Stability Analysis of State Saturation 2D Discrete Time-Delay Systems Based on F-M Model

Abstract

The problem of stability analysis is investigated for a class of state saturation two-dimensional (2D) discrete time-delay systems described by the Fornasini-Marchesini (F-M) model. The delay is allowed to be a bounded time-varying function. By constructing the delay-dependent 2D discrete Lyapunov functional and introducing a nonnegative scalar 𝛽, a sufficient condition is proposed to guarantee the global asymptotic stability of the addressed systems. Subsequently, the criterion is converted into the linear matrix inequalities (LMIs) which can be easily tested by using the standard numerical software. Finally, two numerical examples are given to show the effectiveness of the proposed stability criterion.

1. Introduction

Over the past decades, the two-dimensional (2D) systems have received considerable attention due to their extensive applications [1]. The 2D discrete-time systems have been successfully applied to many practical areas such as signal processing, linear image processing, multidimensional digital filtering, seismographic data processing, water stream heating and thermal processes. To mention a few, the problem of linear image processing has been studied in [2] for 2D discrete systems based on the Roesser model (R model). Aiming at the design of the digital filter, the 2D discrete systems based on the Fornasini-Marchesini model (F-M model) have been discussed in [3]. Recently, the modeling, calculation, and stability of 2D discrete systems have been widely studied and a large amount of results on these topics have been published; see, for example, [4–9].

Saturation, as a common and typical nonlinear constraint for practical control systems, is often encountered in various industrial systems. The phenomena of saturations if not properly handled will inevitably affect the implementations of the designed control schemes and may lead to the occurrence of the zero-input limit cycles [10–13]. Dynamical systems with saturation nonlinearities appear commonly in networks and 2D digital filter [14, 15]. When designing the 2D digital filter by using fixed point arithmetic, saturations are introduced due to the overflow and quantization. Recently, many important results on stability analysis of 2D discrete F-M systems with state saturation have been reported in recent literature; see, for example, [16–21]. To be specific, by using the Lyapunov method and employing the property of matrix norm, some sufficient conditions have been presented in [16, 17] to guarantee the global asymptotic stability of the related systems. In [18], an internally stable condition has been given for the design of 2D filters. By introducing a nonnegative scalar, in [19, 20], the criteria based on linear matrix inequality (LMI) have been proposed to ensure the global asymptotic stability of F-M model. Recently, sufficient conditions have been derived in [21] to guarantee the global asymptotic stability of the addressed F-M systems with state saturation nonlinearities.

As well known, the time delays occur in many practical engineering systems [22–24]. The occurrence of the time delays would yield the instability of the controlled systems in some cases. So far, a great number of results have been reported; see, for example, [25–29]. In particular, considerable research attention has been devoted to the problems of stability analysis of 2D discrete time-delay systems. To be specific, by constructing an appropriate Lyapunov function and a scalar 𝛽, sufficient conditions have been proposed in [26] for state saturation 2D discrete time-delay systems based on the R model. In [27], the state estimation problem has been investigated for 2D complex networks with randomly occurring nonlinearities and probabilistic sensor delays. Accordingly, some sufficient conditions have been established such that the resulted estimation error dynamics are globally asymptotically stable in the mean square sense. To be specific, the problems of global asymptotic stability have been given in [30, 31] for 2D discrete F-M systems with state saturation and constant time delays. However, to the best of author’s knowledge, there has not been much work undertaken on the global asymptotic stability for state saturation 2D discrete time-varying delay systems based on F-M model.

In this paper, we aim to investigate the problem of global asymptotic stability for 2D discrete F-M systems with state saturation and time delays. Here, the delays are assumed to be time varying with known lower and upper bounds. By constructing a delay-dependent 2D discrete Lyapunov functional and introducing a nonnegative scalar 𝛽 based on a row diagonally dominant matrix, a sufficient condition is proposed to guarantee the global asymptotic stability of the addressed systems. Subsequently, the problem of global asymptotic stability is converted into the problem of feasibility by solving the LMIs. Finally, two numerical examples are given to show the effectiveness of the proposed criterion. The main contribution of this paper lies in that new stability criterion is given for state saturation 2D discrete F-M systems with time-varying delays.

Notations. The fundamental notations used in the paper are given as follows: 𝐼_𝑝 and 0 denote the identity matrix and zero matrix with appropriate dimensions. For a matrix 𝐴, 𝐴^𝑇 stands for transpose of the matrix 𝐴. 𝐴>0 means that 𝐴 is positive definite symmetric matrix.

2. Problem Formulation and Preliminaries

In this paper, we consider the following 2D discrete F-M system with state saturation and time-varying delays: 𝑥()𝑦(),𝑦()=𝑦𝑘+1,𝑙+1=𝑓𝑘,𝑙𝑘,𝑙₁(𝑘,𝑙),𝑦₂(𝑘,𝑙),…,𝑦_𝑛(𝑘,𝑙)^𝑇𝑥()𝑥()=𝐴𝑘,𝑙+1𝑘+1,𝑙+𝐴_𝑑𝑥𝑘−𝑑_ℎ(𝑘)𝑥,𝑙+1𝑘+1,𝑙−𝑑_𝑣(𝑙)=𝐴₁₁𝑥(𝑘,𝑙+1)+𝐴₁₂𝑥(𝑘+1,𝑙)+𝐴₂₁𝑥𝑘−𝑑_ℎ(𝑘),𝑙+1+𝐴₂₂𝑥𝑘+1,𝑙−𝑑_𝑣(𝑙), (1) where 𝑥(𝑘,𝑙) is the 𝑛-dimensional state vector and (𝑘,𝑙) is the 2-dimensional time variable along the vertical and horizontal directions satisfying 𝑘≥0,𝑙≥0. 𝐴=[𝐴₁₁,𝐴₁₂]∈𝑅^𝑛×2𝑛, 𝐴_𝑑=[𝐴₂₁,𝐴₂₂]∈𝑅^𝑛×2𝑛, 𝑑_ℎ(𝑘) and 𝑑_𝑣(𝑙) are time-varying delays along the vertical and horizontal directions, respectively. We assume 𝑑_ℎ(𝑘) and 𝑑_𝑣(𝑙) satisfying: ℎ_𝑚≤𝑑_ℎ(𝑘)≤ℎ_𝑀,𝑣_𝑚≤𝑑_𝑣(𝑙)≤𝑣_𝑀, (2) where ℎ_𝑚 and ℎ_𝑀 denote the lower and upper bounds of the delay 𝑑_ℎ(𝑘) along the horizontal directions, 𝑣_𝑚 and 𝑣_𝑀 denote the lower and upper bounds of the delay 𝑑_𝑣(𝑙) along the vertical directions. Throughout this paper, the superscript 𝑇 to any vector (or matrix) stands for the transpose of that vector (or matrix).

The saturation nonlinearity 𝑓(⋅) is given by 𝑓𝑦(𝑘,𝑙)=𝑓₁𝑦₁(𝑘,𝑙),𝑓₂𝑦₂(𝑘,𝑙),…,𝑓_𝑛𝑦_𝑛(𝑘,𝑙)^𝑇, (3) with 𝑓_𝑖𝑦_𝑖(𝑘,𝑙)=1,𝑦_𝑖()𝑦𝑘,𝑙>1,_𝑖(𝑘,𝑙),𝑦_𝑖(𝑘,𝑙)≤1,−1,𝑦_𝑖()()𝑘,𝑙<−1,𝑖=1,2,…,𝑛. (4)

The system (1) has finite initial conditions 𝑥(𝑘,0) and 𝑥(0,𝑙) with fixed yet nonzero values for −ℎ_𝑀≤𝑘≤𝐾 and −𝑣_𝑀≤𝑙≤𝐿, and 𝑥(𝑘,0)=0,𝑘≥𝐾;𝑥(0,𝑙)=0,𝑙≥𝐿, (5) where 𝐾 and 𝐿 are positive integers.

To proceed, we introduce the following definitions which will be used in the subsequent derivations.

Definition 1. The origin 𝑥=0 of the 2D discrete time-delay system is said to be stable (in the sense of Lyapunov) if for every 𝜀>0 there exists a 𝛿=𝛿(𝜀)>0 such that ‖𝑥(𝑘,𝑙)‖<𝜀, (6) for all 𝑘≥0,  𝑙≥0, whenever ‖𝑥(𝑘,0)‖<𝛿(−ℎ_𝑀≤𝑘≤𝐾) and ‖𝑥(0,𝑙)‖<𝛿(−𝑣_𝑀≤𝑙≤𝐿), where 𝐾 and 𝐿 are positive integers and ‖⋅‖ denotes any of the equivalent norms on 𝑅^𝑛.

Definition 2 (see [21]). The origin 𝑥=0 of the 2D system (1) is said to be globally asymptotically stable if the following conditions are simultaneously satisfied:(1)system (1) is stable;(2)every solution of system (1) tends to the origin as 𝑘+𝑙→∞; that is, lim_{𝑘→∞or𝑙→∞}𝑥(𝑘,𝑙)=lim_{𝑘+𝑙→∞}𝑥(𝑘,𝑙)=0. (7)

Definition 3 (see [32]). The matrix 𝑀 is said to be row diagonally dominant if 𝑚_𝑖𝑖≥^𝑛_{𝑗=1,𝑗≠𝑖}𝑚_𝑖𝑗,𝑖∈[1,𝑛]. (8)

Definition 4. For a 2D function 𝑉(𝑘,𝑙)=𝑉^ℎ(𝑘,𝑙)+𝑉^𝑣(𝑘,𝑙), denote its difference Δ𝑉(𝑘,𝑙) as Δ𝑉(𝑘,𝑙)=Δ𝑉^ℎ(𝑘,𝑙)+Δ𝑉^𝑣(𝑘,𝑙), (9) where Δ𝑉^ℎ(𝑘,𝑙)=𝑉^ℎ(𝑘+1,𝑙+1)−𝑉^ℎ(𝑘,𝑙+1),Δ𝑉^𝑣(𝑘,𝑙)=𝑉^𝑣(𝑘+1,𝑙+1)−𝑉^𝑣(𝑘+1,𝑙). (10)

3. Main Results

In this section, we aim to investigate the problem of stability analysis for the 2D discrete F-M system with state saturation and time-varying delays. By resorting to the LMI technique, a sufficient condition is given to ensure the global asymptotic stability of the addressed system.

Theorem 5. The origin 𝑥=0 of 2D discrete system (1) is globally asymptotically stable if there exist matrices 𝑃,𝑄, and 𝑀, where 𝑃=𝑃₁+𝑃₂ and 𝑄=𝑄₁+𝑄₂ with 𝑃₁=[𝑝¹_𝑖𝑗]∈𝑅^𝑛×𝑛, 𝑃₂=[𝑝²_𝑖𝑗]∈𝑅^𝑛×𝑛, 𝑄₁=[𝑞¹_𝑖𝑗]∈𝑅^𝑛×𝑛, and 𝑄₂=[𝑞²_𝑖𝑗]∈𝑅^𝑛×𝑛 being symmetric positive definite matrices and 𝑀=[𝑚_𝑖𝑗]∈𝑅^𝑛×𝑛 is row diagonally dominant with nonnegative diagonal elements, satisfying the following matrix inequality: Π₁𝑀^𝑇𝐴₁₁𝑀^𝑇𝐴₁₂𝑀^𝑇𝐴₂₁𝑀^𝑇𝐴₂₂∗Π₂000∗∗Π₃00∗∗∗−𝑄₁0∗∗∗∗−𝑄₂<0, (11) with Π₁=𝑃₁+𝑃₂−𝑀+𝑀^𝑇,Π₂=𝑄₁−𝑃₁+ℎ_𝑀−ℎ_𝑚𝑄₁,Π₃=𝑄₂−𝑃₂+𝑣_𝑀−𝑣_𝑚𝑄₂. (12)

Proof. In order to prove the global asymptotic stability of the 2D discrete F-M system (1), we construct the following Lyapunov functional as in [26]: 𝑉(𝑘,𝑙)=³_𝑖=1𝑉_𝑖(𝑘,𝑙), (13) where 𝑉₁(𝑘,𝑙)=𝑉^ℎ₁(𝑘,𝑙)+𝑉^𝑣₁(𝑘,𝑙)=𝑥^𝑇(𝑘,𝑙)𝑃₁𝑥(𝑘,𝑙)+𝑥^𝑇(𝑘,𝑙)𝑃₂𝑥(),𝑉𝑘,𝑙₂(𝑘,𝑙)=𝑉^ℎ₂(𝑘,𝑙)+𝑉^𝑣₂()=𝑘,𝑙_{𝑘−1𝑖=𝑘−𝑑ℎ(𝑘)}𝑥^𝑇(𝑖,𝑙)𝑄₁𝑥()+𝑖,𝑙_{𝑙−1𝑗=𝑙−𝑑𝑣(𝑙)}𝑥^𝑇𝑘,𝑗𝑄₂𝑥,𝑉𝑘,𝑗₃(𝑘,𝑙)=𝑉^ℎ₃(𝑘,𝑙)+𝑉^𝑣₃()=𝑘,𝑙_{−ℎ𝑚𝑗=−ℎ𝑀+1𝑘−1𝑖=𝑘+𝑗}𝑥^𝑇(𝑖,𝑙)𝑄₁𝑥()+𝑖,𝑙_{−𝑣𝑚𝑖=−𝑣𝑀+1𝑙−1𝑗=𝑙+𝑖}𝑥^𝑇𝑘,𝑗𝑄₂𝑥𝑘,𝑗 (14) with 𝑃₁>0, 𝑃₂>0, 𝑄₁>0, and 𝑄₂>0 being matrices to be determined.
According to Definition 4, the corresponding difference of Δ𝑉(𝑘,𝑙) along the trajectory of system (1) can be calculated as follows: Δ𝑉(𝑘,𝑙)=³_𝑖=1Δ𝑉_𝑖(𝑘,𝑙), (15) with Δ𝑉₁(𝑘,𝑙)=𝑉₁(𝑘+1,𝑙+1)−𝑥^𝑇(𝑘,𝑙+1)𝑃₁𝑥(𝑘,𝑙+1)−𝑥^𝑇(𝑘+1,𝑙)𝑃₂𝑥(𝑘+1,𝑙),Δ𝑉₂(𝑘,𝑙)=^𝑘_{𝑖=𝑘+1−𝑑ℎ()𝑘+1}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()−𝑖,𝑙+1_{𝑘−1𝑖=𝑘−𝑑ℎ(𝑘)}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()+𝑖,𝑙+1_{𝑙𝑗=𝑙+1−𝑑𝑣()𝑙+1}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥−𝑘+1,𝑗_{𝑙−1𝑗=𝑙−𝑑𝑣(𝑙)}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥𝑘+1,𝑗,Δ𝑉₃(𝑘,𝑙)=^−ℎ𝑚_{𝑗=−ℎ𝑀}^𝑘_{+1𝑖=𝑘+𝑗+1}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()−𝑖,𝑙+1_{−ℎ𝑚𝑗=−ℎ𝑀+1𝑘−1𝑖=𝑘+𝑗}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()+𝑖,𝑙+1_{−𝑣𝑚𝑖=−𝑣𝑀𝑙+1𝑗=𝑙+𝑖+1}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥−𝑘+1,𝑗_{−𝑣𝑚𝑖=−𝑣𝑀+1𝑙−1𝑗=𝑙+𝑖}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥𝑘+1,𝑗. (16) Noting (13)–(15), it can be derived that Δ𝑉₁(𝑘,𝑙)=𝑥^𝑇(𝑘+1,𝑙+1)𝑃₁𝑥(𝑘+1,𝑙+1)+𝑥^𝑇(𝑘+1,𝑙+1)𝑃₂𝑥(𝑘+1,𝑙+1)−𝑥^𝑇(𝑘,𝑙+1)𝑃₁𝑥(𝑘,𝑙+1)−𝑥^𝑇(𝑘+1,𝑙)𝑃₂𝑥(𝑘+1,𝑙),Δ𝑉₂(𝑘,𝑙)=𝑥^𝑇(𝑘,𝑙+1)𝑄₁𝑥(𝑘,𝑙+1)−𝑥^𝑇𝑘−𝑑_ℎ(𝑘),𝑙+1𝑄₁𝑥𝑘−𝑑_ℎ(𝑘)+,𝑙+1_{𝑘−1𝑖=𝑘+1−𝑑ℎ()𝑘+1}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()−𝑖,𝑙+1_{𝑘−1𝑖=𝑘+1−𝑑ℎ(𝑘)}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥(𝑖,𝑙+1)+𝑥^𝑇(𝑘+1,𝑙)𝑄₂𝑥(𝑘+1,𝑙)−𝑥^𝑇𝑘+1,𝑙−𝑑_𝑣(𝑙)𝑄₂𝑥𝑘+1,𝑙−𝑑_𝑣(𝑙+_{𝑙−1𝑗=𝑙+1−𝑑𝑣()𝑙+1})𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥−𝑘+1,𝑗_{𝑙−1𝑗=𝑙+1−𝑑𝑣(𝑙)}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥≤𝑘+1,𝑗_{𝑘−1𝑖=𝑘+1−ℎ𝑀}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()−𝑖,𝑙+1_{𝑘−1𝑖=𝑘+1−ℎ𝑚}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()+𝑖,𝑙+1_{𝑙−1𝑗=𝑙+1−𝑣𝑀}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥−𝑘+1,𝑗_{𝑙−1𝑗=𝑙+1−𝑣𝑚}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥𝑘+1,𝑗+𝑥^𝑇(𝑘,𝑙+1)𝑄₁𝑥(𝑘,𝑙+1)−𝑥^𝑇𝑘−𝑑_ℎ(𝑘),𝑙+1𝑄₁𝑥𝑘−𝑑_ℎ(𝑘),𝑙+1+𝑥^𝑇(𝑘+1,𝑙)𝑄₂𝑥(𝑘+1,𝑙)−𝑥^𝑇𝑘+1,𝑙−𝑑_𝑣(𝑙)𝑄₂𝑥𝑘+1,𝑙−𝑑_𝑣(𝑙)=^{𝑘−ℎ𝑚}_{𝑖=𝑘+1−ℎ𝑀}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()+𝑖,𝑙+1_{𝑙−𝑣𝑚𝑗=𝑙+1−𝑣𝑀}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥𝑘+1,𝑗+𝑥^𝑇(𝑘,𝑙+1)𝑄₁𝑥(𝑘,𝑙+1)−𝑥^𝑇𝑘−𝑑_ℎ(𝑘),𝑙+1𝑄₁𝑥𝑘−𝑑_ℎ(𝑘),𝑙+1+𝑥^𝑇(𝑘+1,𝑙)𝑄₂𝑥(𝑘+1,𝑙)−𝑥^𝑇𝑘+1,𝑙−𝑑_𝑣(𝑙)𝑄₂𝑥𝑘+1,𝑙−𝑑_𝑣(𝑙)Δ𝑉₃(𝑘,𝑙)=ℎ_𝑀−ℎ_𝑚𝑥^𝑇(𝑘,𝑙+1)𝑄₁𝑥()−𝑘,𝑙+1_{𝑘−ℎ𝑚𝑖=𝑘+1−ℎ𝑀}𝑥^𝑇(𝑖,𝑙+1)𝑄₁𝑥()+𝑣𝑖,𝑙+1_𝑀−𝑣_𝑚𝑥^𝑇(𝑘+1,𝑙)𝑄₂𝑥()−𝑘+1,𝑙_{𝑙−𝑣𝑚𝑗=𝑙+1−𝑣𝑀}𝑥^𝑇𝑘+1,𝑗𝑄₂𝑥𝑘+1,𝑗. (17) Denoting 𝜑(𝑘,𝑙)=𝑓𝑦(𝑘,𝑙),𝑥(𝑘,𝑙+1),𝑥(𝑘+1,𝑙),𝑥(𝑘−𝑑_ℎ(𝑘),𝑙+1),𝑥𝑘+1,𝑙−𝑑_𝑣(𝑙)^𝑇, (18) we have Δ𝑉(𝑘,𝑙)≤𝜑^𝑇(𝑘,𝑙)Θ𝜑(𝑘,𝑙), (19) where𝑃Θ=¹+𝑃₂∗ℎ0000_𝑀−ℎ_𝑚+1𝑄₁−𝑃₁𝑣000∗∗_𝑀−𝑣_𝑚+1𝑄₂−𝑃₂00∗∗∗−𝑄₁0∗∗∗∗−𝑄₂. (20)
Now, construct the following parameter 𝛽: 𝛽=2^𝑛_𝑖=1𝑦_𝑖(𝑘,𝑙)−𝑓_𝑖𝑦_𝑖(×𝑚𝑘,𝑙)_𝑖𝑖𝑓_𝑖𝑦_𝑖(𝑘,𝑙)+^𝑛_{𝑗=1,𝑗≠𝑖}𝑚_𝑖𝑗𝑓_𝑗𝑦_𝑗(𝑘,𝑙)=𝑦^𝑇(𝑘,𝑙)𝑀𝑓𝑦(𝑘,𝑙)+𝑓^𝑇𝑦(𝑘,𝑙)𝑀^𝑇𝑦(𝑘,𝑙)−𝑓^𝑇𝑦(𝑘,𝑙)𝑀+𝑀^𝑇𝑓𝑦(𝑘,𝑙). (21) Let 𝛼_𝑖=𝑦_𝑖(𝑘,𝑙)−𝑓_𝑖𝑦_𝑖()×𝑚𝑘,𝑙_𝑖𝑖𝑓_𝑖𝑦_𝑖(𝑘,𝑙)+^𝑛_{𝑗=1,𝑗≠𝑖}𝑚_𝑖𝑗𝑓_𝑗𝑦_𝑗(𝑘,𝑙). (22) The saturation region can be divided into two regions, and we have 𝑓_𝑖𝑦_𝑖()𝑦𝑘,𝑙=1,_𝑖(𝑘,𝑙)−𝑓_𝑖𝑦_𝑖()𝑓𝑘,𝑙>0,inRegion1,_𝑖𝑦_𝑖()𝑦𝑘,𝑙=−1,_𝑖(𝑘,𝑙)−𝑓_𝑖𝑦_𝑖(𝑘,𝑙)<0,inRegion2. (23) Thus, we obtain 𝛼_𝑖=𝑦_𝑖(𝑘,𝑙)−1𝑚_𝑖𝑖+^𝑛_{𝑗=1,𝑗≠𝑖}𝑚_𝑖𝑗𝑓_𝑗𝑦_𝑗(),𝑦𝑘,𝑙inRegion1,_𝑖(𝑘,𝑙)+1−𝑚_𝑖𝑖+^𝑛_{𝑗=1,𝑗≠𝑖}𝑚_𝑖𝑗𝑓_𝑗𝑦_𝑗(𝑘,𝑙),inRegion2. (24)
Subsequently, due to the property of row diagonally dominant matrix 𝑀, we have 𝛼_𝑖≥0. Furthermore, note that 𝛽 is the sum of nonnegative scalars, and hence 𝛽≥0. Therefore, Δ𝑉≤𝜑^𝑇(𝑘,𝑙)𝑊𝜑(𝑘,𝑙)−𝛽, (25) where Π𝑊=¹𝑀^𝑇𝐴₁₁𝑀^𝑇𝐴₁₂𝑀^𝑇𝐴₂₁𝑀^𝑇𝐴₂₂∗Π₂000∗∗Π₃00∗∗∗−𝑄₁0∗∗∗∗−𝑄₂. (26) If 𝑊<0, then Δ𝑉(𝑘,𝑙)<0; that is 𝑉^ℎ(𝑘+1,𝑙+1)+𝑉^𝑣(𝑘+1,𝑙+1)≤𝑉^ℎ(𝑘,𝑙+1)+𝑉^𝑣(𝑘+1,𝑙). (27) Hence, for any nonnegative integer 𝑑≥max{𝐾,𝐿}, we have _{𝑘+𝑙=𝑑+1}𝑉(𝑘,𝑙)=𝑉(1,𝑑)+𝑉(2,𝑑−1)+⋯+𝑉(𝑑,1)≤𝑉^ℎ(0,𝑑)+𝑉^𝑣(1,𝑑−1)+𝑉^ℎ(1,𝑑−1)+𝑉^𝑣(2,𝑑−2)+⋯+𝑉^ℎ(𝑑−1,1)+𝑉^𝑣()=𝑑,0_{𝑘+𝑙=𝑑}𝑉(𝑘,𝑙). (28) It is clear that the sum of the Lyapunov functional is a decreasing function along the state trajectories of system (1). Then, noting the initial condition 𝑥(0,𝑑)=𝑥(𝑑,0)=0, we have lim_{𝑘→∞or𝑙→∞}𝑥(𝑘,𝑙)=lim_{𝑘+𝑙→∞}𝑥(𝑘,𝑙)=0. (29) Summarizing the above discussions, it can be shown that (11) is a sufficient condition which ensures the global asymptotic stability of system (1). This completes the proof of Theorem 5.

As special cases, if there is no time delay in system (1), that is, 𝐴₂₁=0 and 𝐴₂₂=0, or the time delays are constant, then we can have the following corollaries.

Corollary 6. Consider the system (1) without time delay. If there exist matrices 𝑃 and 𝑀, with 𝑃=𝑃₁+𝑃₂, 𝑃₁=[𝑝¹_𝑖𝑗]∈𝑅^𝑛×𝑛, and 𝑃₂=[𝑝²_𝑖𝑗]∈𝑅^𝑛×𝑛 being symmetric positive definite matrices and 𝑀=[𝑚_𝑖𝑗]∈𝑅^𝑛×𝑛 is row diagonally dominant with nonnegative diagonal elements, satisfying the following matrix inequality: 𝑃−𝑀+𝑀^𝑇𝑀^𝑇𝐴₁₁𝑀^𝑇𝐴₁₂∗−𝑃₁0∗0−𝑃₂<0 (30) then the origin 𝑥=0 of 2D discrete system (1) without time delay is globally asymptotically stable.

Corollary 7. Consider the constant time delay 𝑑_ℎ(𝑘)=𝑑_ℎ, 𝑑_𝑣(𝑙)=𝑑_𝑣. If there exist matrices 𝑃, 𝑄, and 𝑀, where 𝑃=𝑃₁+𝑃₂ and 𝑄=𝑄₁+𝑄₂, with 𝑃₁=[𝑝¹_𝑖𝑗]∈𝑅^𝑛×𝑛, 𝑃₂=[𝑝²_𝑖𝑗]∈𝑅^𝑛×𝑛, 𝑄₁=[𝑞¹_𝑖𝑗]∈𝑅^𝑛×𝑛, and 𝑄₂=[𝑞²_𝑖𝑗]∈𝑅^𝑛×𝑛 being symmetric positive definite matrices and 𝑀=[𝑚_𝑖𝑗]∈𝑅^𝑛×𝑛 is row diagonally dominant with nonnegative diagonal elements, satisfying the following matrix inequality:𝑃₁+𝑃₂−𝑀+𝑀^𝑇𝑀^𝑇𝐴₁₁𝑀^𝑇𝐴₁₂𝑀^𝑇𝐴₂₁𝑀^𝑇𝐴₂₂∗𝑄₁−𝑃₁000∗∗𝑄₂−𝑃₂00∗∗∗−𝑄₁0∗∗∗∗−𝑄₂<0, (31)then the origin 𝑥=0 of 2D discrete system (1) with constant time delay is globally asymptotically stable.

Remark 8. Note that the problem of global asymptotic stability has been investigated in [33] for state saturation 2D discrete system without time delay. Accordingly, a stability criterion has been derived in [33] but a positive diagonally matrix is needed to be searched. It can be easily seen that the stability condition in [33] is a special case of (30) in Corollary 6. Therefore, the stability condition proposed in this paper is less conservative than the one in [33]. Meanwhile, note that a stability condition has been proposed in [21] where an unknown matrix 𝐺 must be given firstly. The values of 𝐺 must take specific values. Compared with the results in [21], we only need to find matrix 𝑀. It concludes that the stability condition in [21] is more conservative than our result.

Remark 9. It is worth mentioning that the delay-fractioning approach has been employed in [34, 35] to reduce the conservativeness of the time delay. It has been shown that the developed approach performs well when dealing with the time delay compared with other methods. Accordingly, some effective SMC/SMO schemes based on the delay-fractioning idea have been proposed for discrete time-delay nonlinear stochastic systems with randomly occurring incomplete information. Motivated by the results in [34, 35], we are now researching into the stability criterion based on the delay-fractioning approach for the state saturation 2D discrete time-delay systems. The corresponding results will appear in the near future. Moreover, note that the recursive filters have been designed in [36–38] for time-varying networked nonlinear systems with missing measurements. It is also interesting to consider the analysis and synthesis of 2D discrete nonlinear systems with state saturations and missing measurements.

Remark 10. It is worth noting that the conditions in Theorem 5 are not strict LMI due to the fact that the matrix 𝑀=[𝑚_𝑖𝑗]∈𝑅^𝑛×𝑛 is row diagonally dominant with nonnegative diagonal elements. Hence, the results of Theorem 5 and Corollary 6 are not convex which lead to the computational difficulties. In the following, an alternative approach is developed to deal with the nonconvex problem.

Let 𝑒_𝑙 be 𝑛-dimensional column vectors in which the 𝑙th element is 1 and other elements are 0. Let 𝑌_𝑙 be the set of 𝑛-dimensional column vectors in which the 𝑙th element is −1 and other elements are either 1 or −1 and 𝑌_𝑙𝑠∈𝑌_𝑙(𝑠∈[1,2^𝑛−1]). Then, the condition where matrix 𝑀=[𝑚_𝑖𝑗]∈𝑅^𝑛×𝑛 is row diagonally dominant and the diagonal is composed of nonnegative elements can be equivalently converted into the following LMIs: 𝑒^𝑇_𝑙𝑀𝑌_𝑙𝑠<0,𝑙=1,2,…,𝑚+𝑛;𝑠∈1,2^{𝑚+𝑛−1}. (32) Together with (11) and (32), we can see that the proposed stability condition is a convex one and then can be easily solved by using the standard numerical software.

4. Numerical Examples

In this section, two numerical examples are given here to illustrate the effectiveness of the main results.

Example 1. Consider the state saturation 2D discrete time-delay system (1) with 𝐴₁₁=,𝐴1.6780.2853−0.2432−0.1246−0.840.23−0.11.450.40.02170.17850.16620.0245−0.008840.0321−0.2428₁₂=,𝐴−0.495−0.687−0.030−0.3013−0.1630−0.4817−0.9930.18140.28420.1790−0.55510.0799−0.21120.18870.08950.5604₂₁=,𝐴−9.987−0.2000−0.150−0.100−0.780−0.2300−0.98₂₂=.5.763−0.06700.04−2.3140.3420.098−5.784.560.001−0.980.2310.23−0.760.0438.6 (1) The lower and upper bounds are given by ℎ_𝑀=3, ℎ_𝑚=1, 𝑣_𝑀=4, and 𝑣_𝑚=2.
Solving the inequalities (11) and (32) in the Matlab environment, we can obtain 𝑃₁=,𝑃3.0067−0.0360−0.00430.0125−0.03603.14200.16350.2573−0.00430.16352.94430.12530.01250.25730.12532.9694₂=,𝑄2.9848−0.0225−0.00060.0463−0.02253.14940.17570.2382−0.00060.17573.00490.09900.04630.23820.09902.9051₁=,𝑄0.5521−0.0048−0.00170.0417−0.00480.61680.04610.0573−0.00170.04610.56310.03560.04170.05730.03560.4917₂=,.0.5676−0.0144−0.00430.0175−0.01440.61150.03740.0709−0.00430.03740.51980.05450.01750.07090.05450.5376𝑀=0.28120.0241−0.0475−0.00200.03530.68770.06090.2516−0.07690.06540.70740.01280.00270.1886−0.00180.5205 (34) Then, we can see that there exist the required matrices 𝑃₁, 𝑃₂, 𝑄₁, 𝑄₂, and 𝑀 satisfying LMIs (11) and (32). As such, according to Theorem 5, the origin 𝑥=0 of system (1) is globally asymptotically stable which confirms the feasibility of the proposed main results.

Example 2. Consider a 2D system described by (1) without time delay. The system parameters are given as 𝐴₁₁=,𝐴1.2−2.80.10₁₂=.00.0100.02 (35)
By using the Matlab LMI Toolbox, it can be easily verified that the following feasible solutions can be obtained 𝑃₁=1421.7−203.9−203.91698.1,𝑃₂=,.132125251167.5𝑀=434.848−11.3033−9.1992818.5481 (36) That is, there exist the required matrices 𝑃₁, 𝑃₂, and 𝑀 satisfying LMI (30). According to Corollary 6, the origin of system (1) is globally asymptotically stable which confirms the effectiveness of the presented results. However, it can be tested that the conditions in [33] are infeasible. Therefore, the result in our paper is less conservative than the one in [33].

5. Conclusions

In this paper, we have discussed the problem of global asymptotic stability for 2D discrete F-M systems with state saturation and time-varying delays. By constructing the 2D discrete-time Lyapunov functional, a new stability criterion has been established to ensure that the addressed system is globally asymptotically stable. The proposed stability criterion is in terms of the LMIs which can be easily tested by using the Matlab matrix toolbox. Two numerical examples have been given to demonstrate the feasibility of the proposed stability condition. It is worth mentioning that the construction of the scalar 𝛽 has taken full effects from time delays with hope to reduce the conservativeness. One of the future research topics would be the extension of the proposed main results to more general state saturation 2D discrete systems as in [39–41] with network-induced phenomena.

Acknowledgments

This work was supported in part by the Science Research Project of Abroad Scholars of Educational Commission of Heilongjiang Province under Grant 1152hq09 and the National Natural Science Foundation of China under Grants 11271103 and 11301118.