulsive time sequence satisfies and . It is well-known that the Dirac delta function has two properties: for any constants and and function , (1) only when ; (2) . Then, by virtue of (2) and (3) and the properties of function , what we can see is that at and , and for any constant that is small enough,where , which can be regarded as the convolution in the interval based on the properties of function , represents the sum of the effects of all unit impulses on the observer state over .

Let and ; then we can infer that

By means of the above calculation, the controller with function can make the observer state change instantaneously in the discrete time sequence so as to achieve the impulsive effect. Thus, observer (2) is converted into an impulsive control system as follows:where the left-continuous case of the estimate of is always considered; that is, .

According to the aforementioned observer, the control inputs and are devised aswhere is the gain of . In a similar way, uncertain system (1) can be transformed into

Without loss of generality, we always assume that the state of system (8) is left-continuous. For forward complete impulsive systems, there is no fundamental difference between utilizing left-continuous model and employing right-continuous model. Then by defining the tracking error , the error system can be expressed as

*Remark 1. *Since uncertain system (1) is a category of impulsive systems, coupled with the incomplete measurability of the system state, we need to construct an appropriate observer and subsequently establish an applicable error system related to plant and observer. Based on the prerequisite of ensuring real-time monitoring, fault-tolerant control, easy realization, and so on, as a result, it is necessary and natural to construct observer (2) which is only influenced by impulsive. One more point needs noting that the controllers and designed by us can exert positive effects on the unstable systems and meanwhile control them only at the impulsive instant. In this way, and can stabilize systems (1) and (2), respectively, for ages, while reducing unnecessary computing costs.

Letand then the argument system can be deduced aswhereFurthermore, , argument system (11) is rewritten aswhere ;An adaptive event-triggered mechanism, determining the continuously updated controller works at the instants known as the triggered time sequence, is introduced to decrease the burden of updating and communication in control. It is notable that the system states are imperfectly accessible, so the event-triggered mechanism included exogenous disturbances as well as the system and observer output is designed. By definingthe event-triggered mechanism is formulated aswhere the event generator function ; parameters , , and in which is a forced triggered constant. Denote by the event-triggered time sequence that it is determined by function . For , the next event will be triggered only when the correlative measurement reaches or surpasses the stated threshold, and then, the next triggered instant (impulsive instant) will be generated by comparing the obtained event-triggered time with the forced triggered time. In addition, it is worth mentioning that the aforesaid two sequences may differ depending on the selected parameters , , and .

In what follows, two assumptions are proposed around the uncertain terms and the nonlinear function.

*Assumption 1. *in which is the unknown time-varying matrix with , the adjustment coefficient of the uncertain term , and , , , , and are constant matrices with compatible dimensionality.

*Remark 2. *Different from existing achievements, such as [21, 30, 31], this paper has more uncertain parameters. Note that, in the practical application, each program will inevitably be subjected to the actual limitation of imprecise modeling for controlled plant and affected by external factors like environmental noise. Therefore, it is favorable and urgent for increasing the number of uncertainties to describe a larger range nonlinear system.

*Remark 3. *Only the norm-bounded uncertainties are taken into account in this article to efficaciously avoid needlessly intricate notations and restrain parameter uncertainty. In accordance with Assumption 1 and several linear matrix inequalities, the uncertainties , , , , and can be reasonably eliminated. Moreover, compared with the conventional constraint conditions of uncertain terms, the adjustment coefficient is added in this paper, which not merely does not change the norm value range of the uncertain terms but also can ingeniously resolve the input-to-state stability problem of fairly sophisticated system. Even though the uncertainty parameters are also present at other singular structures, the subsequent results could be popularized to this circumstance in parallel.

*Assumption 2. *Suppose that there exists a scalar such that the nonlinearity satisfies , . Particularly, .

Hereafter, a definition and several lemmas are introduced for latter use.

*Definition 1. *For every initial condition and each measurable locally bounded exogenous disturbance (see [1]), system (13) is said to be input-to-state stabilizable under the given event-triggered mechanism (16) if there exist functions and such that the solution satisfies

Lemma 1. *Given constant matrices , , and with suitable dimensionality and a matrix function (see [23, 32]),*(1)* and , then**(2)** such that and , then*

Particularly, when , we obtain

Lemma 2. * (see [33]), the inequalityholds, where is a positive definite matrix.*

Lemma 3. *Given constant matrices (see [34]), where and , thenif and only if*

#### 3. Main Results

This section is devoted to the following tripartite through theoretical analysis and demonstrates the following:â€‰(T1) The presence of the lower bound of adjacent impulse instants is testified, whereafter, the Zeno behavior can be excluded.â€‰(T2) The resultant augmented system which is equipped with parameter uncertainties and exogenous disturbances is input-to-state stabilizable where the uncertainties are tactfully subdued.â€‰(T3) The control gains and event-triggered scheme parameters are devised without strong constrained condition under system (13) corresponding stability.

Before verifying the above statements, it is necessary to introduce some symbols:

According to Assumption 1, we havewhere

Theorem 1. *Under event-triggered scheme (16), then (T1) holds, where the positive lower bound of adjoining impulsive moments conforms to , constants , , and are specified in (16), and*

*Proof. * For the sake of checking on (T1), we consider the following three cases:â€‰*Case (i).* The generation of the triggered time sequence depends entirely on the event-triggered time sequence . Based on , in this case, the upper right Dini derivative of in the interval is calculated asâ€‰Let ; then,â€‰Using to premultiplication and postmultiplication in (30) results inâ€‰By considering and integrating both sides of (31) from to , we haveâ€‰When mechanism (16) is triggered, we can derive . Together with , it follows thatâ€‰which indicates thatâ€‰*Case (ii).* Only the forced triggered time sequence exists in the sequence . Obviously, in this case, .â€‰*Case (iii).* The sequence is composed of the event-triggered instants and the forced triggered instants . If the Zeno behavior lives in argument system (13), it must be that a finite time interval owns infinite impulse jumps in this case. To exhibit this phenomenon, suppose that presents the accumulation time (or Zeno time) on the finite time interval . By defining , it is apparent that countless of impulsive instants appears in the interval . Denote by the subsequence of satisfying as , where integer . If there is no forced triggered moment in , similar to the discussion of Case (i), we conclude that as , which contradicts the definition of the accumulation time . If there exists for some over the interval , recalling the definition of , it can be deduced that only one , which implies that the triggered moments totally consist of the event-triggered moments in . Then it follows from Case (i) that as . Hence, the Zeno behavior is precluded in Case (iii).From the foregoing discussion, the lower bound of neighboring impulsive instants is ultimately acquired, which symbolizes that the Zeno behavior can be eliminated.

*Remark 4. *When designing event-triggered mechanism (16), cannot be superseded by . It is on account of a discovery that the correlation over and is not easily obtained in the abovementioned reasoning process of excluding the Zeno phenomenon. Therefore, it is essential for the device of (16) to take the supremum of in . In addition, the study in [29] assists us in detecting the indispensability of the forced trigger condition in (16).

*Remark 5. *As described earlier, mechanism (16) is presented to select the optimal triggered moment according to the steady state of resulting system (13). Based on the results of Theorem 1, it can be proved that the designed event-triggered mechanism is effective. By comparison with the existing results of the uncertain models, such as [21, 35, 36], although they reduce the transmission of information, impulsive controller with fixed impulsive moments, in design, is still conservative. Now, in this paper, the controller based on event-triggered mechanism only is updated at the triggered moment. By this means, the burden of controller update can be decreased without affecting accurate control. Moreover, the sampling control adopted in this paper is only dependent on the state information of observer at the triggered moment, which can reduce the communication between the equipment under test and the observer.

*Assumption 3. * is a nonsingular matrix, and there exists a constant such that .

Theorem 2. *Uncertain system (13) is input-to-state stable via the event-triggered scheme (16); suppose that for given parameters , , and and the control gains , , and , there exist matrices and and constants , , , and , such thatwhere*

*Proof. * Suppose that is the solution of system (13) with the initial value . Due to the imperfect measurability of the system states, the Lyapunov functional related to the relevant outcomings is considered as . , , the derivative of can be calculated thatIt follows from Lemma 1 and conditions (26) and (35) that