Abstract

This paper concerns network based sliding mode control of linear plants with state measurement delay. The considered plants are subject to unbounded disturbance, but it is assumed that the change of disturbance value between each two subsequent sampling instants is limited. In order to combat the unpredictable disturbance in the environment with state measurement delay, a novel sliding mode controller has been introduced. It utilizes two nominal models of the plant to drive the system state along a desired trajectory and counteract the predicted effect of the past disturbance on the system. It has been proven that applying the new control strategy to the plant confines the system state to a defined band around the sliding hyperplane.

1. Introduction

Continuous-time variable structure control (VSC) has gained a considerable popularity in control engineering community since its introduction in the early 1960s [1, 2]. Its main advantages are computational efficiency and insensitivity with respect to matched disturbance [3]. These attractive properties resulted in a considerable amount of research dedicated to sliding mode control, which is the most widely used method of VSC implementation [4–8]. In mid-1980s, the theory of variable structure control has been extended to discrete-time systems [9, 10], which in turn inspired many significant works related to discrete-time sliding mode control [11–33]. In particular, discrete-time sliding mode control has great significance in digital environments [34, 35], where no continuous states are available.

Since the introduction of relatively low-cost control networks [36–41], the problem of effectively controlling discrete-time systems with time delay [42–46] gained importance, as the time it takes for the sensor data and control signal to travel through the network is often nonnegligible. With this in mind, we will design an effective sliding mode control strategy for discrete-time systems with time delay. As opposed to the control method proposed in our previous paper [47], the new approach will realize trajectory tracking in a networked environment as well as ensuring upper bounded convergence rate to the vicinity of the switching hyperplane. To that end, we will introduce two nominal models of the controlled plant. One of them will be compared to the original plant in order to extract delayed information about disturbance affecting the system. Then, the current effect of the disturbance on the system will be estimated and reduced to zero with a dead-beat controller. The other model will be utilized to drive the system from its initial state to a desired one in finite time and realize trajectory tracking in subsequent time instants.

Typically, in the design process of a sliding mode controller, an upper bound of the disturbance affecting the controlled plant is assumed to be known [14, 16, 21, 47]. However, in this paper we will abandon that assumption in favor of the one proposed in [15]. In other words, we assume that the change of the disturbance value between every two subsequent time instants is bounded by a certain constant , while the value of the disturbance itself does not have to be bounded. Under this assumption, we will design a sliding mode controller that will effectively combat the effect of the disturbance on the system in a networked environment.

The remainder of this paper is organized in the following way. Section 2 states the considered problem, after which the new control strategy is introduced in Section 3. The control law for the considered system is established and the properties of the controlled system are formally proven in the same section. In Section 4, two simulation examples that show the effectiveness of the proposed method are presented and concluding remarks are given in Section 5.

2. Problem Statement

Let us consider a single input single output, linear, time invariant plant: where is the state matrix such that , and are vectors of appropriate dimensions that represent input distribution and system state, respectively, and and are scalars that represent the control signal and disturbance, respectively. We assume that the plant operates in a networked environment, which means that it is subject to network-induced delay . Therefore, the control signal in each time instant is calculated based on the system states up to the moment . We do not assume that the disturbance affecting the system is bounded, but that the change of the disturbance value between every two subsequent time instants is limited by . In other words, for any .

3. Proposed Control Strategy

We begin by introducing two models of the considered system (1). The first onewill be used to extract information about disturbance from the system. The matrix and vector in the model are the same as in the original system and . The model is controlled with the same signal as plant (1). Moreover, let us define another modelcontrolled with a different signal . Again, and are the same as in (1) and . The second model will be used solely to drive the system along the desired trajectory . We take into account the sliding hyperplane defined aswhere and is such a vector thatIt can be seen that vector would allow us to design a dead-beat controller for the delay-free system. Although the existing delay prevents us from applying such a controller to the system (1), vector is chosen in such a way that allows us to compensate for the effect of past disturbances on the system as well as partially reducing the ones that have not yet been estimated due to the delay. We first define the following sequence of dimensional vectors:that will be used to extract the information about past disturbances from the system. We assume that for any , which gives for . For any , Therefore, at any time instant , we haveWe will now design a control law that combats the effect of the disturbance obtained from (8) on the system and partially reduces the effect of subsequent disturbances that have not yet been explicitly obtained due to the delay . This is possible since the disturbance rate of change is limited. Let us define vectors in the following way:whereIt can be seen from (9) that each vector can be expressed asVectors possess an important property described by Lemma 1.

Lemma 1. For any , the vector . Furthermore, for , the product .

Proof. The proof of this property is given in our previous work [47].

The control laws (10) will be used to reduce each individual disturbance value affecting the system (1) to zero in finite time. This property will be demonstrated in Theorem 2 later in this paper. We will now define the control law that will drive the system from its initial position to a desired trajectory . Let and let us consider the reaching law [15]:where is a natural number chosen by the designer. It can be seen that the reaching law designed in such a way guarantees that the reference sliding variable will be reduced to 0 at the moment and will retain that value in every subsequent step. The control law for model (3) obtained from reaching law (12) can be expressed asWe propose the following control law for system (1):where is defined by (13), are defined by (10), and is defined by (8). Control law designed in such a way will drive the system from its initial position to a desired state in a finite amount of steps, combat the effect of disturbances up to the moment on the sliding variable, and partially reduce the effects of disturbances that occurred after . Additionally, the element will partially counteract the disturbance that occurs at the moment of applying the control law. As a result, the system state will be confined to a certain band around the sliding hyperplane. The width of the band is specified by Theorem 2 formulated below.

Theorem 2. If the control signal for system (1) is defined by (14) and the system is subject to disturbance , which satisfies for any , then there exists a natural number such that for every the system state is confined to a quasi-sliding mode band defined as

Proof. Let . First, we express asWe rewrite each disturbance as , where . We obtainWe now rearrange the elements in (17) in the following way. For every , elements containing multiplied by a different power of are grouped together. Then, in each group, the highest power of is factored out. In this way we getLet us introduce the following notation:We substitute from (9) and from (19) into (18) and obtainThen, we substitute from (14) into (20) and getRelation (8) gives for any . Therefore, from (9), we know that for all and consequently for any . From (10) we know that is a function of , which gives for . Taking that into consideration, we rewrite (21) asIt can be seen from (3) thatMoreover, from (8), we know thatTaking into consideration (23) and (24), we rewrite (22) asWe rearrange the elements of (25) as follows:Using (11), we simplify (26) in the following way:From Lemma 1, we know that for each vector , which gives usThe product can therefore be expressed as Lemma 1 states that, for each and , the product , which results inFrom (12), we getfor , which gives usand consequently Therefore, the bound of the absolute value of can be expressed asWe conclude that (15) holds for any .