Stabilization Strategies of Supply Networks with Stochastic Switched Topology

Abstract

In this paper, a dynamical supply networks model with stochastic switched topology is presented, in which the stochastic switched topology is dependent on a continuous time Markov process. The goal is to design the state-feedback control strategies to stabilize the dynamical supply networks. Based on Lyapunov stability theory, sufficient conditions for the existence of state feedback control strategies are given in terms of matrix inequalities, which ensure the robust stability of the supply networks at the stationary states and a prescribed 𝐻_∞ disturbance attenuation level with respect to the uncertain demand. A numerical example is given to illustrate the effectiveness of the proposed method.

1. Introduction

Supply network is a complicated dynamical system which is composed of a set of facilities, connected by transportation links transforming the raw materials or resources into intermediate and the finished products into the end consumers. Supply networks model has been intensively studied in the last few years, which is a more realistic model to describe nonlinear interactions and dynamics of the flow of materials through networks [1–4]. Helbing et al. [5] proposed a supply network that is governed by balance equations and equations for the adaptation of production speeds and studied the stability and dynamics of supply networks. The study in [6] examined the stability of multistage supply chains under arbitrary demand conditions and presented commitment-based polices that can maintain any desired inventory level for any demand rate. A supply network based on stochastic discrete-time controlled dynamical system was proposed in [7], in which an explicit state-feedback control policy was derived to control the material flow of the supply network. In [8], the propagation and amplification of order fluctuations in supply chain networks were analyzed, and based on inventory management policies, the robust analytical conditions were proposed to predict the presence of the bullwhip effect for any network structure. The study in [9] developed a model of a general closed-loop supply chain network and formulated and optimized the equilibrium state of the network by using the variational inequalities method. In [10], a supply chain network model with reentrant nodes based on partial differential equation was proposed, which can accurately reflect the impacts of the reentrant degree of the product on the system performance. In [11], a nonlinear complementarity formulation for the supply chain network equilibrium models was established, and the equilibrium state of the network was formulated by using the variational inequalities method.

In practice, the topology structure of the supply networks will change with the time due to interconnections changing between the suppliers, which will lead to the instability of the supply network. To describe such switching behavior, Markov jump process is popular in modeling many practical complex network models with abrupt random changes in their structures. In [12, 13], the general complex networks model with Markov jump structure is developed, which shows that complex networks with Markov jump structure have great application potential in a variety of areas. As a special case of complex networks, the random structure switching behavior of supply networks can be modeled by a Markov process. In [14], the supply chain with stochastic system parameters was modeled as a Markov jump linear system, and the bullwhip effect was analyzed for the supply chain. The work in [15] proposed a discrete-time Markov chain model to characterize the unreliable production capacity in serial supply chain networks. Additionally, the consumer demands in supply networks are uncertain in reality. In [7], the demand of the end customers was regarded as an external disturbance, and the range of the external disturbance was supposed to be bounded. In [16], the demand rate of the production system is assumed to be composed of a known constant term plus an unknown time-varying component with finite energy. However, little literature can be found to study the stability and dynamics of supply networks with stochastic switched topology and uncertain demand.

Motivated by the above discussions, in this paper, we will present a supply networks model with stochastic switched topology that is dependent on a continuous time Markov process and study the stabilization strategies of supply networks model with stochastic switched topology and uncertain demand. By the Lyapunov stability theory and robust control method for the continuous time Markov jumping system [17], the sufficient condition for the existence of the state feedback control strategy for the stability of the supply networks model with stochastic switched topology is given in the form of matrix inequalities, which ensure the robust stability and a prescribed 𝐻_∞ disturbance attenuation level for dynamical supply networks with uncertain demand. From the numerical example, we can observe that the proposed control method can effectively make the supply networks realize robust stability at the stationary states.

The rest of this paper is organized as follows. In Section 2, a supply networks model with stochastic switched topology and uncertain demand is presented. In Section 3, the stabilization strategies of the supply networks model with stochastic switched topology and uncertain demand are derived. In Section 4, a numerical example is provided to demonstrate the effectiveness of the proposed methods. We conclude this paper in Section 5.

2. The Supply Networks Model

Let us consider a supply network that consists of 𝑛 suppliers, in which the supplier 𝑖 delivers materials or resources to the other supplier 𝑗(𝑗≠𝑖) with a delivery rate 𝑐_𝑖𝑗𝑥_𝑗(𝑡), where 𝑐_𝑖𝑗 is the connection weight coefficient of the suppliers 𝑖 and 𝑗. We denote the stock level of supplier 𝑖 at time 𝑡 by 𝑦_𝑖(𝑡), and the delivery rate of supplier 𝑖 at time 𝑡 by 𝑥_𝑖(𝑡). Therefore, the whole supply network can be described by a directed complex networks model, where each supply 𝑖 is the node of the networks and the materials flow represents the connection among the nodes. The dynamics of supply networks include two sets of equations: one is the change of inventories 𝑦_𝑖(𝑡) with time 𝑡, and another is the adaptation of the delivery rates 𝑥_𝑖(𝑡) of the suppliers. Then the inventory 𝑦_𝑖(𝑡) of each supplier 𝑖 is described by material balance equation as follows: 𝑦̇_𝑖(𝑡)=𝑥_𝑖(𝑡)−^𝑛_𝑗=1𝑐_𝑖𝑗𝑥_𝑗(𝑡)+𝑑_𝑖(𝑡),𝑖=1,2,…,𝑛, (1) where 𝑑_𝑖(𝑡) is the demand rate of the end customers for supplier 𝑖. It is assumed that the demand rate 𝑑_𝑖(𝑡) is composed of a known constant component 𝑑_𝑖 (obtained by forecasting) plus an unknown time-varying (uncertain fluctuating) component 𝑤_𝑖(𝑡) with finite energy. Then the form of 𝑑_𝑖(𝑡) is given as 𝑑_𝑖(𝑡)=𝑑_𝑖+𝑤_𝑖(𝑡). (2)

For the delivery rates 𝑥_𝑖(𝑡) of each supplier 𝑖, it is reasonable to assume that the temporal change of the delivery rate is proportional to the deviation of the actual delivery rate from the desired one, and its adaptation takes on the average time interval 𝑇. Based on this, the delivery rate 𝑥_𝑖(𝑡) of the supplier 𝑖 is described by 𝑥̇_𝑖(𝑡)=1𝑇𝐹𝑦_𝑖(𝑡)−𝑥_𝑖(𝑡)+𝑢_𝑖(𝑡),𝑖=1,2,…,𝑛, (3) where 𝑇 is the adaptation time interval, 𝑢_𝑖(𝑡) is the control strategy to be designed, and the function 𝐹(𝑦_𝑖(𝑡)) is the desired delivery rate. The desired delivery rate 𝐹(𝑦_𝑖(𝑡)) is usually reduced with the increase in stock levels 𝑦_𝑖(𝑡). According to the literature [2], we choose the form 𝐹(𝑦_𝑖(𝑡)) as follows: 𝐹𝑦_𝑖(𝑡)𝑦=1−tanh_𝑖(𝑡)−𝑦_𝑐+tanh𝑦_𝑐2, (4) where 𝑦_𝑐 is the safe stock level.

According to the form of function 𝐹(𝑥), for all 𝑥∈𝑅^𝑚, it holds that d𝐹(𝑥)/d𝑥≤1/2, which implies that the function 𝐹(𝑥) satisfies the following Lipschitz condition: 𝐹𝑦₁(𝑡)−𝐹𝑦₂(𝑡≤1)2𝑦₁(𝑡)−𝑦₂(𝑡),∀𝑦₁(𝑡),𝑦₂(𝑡)∈𝑅^𝑚. (5)

The above supply network is based on fluid-dynamic model, which is better suited for online control under dynamically changing condition. Especially in some settings a supplier may need more flexibility under the dynamically changing condition, and the time step needs tb be very small, such as the hardware as well as the semiconductor industries [5, 18].

In practice, the topology structure of the supply networks will change with the time due to interconnections changing between the suppliers. Then we assume that the connection 𝐶=[𝑐_𝑖𝑗]_𝑛×𝑛 of the supply networks is stochastic switching, which is dependent on a continuous time Markov process. Let {𝜃_𝑡,𝑡≥0} be a right-continuous Markov chain on a probability space taking values in a finite set 𝒮={1,2,…,𝑠} with transition probability matrix given by Pr𝜃_𝑡+ℎ=𝑗∣𝜃_𝑡=𝜋=𝑖^𝑖𝑗ℎ+𝑜(ℎ),𝑖≠𝑗1+𝜋_𝑖𝑖ℎ+𝑜(ℎ),𝑖=𝑗, (6) where ℎ>0 and 𝜋_𝑖𝑗≥0, for 𝑗≠𝑖, is the transition rate from state 𝑖 at time 𝑡 to state 𝑗 at time 𝑡+ℎ and satisfies 𝜋_𝑖𝑖=−_𝑗≠𝑖𝜋_𝑖𝑗. (7)

Then by combining (1)–(6), the whole supply networks model with stochastic switched topology is presented as 𝑦̇_𝑖(𝑡)=𝑥_𝑖(𝑡)−^𝑛_𝑗=1𝑐_𝑖𝑗𝜃_𝑡𝑥_𝑗(𝑡)+𝑑_𝑖+𝑤_𝑖(𝑡),𝑥̇_𝑖(𝑡)=1𝑇𝐹𝑦_𝑖(𝑡)−𝑥_𝑖(𝑡)+𝑢_𝑖(𝑡),𝑖=1,2,…,𝑛. (8)

Let 𝑦_𝑖(𝜃_𝑡) and 𝑥_𝑖(𝜃_𝑡) be the stationary state values of the supply network dynamics (8) under the different topology structures. The values of 𝑥_𝑖(𝜃_𝑡) can be obtained from the following equation: 𝑥_𝑖𝜃_𝑡−^𝑛_𝑗=1𝑐_𝑖𝑗𝜃_𝑡𝑥_𝑗𝜃_𝑡+𝑑_𝑖=0,𝑖=1,2,…,𝑛. (9) Then according to the condition that 𝐹(𝑦_𝑖)=𝑥_𝑖, we can get the values of 𝑦_𝑖(𝜃_𝑡).

The objective of this paper is to design the state feedback control 𝑢_𝑖(𝑡) to guarantee that the supply network with stochastic switched topology is robustly stabilized at the stationary state with respect to the unknown fluctuating demand 𝑤_𝑖(𝑡).

3. Stability Strategies of the Supply Networks Model

We consider that the state feedback control 𝑢_𝑖(𝑡) for each supplier 𝑖 is based on the information of the difference between the present stock level 𝑦_𝑖(𝑡) and the stationary state 𝑦_𝑖(𝜃_𝑡) and the difference between the present delivery rate 𝑥_𝑖(𝑡) and the stationary state 𝑥_𝑖(𝜃_𝑡). Then the state feedback control 𝑢_𝑖(𝑡) is designed as follows: 𝑢_𝑖(𝑡)=𝑘_1𝑖𝑦_𝑖(𝑡)−𝑦_𝑖𝜃_𝑡+𝑘_2𝑖𝑥_𝑖(𝑡)−𝑥_𝑖𝜃_𝑡,𝑖=1,2,…,𝑛, (10) where 𝑘_1𝑖,𝑘_2𝑖 are the control parameters to be determined.

Let 𝑦_𝑖(𝑡)=𝑦_𝑖(𝑡)−𝑦_𝑖(𝜃_𝑡),𝑥_𝑖(𝑡)=𝑥_𝑖(𝑡)−𝑥_𝑖(𝜃_𝑡). Applying the state feedback control (10) to the supply networks model (8), one can obtain the error dynamic for supply networks model (8) at the stationary states as follows: 𝑦̇_𝑖(𝑡)=𝑥_𝑖(𝑡)−^𝑛_𝑗=1𝑐_𝑖𝑗𝜃_𝑡𝑥_𝑗(𝑡)+𝑤_𝑖(𝑡),𝑥̇_𝑖(𝑡)=1𝑇𝐹𝑦_𝑖(𝑡)−𝑥_𝑖(𝑡)+𝑘_1𝑖𝑦_𝑖(𝑡)+𝑘_2𝑖𝑥_𝑖(𝑡),𝑖=1,2,…,𝑛, (11) where 𝐹(𝑦_𝑖(𝑡))=𝐹(𝑦_𝑖(𝑡))−𝐹(𝑦_𝑖).

Let 𝑦(𝑡)=[𝑦₁(𝑡),𝑦₂(𝑡),…,𝑦_𝑛(𝑡)]^𝑇, 𝑥(𝑡)=[𝑥₁(𝑡),𝑥₂(𝑡),…,𝑥_𝑛(𝑡)]^𝑇, 𝑤(𝑡)=[𝑤₁(𝑡),𝑤₂(𝑡),…,𝑤_𝑛(𝑡)]^𝑇. Reformulate the error dynamic equation (11) as 𝑦̇(𝑡)=𝐵𝜃_𝑡𝑥(𝑡)(𝑡),𝑥̇(𝑡)𝐹𝑦(𝑡)−𝑥(𝑡)−𝑤=𝐴+𝐾₁𝑦(𝑡)+𝐾₂𝑥(𝑡), (12) where 𝐵(𝜃(𝑡))=[𝑏_𝑖𝑗(𝜃_𝑡)]_𝑛×𝑛and the diagonal element 𝑏_𝑖𝑖(𝜃_𝑡)=1, and if 𝑖≠𝑗, 𝑏_𝑖𝑗(𝜃_𝑡)=−𝑐_𝑖𝑗(𝜃_𝑡), 𝐴=diag{1/𝑇,1/𝑇,…,1/𝑇}, and 𝐾₁=diag{𝑘₁₁,𝑘₁₂,…,𝑘_1𝑛}, 𝐾₂=diag{𝑘₂₁,𝑘₂₂,…,𝑘_2𝑛}.

It is obvious that if the solutions of the error dynamic equation (11) converge to zero, then the supply networks model (8) will be stabilized at the stationary states.

For the unknown fluctuating demand 𝑤(𝑡), the robust stability of the supply networks model (8) is to design the control gains 𝐾₁ and 𝐾₂ such that the error dynamic equation (12) with 𝑤(𝑡)=0 is stable, and for a certain given prescribed 𝐻_∞ disturbance attenuation level 𝛾>0, the following condition holds: ₀^+∞𝛿²(𝑡)d𝑡^1/2≤𝛾₀^+∞𝑤²(𝑡)d𝑡^1/2, (13) for any nonzero 𝑤(𝑡)∈𝐿₂[0,+∞), where 𝛿(𝑡)=[𝑦(𝑡),𝑥(𝑡)]^𝑇.

Based on the Lyapunov stability theory, the following theorem provides a sufficient condition for the existence of the state feedback control strategies for the robust stability of the supply networks model.

Theorem 1. Consider the supply networks model (8) with the state feedback control (10). Let 𝐻₁=𝐼0^𝑇,𝐻₂=0𝐼^𝑇. For a given constant 𝛾>0, if there exist a positive scalar 𝜀, positive definite matrixes 𝑋(𝑖),𝑖=1,2,…,𝑠, and control parameter matrices 𝐾₁,𝐾₂ such that the following matrix inequalities hold: Ω₁(𝑖)12𝑋(𝑖)𝐻₁Ω₂(𝑖)𝑋(𝑖)−𝐻₂12𝐻^𝑇₁𝑋(𝑖)Ω−𝜀𝐼000_𝑇2(𝑖)0−Ω₃(𝑖)𝑋(𝑖)0000−𝐼0−𝐻^𝑇₂000−𝛾²𝐼<0, (14) where Ω₁(𝑖)0𝐴=𝜀0𝐴^𝑇+(𝑖)𝑋(𝑖)(𝑖)(𝑖)0𝐵0−𝐴+𝑋0𝐵0−𝐴^𝑇+𝐾00₁0𝑋(𝑖)(𝑖)𝐾+𝑋00₁0^𝑇+000𝐾₂𝑋(𝑖)+𝑋(𝑖)000𝐾₂^𝑇+𝜋_𝑖𝑖𝑋(𝑖),Ω₂(𝑖)=√𝜋_𝑖1𝑋(𝑖),√𝜋_𝑖2𝑋(𝑖),…,𝜋_{𝑖(𝑖−1)}𝑋(𝑖),𝜋_𝑖(𝑖+1)𝑋(𝑖),…,√𝜋_𝑖𝑠𝑋(𝑖,Ω₃)(𝑖){𝑋(1)(2)𝑋()()(𝑠,=diag,𝑋,…,𝑖−1,𝑋𝑖+1,…,𝑋)} (15)(16)(17) then the state feedback control gains 𝐾₁ and 𝐾₂ are obtained to guarantee the robust stability of the supply networks model (8) with a given prescribed 𝐻_∞ disturbance attenuation level 𝛾>0 at the stationary states.

Proof. At first, we will establish the stability condition for the error dynamics (12) of the supply networks model with 𝑤(𝑡)=0. The Lyapunov function candidate for the error dynamics equation (12) of the supply networks model is constructed as 𝑉𝛿(𝑡),𝜃_𝑡=𝛿^𝑇(𝑡)𝑃𝜃_𝑡𝛿(𝑡), (18) where 𝑃(𝜃_𝑡)>0.
Let ℒ be the weak infinitesimal generator of the stochastic process (𝑒_𝑡, 𝜃_𝑡). Then, for each 𝜃_𝑡=𝑖,𝑖∈𝒮, we have ℒ𝑉𝛿(𝑡),𝜃_𝑡=2𝛿^𝑇(𝑡)𝑃𝜃_𝑡×0𝐴𝐹𝑦(𝑡)+0𝐵𝜃_𝑡𝛿(𝑡)+𝐾0−𝐴00₁0𝛿(𝑡)+000𝐾₂𝛿(𝑡)+𝛿^𝑇(𝑡)^𝑠_𝑗=1𝜋_{𝜃𝑡,𝑗}𝑃𝑗𝛿(𝑡). (19)
By the Cauchy inequality 𝐴^𝑇𝐵+𝐵^𝑇𝐴≤𝛼𝐴^𝑇𝐴+𝛼⁻¹𝐵^𝑇𝐵, for all 𝛼>0, and matrices 𝐴 and 𝐵 with appropriate dimensions, together with condition (5), it holds that 2𝛿^𝑇(𝑡)𝑃𝜃_𝑡0𝐴𝐹𝑦(𝑡)≤𝜀𝛿^𝑇(𝑡)𝑃𝜃_𝑡0𝐴0𝐴^𝑇𝑃𝜃_𝑡𝛿(𝑡)+𝜀⁻¹4𝛿^𝑇(𝑡)𝐼0𝐼0𝛿(𝑡). (20)
Combining (19)-(20), one has ℒ𝑉𝛿(𝑡),𝜃_𝑡≤𝛿^𝑇(𝑡)Ψ𝜃_𝑡𝛿(𝑡), (21) where Ψ𝜃_𝑡=𝜀𝑃𝜃_𝑡0𝐴0𝐴^𝑇𝑃𝜃_𝑡+𝜀⁻¹4𝐼0𝐼0+𝑃𝜃_𝑡0𝐵𝜃_𝑡+𝜃0−𝐴0𝐵_𝑡0−𝐴^𝑇𝑃𝜃_𝑡+𝑃𝜃_𝑡𝐾00₁0+𝐾00₁0^𝑇𝑃𝜃_𝑡+𝑃𝜃_𝑡000𝐾₂+000𝐾₂^𝑇𝑃𝜃_𝑡+^𝑠_𝑗=1𝜋_{𝜃𝑡,𝑗}𝑃𝑗. (22) In addition, let 𝑋⁻¹(𝜃_𝑡)=𝑃(𝜃_𝑡). By Schur Complement [19], pre- and postmultiplying both sides of (14) by diag{𝑋⁻¹(𝜃_𝑡),𝐼,…,𝐼}, one can obtain that inequalities (14) are equivalent to the following inequalities: Ψ(𝑖)+𝐼−𝑃(𝑖)𝐻₂−𝐻^𝑇₂𝑃(𝑖)−𝛾²𝐼<0, (23) which implies that Ψ(𝑖)<0. Thus it holds that ℒ𝑉(𝛿(𝑡),𝜃_𝑡)<0. Then based on Lyapunov stability theory, the error dynamic equation (12) of the supply networks model with 𝑤(𝑡)=0 is stable.
Next, for all nonzero 𝑤(⋅)∈𝐿₂[0,∞), calculating the derivative of 𝑉(𝑡) along the trajectory of error dynamics equation (12) yields that ℒ𝑉𝛿(𝑡),𝜃_𝑡≤𝛿(𝑡)𝑤(𝑡)^𝑇Ψ𝜃_𝑡−𝑃𝜃_𝑡𝐻₂−𝐻^𝑇₂𝑃𝜃_𝑡0𝛿(𝑡)𝑤(𝑡). (24) It obviously follows from (23) and (24) that ℒ𝑉𝛿(𝑡),𝜃_𝑡+𝛿^𝑇(𝑡)𝛿(𝑡)−𝛾²𝑤^𝑇(𝑡)𝑤(𝑡)<0. (25)
Thus, under the zero initial condition, integrating both sides of (25) from 0 to +∞ and noting that lim_𝑡→∞𝔼{𝑉(𝛿(𝑡),𝜃_𝑡)}=0, we have 𝔼‖𝛿(𝑡)‖²≤𝛾²‖𝑤(𝑡)‖². (26) The supply networks model (8) satisfies (13) for any nonzero 𝑤(𝑡)∈𝐿₂[0,+∞).
Therefore, the supply networks model (8) is robustly stable with the 𝐻_∞ disturbance attenuation level 𝛾 at the stationary states. The proof is completed.

Remark 2. Theorem 1 gives a sufficient condition to choose proper state feedback control gains 𝐾₁ and 𝐾₂ such that the supply networks model (8) is robustly stable for the external disturbance 𝑤(𝑡). For the given prescribed 𝐻_∞ disturbance attenuation level 𝛾>0, condition (14) in Theorem 1 takes the form of bilinear matrix inequalities (BMIs) about the matrix variables. Then according to the proposed mixed algorithm to solve the BMIs problem in [20, 21], the state feedback control gains 𝐾₁ and 𝐾₂ can be solved numerically efficiently with the available software.

Remark 3. The stability of the supply networks at the stationary inventory states can effectively reduce the stocking or shortage costs and increase the profits for the suppliers of supply networks. Furthermore, the proposed robust stability strategy in this paper can effectively suppress the bullwhip effect under the uncertain customer demand so as to reduce the extra operation costs and improve customer service level of the whole networks.

4. Numerical Simulations

In this section, a numerical example is given to illustrate the effectiveness of the proposed methods for the stabilization of the supply networks model.

Let us consider a supply network with 6 suppliers. The adaptation time interval is chosen as 𝑇=10. The safe stock level 𝑦_𝑐=3. Here we suppose that the supply networks model has two switched topology modes, which is dependent on a continuous time Markov process. The two switched topologies are given as Figure 1.

(a)

(b)

Figure 1 

The two switched topologies of supply networks.

According to the two switched topologies presented in Figure 1, we consider the following two switched coupling matrices: 𝐵(1)=21−3400000100−50−13−201000330−42100000−3110000−3,𝐵(2)=2011−310000010−3−1003−201000330−4110000−5−13110000−3.01 (27) Suppose the transition rate matrix is given by Π=−0.50.50.4−0.4. (28) Figure 2 is one of the possible realizations of the Markovian jumping mode.

Figure 2 

The numerical simulation of the Markovian jumping mode.

For a given small prescribed 𝐻_∞ disturbance attenuation level 𝛾=2.5, by solving condition (14) in Theorem 1 with the proposed mixed algorithm proposed in [20, 21], a set of feasible state feedback control parameters can be obtained as 𝑘_1𝑖=−1.9,𝑘_2𝑖=−1.5. The initial error inventory level 𝑦_𝑖(0) is chosen randomly from the interval [−5,5], and the initial error delivery rates 𝑥_𝑖(0) are chosen randomly from the interval [−1,1]. Suppose that 𝑤_𝑖(𝑡)=0.1sin(𝑡)/𝑡 and the average demand rates 𝑑=[𝑑₁,𝑑₂,…,𝑑₆]=[0.10.20.20.20.10.10.2]. Figure 3 is the evolution curves of the error states of supply networks without control, which shows that the supply network is not stable at the stationary states. Under the control (10), the evolution curves of the error states of supply networks depicted in Figure 4, from which we can see that under control (10), the supply networks model (8) realizes the robust stabilization at the stationary states at 𝑡=20. The simulations show the effectiveness of the proposed methods in this paper.

(a)

(b)

Figure 3 

The evolution curves of the error states of supply networks without control.

(a)

(b)

Figure 4 

The evolution curves of the error states of supply networks under control (10).

5. Conclusion

In this paper, a supply networks model with stochastic switched topology is developed, and the stabilization strategies of the supply networks with stochastic switched topology are designed. Based on the Lyapunov stability theory, a sufficient condition for the existence of state feedback control strategies is given in terms of matrix inequalities, which ensure the robust stability of the supply networks at the stationary states and a prescribed 𝐻_∞ disturbance attenuation level for the uncertain demand. This paper provides a new way to study the stabilization of the supply networks with varying topology. A numerical example shows the effectiveness of the proposed methods. Additionally, in reality there will be benefit conflicts between suppliers. Then how to design stability strategy of supply network with conflicts between suppliers may be an interesting issue and need to be investigated in the future. Moreover, the stability strategy of supply network with partial information will be another future research.

Acknowledgments

This research was supported by the National Natural Science Foundation of China (nos. 61004015 and 71271020), the Program for New Century Excellent Talents in Universities of China (nos. NCET-11-0377), and the Research Foundation of State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, (no. RCS2012ZZ001).