 Research
 Open Access
 Published:
Pinning control design for feedback stabilization of constrained Boolean control networks
Advances in Difference Equations volume 2016, Article number: 182 (2016)
Abstract
We study the pinning control design for feedback stabilization of constrained Boolean control networks (BCNs) via the semitensor product of matrices. Firstly, a constrained algebraic representation is obtained for constrained BCNs with pinning control, and a necessary and sufficient condition is established for the reachability of constrained BCNs with pinning control. Secondly, a general procedure is proposed for the pinning control design of state feedback stabilization of constrained BCNs. Thirdly, a necessary and sufficient condition is presented for the output feedback stabilization of constrained BCNs with pinning control. Finally, two illustrative examples are worked out to demonstrate the effectiveness of the obtained new results.
Introduction
In recent years, with the development of systems biology and medical science, there have been a lot of computational models to simulate gene regulatory networks (GRNs) [1, 2]. In 1969, Kauffman [3] firstly proposed the concept of Boolean networks to model GRNs. From then on, the study of Boolean networks has attracted attention of many scholars [4–7]. In a Boolean network, genes are simulated as 1 and 0 for studying the activity of genes. Since the dynamics of Boolean networks does not contain parameters, we can use Boolean networks to model largescale GRNs.
Recently, a semitensor product method [8] has been proposed for the study of Boolean networks. Using this novel method, we can convert a Boolean (control) network into a linear (bilinear) form. Then, we can conveniently investigate Boolean networks with the help of classical control theory. This framework is called the algebraic state space representation (ASSR) of logical systems. Up to now, many scholars have applied the ASSR framework to the analysis and control of Boolean networks and obtained lots of interesting results [9–23]. Moreover, the semitensor product method has also been applied to networked evolutionary games [24, 25] and feedback shift registers [26].
As one of the most important issues in the study of GRNs, the stabilization of Boolean control networks (BCNs) has been found to be widely applied in the design of therapeutic interventions and the explanation of some living phenomena [27]. The state feedback stabilization of BCNs was studied in [28, 29], and a novel design procedure was established. Later, the design of output feedback stabilizers of BCNs was investigated [30–32], and some necessary and sufficient conditions were presented.
Note that the pinning control problem has been introduced to the study of Boolean networks in some recent works [33, 34]. With the pinning control, we can realize the control objective by controlling a small number of nodes. Using the ASSR framework, the pinning control design for the stabilization of Boolean networks was studied in [34], and a novel design procedure was established. Moreover, in GRNs, some states and inputs are undesirable because they may lead to a dangerous situation such as the deterioration of a disease, the metastasis of a cancer [7, 35], and so on. Hence, we need put some constraints to the undesirable states and inputs in BCNs. It should be noticed that the pinning control design for feedback stabilization of constrained BCNs is still a challenging problem, and there are fewer results on this problem. In addition, the pinning stabilization control design procedure proposed in [34] cannot be directly applied to the pinning control design for feedback stabilization of constrained BCNs. This motivates us to study the pinning control design for feedback stabilization of constrained BCNs.
In this paper, using the ASSR framework, we investigate the pinning control design for feedback stabilization of constrained BCNs and present a number of new results. The main contributions of this paper are as follows. On one hand, we convert the dynamics of constrained BCNs with pinning control into a constrained algebraic form, which facilitates the reachability analysis and control design of constrained BCNs with pinning control. On the other hand, we present some necessary and sufficient conditions for the pinning control design of feedback stabilizers of constrained BCNs, which can be easily verified with the help of MATLAB.
The rest of this paper is organized as follows. In Section 2, we give some preliminaries on the semitensor product of matrices. In Section 3, we investigate the pinning control design for feedback stabilization of constrained BCNs and present the main results of this paper. Two illustrative examples are worked out to verify the obtained new results in Section 4, which is followed by a brief conclusion in Section 5.
Notation
\(\mathbb{R}\), \(\mathbb{N}\), and \(\mathbb{Z}_{+}\) denote the sets of real numbers, natural numbers, and positive integers, respectively; \(\mathcal{D}:=\{1,0\}\), \(\mathcal {D}^{n}:=\underbrace{\mathcal{D}\times\cdots\times\mathcal {D}}_{n}\), \(\Delta _{n}:=\{\delta _{n}^{k} \mid 1\leq k\leq n\}\), where \(\delta _{n}^{k}\) denotes the kth column of \(I_{n}\), and for compactness, \(\Delta :=\Delta_{2}\). An \(n\times t\) matrix M is called a logical matrix if \(M=[\delta _{n}^{i_{1}}\ \delta _{n}^{i_{2}}\ \cdots\ \delta _{n}^{i_{t}}]\). We express M briefly as \(M=\delta _{n}[i_{1}\ i_{2}\ \cdots\ i_{t}]\). Denote the set of \(n\times t\) logical matrices by \({\mathcal {L}}_{n\times t}\). An \(n\times t\) matrix \(M=(M_{i,j})\) is called a Boolean matrix if \(M_{i,j}\in\mathcal{D}\), \(i=1,\ldots,n\), \(j=1,\ldots,t\). Denote the set of \(n\times t\) Boolean matrices by \({\mathcal {B}}_{n\times t}\). Given a real matrix \(A\in\mathbb{R}^{m\times n}\), \(\operatorname{Col}_{i}(A)\) denotes the ith column of A, and \(\operatorname{Row}_{i}(A)\) denotes the ith row of A; ‘∗’ denotes the KhatriRao product of matrices.
Preliminaries
First of all, we give some fundamental preliminaries on the semitensor product of matrices. For details, we refer to [8].
Definition 1
([8])
The semitensor product of two matrices \(A\in\mathbb{R}^{m\times n}\) and \(B\in \mathbb{R}^{p\times q}\) is
where \(\alpha=\operatorname{lcm}(n,p)\) represents the least common multiple of n and p, and ⊗ is the Kronecker product.
Obviously, when \(n=p\), the semitensor product of A and B becomes the conventional matrix product. Thus, we can simply call it ‘product’ and omit the symbol ‘⋉’ if no confusion arises.
Proposition 1
([8])
Let \(X\in \mathbb{R}^{t\times1}\) be a column vector, and \(A\in\mathbb {R}^{m\times n}\). Then
Proposition 2
([8])
Let \(X\in \mathbb{R}^{m\times1}\) and \(Y\in\mathbb{R}^{n\times1}\) be two column vectors. Then
where \(W_{[m,n]}\in\mathcal{L}_{mn\times mn}\) is the socalled swap matrix defined as
Proposition 3
([8])
Define the matrix, \(M_{r,k}\), called the kvalued powerreducing matrix, as
where \(0_{k}\in\mathbb{R}^{k}\) is the zero vector.
Denoting \(1\sim\delta _{2}^{1}\) and \(0\sim\delta _{2}^{2}\), we have \(\Delta \sim\mathcal{D}\), where ‘∼’ denotes two different forms of the same object. In most places of this paper, we use \(\delta _{2}^{1}\) and \(\delta _{2}^{2}\) to express logical variables and call them the vector form of logical variables. The following lemma is fundamental for the matrix expression of logical functions.
Lemma 1
([8])
Let \(f(x_{1}, x_{2}, \ldots, x_{s}): \mathcal{D}^{s}\mapsto\mathcal{D}\) be a logical function. Then, there exists a unique matrix \(M_{f}\in{\mathcal {L}}_{2\times2^{s}}\), called the structural matrix of f, such that
For example, the structural matrices for negation (¬), conjunction (∧), disjunction (∨), conditional (→), biconditional (↔), and exclusive Or (\(\bar{\vee}\)) are \(M_{n}=\delta _{2}[2\ 1]\), \(M_{c}=\delta _{2}[1\ 2\ 2\ 2]\), \(M_{d}=\delta _{2}[1\ 1\ 1\ 2]\), \(M_{i}=\delta _{2}[1\ 2\ 1\ 1]\), \(M_{e}=\delta _{2}[1\ 2\ 2\ 1]\) and \(M_{p}=\delta _{2}[2\ 1\ 1\ 2]\), respectively.
Based on Lemma 1, we can convert a BCN into an algebraic representation. For details, refer to [8].
Main results
In this section, we give the main results of this paper. Firstly, we convert the dynamics of a constrained Boolean network with pinning control into its equivalent constrained algebraic form, based on which we obtain a necessary and sufficient condition for the reachability of constrained BCNs with pinning control. Secondly, we present the pinning control design procedure for the feedback stabilization of constrained BCNs.
Constrained algebraic form
A Boolean network with n network nodes and r pinning controls can be described as
where nodes \(i_{1},\ldots,i_{r}\) are selected to be pinning controlled, and \(1\leq r< n \). Without loss of generality, we assume that \(i_{s}=s\), \(s=1,\ldots, r\). \(x_{i}(t)\in\mathcal{D}\), \(i=1,\ldots ,n\), \(u_{i}(t)\in\mathcal{D}\), \(i=1,\ldots,r\), and \(y_{i}(t)\in\mathcal {D}\), \(i=1,\ldots,p\), are the states, the control inputs, and the outputs of system (6), respectively, and \(f_{i}: \mathcal {D}^{n+1}\mapsto\mathcal{D}\), \(i=1,\ldots,r\), \(f_{i}: \mathcal {D}^{n}\mapsto\mathcal{D}\), \(i=r+1,\ldots,n\), and \(h_{i}: \mathcal {D}^{n}\mapsto\mathcal{D}\), \(i=1,\ldots,p\), are logical functions.
In order to convert system (6) into an algebraic form, we define \(x(t)=\ltimes_{i=1}^{n}x_{i}(t)\in\Delta_{2^{n}}\), \(x^{1}(t)=\ltimes_{i=1}^{r}x_{i}(t)\in\Delta_{2^{r}}\), \(x^{2}(t)=\ltimes_{i=r+1}^{n}x_{i}(t)\in\Delta_{2^{nr}}\), \(u(t)=\ltimes_{i=1}^{r}u_{i}(t)\in\Delta_{2^{r}}\), and \(y(t)=\ltimes _{i=1}^{p}y_{i}(t)\in\Delta_{2^{p}}\). Assume that the structural matrix of \(f_{i}\) and \(h_{i}\) are \(F_{i}\), \(i=1,\ldots, n\), and \(H_{i}\), \(i=1,\ldots,p\), respectively. Using Lemma 1, system (6) can be expressed as
and
where
and
Summarizing, we obtain the following algebraic form of system (6):
where \(Q=L_{1}(I_{2^{n+r}}\otimes L_{2})W_{[2^{n},2^{n+r}]}M_{r,2^{n}}W_{[2^{r},2^{n}]}\in\mathcal {L}_{2^{n}\times2^{n+r}}\).
Now, we consider system (10) with state and input constraints. For any \(t\in\mathbb{N}\), we assume that \(x(t)\in S_{x}\subseteq\bigtriangleup_{2^{n}}\) and \(u(t)\in S_{u}\subseteq \bigtriangleup_{2^{r}}\). Let \(S_{x}=n_{1}\leq2^{n}\) and \(S_{u}=r_{1}\leq2^{r}\), where \(S_{x}\) denotes the cardinality of the set \(S_{x}\). Then, \(S_{x}\) and \(S_{u}\) can be expressed as
Denote the trajectory of system (10) with a pinning control sequence \(\{(u_{1}(t), u_{2}(t),\ldots, u_{r}(t)):t \in\mathbb{N}\} \subseteq S_{u}\) and an initial state \(x_{0}\in S_{x}\) by \(x(t;x_{0},(u_{1}(t), u_{2}(t),\ldots, u_{r}(t)))\).
In the following, we convert system (10) with state and input constraints into an equivalent constrained algebraic form.
Define the following set of matrices:
where \(J_{i}^{(p,q)}\in\mathbb{R}^{q\times pq}\), \(i=1,2,\ldots,p\), \(0_{q\times q}\) denotes the \(q\times q\) zero matrix, and \(I_{q}\in \mathcal{L}_{q\times q}\) is the \(q\times q\) identity matrix.
Proposition 4
([35])
1. Given a matrix \(A\in\mathbb{R}^{pq\times r}\), split A as
where \(A_{i}\in\mathbb{R}^{q\times r}\). Then,
2. Given a matrix \(B\in\mathbb{R}^{r\times pq}\), split B as
where \(B_{i}\in\mathbb{R}^{r\times q}\). Then,
Based on Proposition 4, set
Denote \(\delta_{n_{1}}^{0}=0_{n_{1}\times1}\) and \(\delta _{r_{1}}^{0}=0_{r_{1}\times1}\). Then, we convert the state \(x(t)\in \Delta _{2^{n}}\) and input \(u(t)\in \Delta _{2^{r}}\) of the constrained system into the following form:
where \(\widehat{S}_{x}=\{\delta_{n_{1}}^{1},\delta _{n_{1}}^{2},\ldots,\delta_{n_{1}}^{n_{1}}\}\cup\{\delta _{n_{1}}^{0}\}\) and \(\widehat{S}_{u}=\{\delta_{r_{1}}^{1},\delta _{r_{1}}^{2},\ldots,\delta_{r_{1}}^{r_{1}}\}\cup\{\delta _{r_{1}}^{0}\}\).
For system (10), let \(Q=[Q_{1},\ldots,Q_{2^{r}}]\), \(Q_{l}\in \mathcal{L}_{2^{n}\times2^{n}}\), \(l\in\{1,\ldots,2^{r}\}\). Set
where
Remark 1
Since Q is a logical matrix, we can easily conclude that each column of Q̂ has at most one element ‘1’.
Based on the above transformation, we convert system (10) into the following form:
Proposition 5
The state trajectories of system (10) with \(S_{x}\) and \(S_{u}\) are equivalent to that of system (25) with \(\widehat{S}_{x}\) and \(\widehat{S}_{u}\).
Proof
On one hand, \(\forall t\in\mathbb{N}\), \(\forall u(t)=\delta _{2^{r}}^{j_{s}}\in S_{u}\), and \(\forall x(t)=\delta_{2^{n}}^{i_{\alpha}}\in S_{x}\), if \(x(t+1)=Qu(t)x(t)\in S_{x}\), say, \(x(t+1)=\delta_{2^{n}}^{i_{\beta }}\), \(\beta\in\{1, \ldots, n_{1}\}\), a simple calculation shows that
where \(\widehat{x}(t)=\delta_{n_{1}}^{\alpha}\in\widehat{S}_{x}\) and \(\widehat{u}(t)=\delta_{r_{1}}^{s}\in\widehat{S}_{u}\). If \(x(t+1)=Qu(t)x(t)=\delta_{2^{n}}^{i}\notin S_{x}\), then \(\widehat {x}(t+1)=\widehat{Q}\widehat{u}(t)\widehat{x}(t)=\delta_{n_{1}}^{0}\). Hence, in both cases, \(\widehat{x}(t+1)=\Psi_{x}x(t+1)\).
On the other hand, \(\forall t\in\mathbb{N}\), \(\forall \widehat {u}(t)=\delta_{r_{1}}^{s}\in\widehat{S}_{u}\setminus\{\delta _{r_{1}}^{0}\}\), and \(\forall\widehat{x}(t)=\delta_{n_{1}}^{\alpha}\in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\), if \(\widehat {x}(t+1)=\widehat{Q}\widehat{u}(t)\widehat{x}(t)\in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\), say, \(\widehat {x}(t+1)=\delta_{n_{1}}^{\beta}\), \(\beta\in\{1, \ldots, n_{1}\}\), we can obtain that
If \(\widehat{x}(t+1)=\widehat{Q}\widehat{u}(t)\widehat{x}(t)=\delta _{n_{1}}^{0}\), then
Hence, we have \(\widehat{x}(t+1)=\Psi_{x}x(t+1)\).
Therefore, the state trajectories of system (10) with \(S_{x}\) and \(S_{u}\) are equivalent to that of system (25) with \(\widehat{S}_{x}\) and \(\widehat{S}_{u}\). □
Remark 2
We call (25) the constrained algebraic form of the original system. Based on Proposition 5, we can convert the feedback stabilization of the original system to that of system (25).
Reachability analysis
In this subsection, we study the reachability of system (25), which is crucial to pinning control design for the feedback stabilization.
We give the definition of the reachability for system (25) as follows.
Definition 2
For system (25), given two states \(\widehat{x}_{0}, \widehat{x}_{d} \in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\) and a integer \(k>0\), \(\widehat{x}_{d}\) is said to be reachable from \(\widehat{x}_{0}\) at time k if there is a pinning control sequence \((\widehat{u}_{1}(t), \widehat{u}_{2}(t),\ldots,\widehat{u}_{r}(t))\) with \(\widehat {u}(t)=\ltimes_{i=1}^{r}\widehat{u}_{i}(t)\in\widehat{S}_{u}\setminus \{\delta_{r_{1}}^{0}\}\), \(t\in\{1,2,\ldots,k1\}\), such that \(\widehat{x}(k; \widehat{x}_{0}, (\widehat{u}_{1}(t), \widehat {u}_{2}(t),\ldots,\widehat{u}_{r}(t)))=\widehat{x}_{d}\).
For system (25), consider two given states \(\widehat {x}_{0}=\delta_{n_{1}}^{\alpha}\in\widehat{S}_{x}\setminus\{\delta _{n_{1}}^{0}\}\), \(\widehat{x}_{d}=\delta_{n_{1}}^{\beta}\in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\) and a given integer \(k>0\). Let \(P(k;\widehat{x}_{0}, \widehat{x}_{d})\) denote the number of different paths such that \(\widehat{x}_{d}\) is reachable from \(\widehat{x}_{0}\) at time k.
Lemma 2
Consider system (25) with two given states \(\widehat{x}_{0}=\delta_{n_{1}}^{\alpha}\in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\), \(\widehat{x}_{d}=\delta _{n_{1}}^{\beta}\in\widehat{S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\) and a given integer \(k>0\). Then,
where \(\overline{Q}=\sum_{i=1}^{r_{1}}\widehat{Q}_{i}\), and \(\widehat{Q}_{i}\) is defined in (24).
Proof
We prove this lemma by induction.
Firstly, letting \(k=1\), assume that \(\widehat{u}^{1}=\delta _{r_{1}}^{\rho_{1}},\widehat{u}^{2}=\delta_{r_{1}}^{\rho _{2}},\ldots,\widehat{u}^{s}=\delta_{r_{1}}^{\rho_{s}}\in\widehat {S}_{u}\setminus\{\delta_{r_{1}}^{0}\}\) are different control sequences such that \(\widehat{x}_{d}\) is reachable from \(\widehat {x}_{0}\) at one step. Let \(\widehat{u}^{s+1}=\delta_{r_{1}}^{\rho _{s+1}},\widehat{u}^{s+2}=\delta_{r_{1}}^{\rho_{s+2}},\ldots ,\widehat{u}^{r_{1}}=\delta_{r_{1}}^{\rho_{r_{1}}}\in\widehat {S}_{u}\setminus\{\delta_{r_{1}}^{0}\}\) be different control sequences such that \(\widehat{x}_{0}\) cannot reach \(\widehat{x}_{d}\) in one step. Hence, it is easy to see that \((\widehat{Q}_{\rho_{l}})_{\beta ,\alpha}=1\), \(\forall l\in\{1,\ldots,s\}\), and \((\widehat{Q}_{\rho _{l}})_{\beta,\alpha}=0\), \(\forall l\in\{s+1,\ldots,r_{1}\}\), which implies that \((\sum_{i=1}^{r_{1}}\widehat{Q}_{i} )_{\beta ,\alpha}=s\). Since \((\widehat{x}_{d}^{T})(\overline{Q})(\widehat {x}_{0})= (\sum_{i=1}^{r_{1}}\widehat{Q}_{i} )_{\beta,\alpha }\), we have
Thus, (26) holds for \(k=1\).
Suppose that (26) holds for an integer \(k\geq1\). Then, we consider the case of \(k+1\). It is easy to see that
which shows that (26) holds for \(k+1\).
By induction, (26) holds for any integer \(k>0\). This completes the proof. □
Based on Lemma 2, we give the following result on the reachability of system (25).
Theorem 1
For system (25), \(\widehat {x}_{d}=\delta_{n_{1}}^{\beta}\in\widehat{S}_{x}\setminus\{\delta _{n_{1}}^{0}\}\) is reachable from \(\widehat{x}_{0}=\delta_{n_{1}}^{\alpha}\in\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\) at time k if and only if
Proof
By Lemma 2 we get \(P(k;\widehat{x}_{0},\widehat {x}_{d})=(\delta_{n_{1}}^{\beta})^{T}(\overline{Q})^{k}(\delta _{n_{1}}^{\alpha})= (\overline{Q}^{k} )_{\beta,\alpha}\). Therefore, \(P(k;\widehat{x}_{0},\widehat{x}_{d})\) denotes the number of different paths from \(\widehat{x}_{0}\) to \(\widehat{x}_{d}\) at time k. If \((\overline{Q}^{k})_{\beta,\alpha}>0\), then there are \((\overline{Q}^{k})_{\beta,\alpha}\) different paths from \(\widehat {x}_{0}\) to \(\widehat{x}_{d}\). Thus, \(\widehat{x}_{d}\) can be reached from \(\widehat{x}_{0}\) at time k. Conversely, if \(\widehat{x}_{d}\) is reachable from \(\widehat{x}_{0}\) at time k, then we can find at least one pinning control sequence \((\widehat{u}_{1}(t), \widehat {u}_{2}(t),\ldots,\widehat{u}_{r}(t))\) with \(\widehat{u}(t)=\ltimes _{i=1}^{r}\widehat{u}_{i}(t)\in\widehat{S}_{u}\setminus\{\delta _{r_{1}}^{0}\}\), \(t\in\{1,2,\ldots,k1\}\), such that \(\widehat{x}(k; \widehat{x}_{0}, (\widehat{u}_{1}(t), \widehat{u}_{2}(t),\ldots ,\widehat{u}_{r}(t)))=\widehat{x}_{d}\), which implies that \((\overline{Q}^{k} )_{\beta,\alpha}=P(k;\widehat{x}_{0},\widehat {x}_{d})>0\). □
Remark 3
From the proof of Theorem 1 we easily see that
All of them denote the number of different paths such that \(\widehat {x}_{d}\) is reachable from \(\widehat{x}_{0}\) at time k.
Feedback stabilization pinning control design
In this part, we study the pinning control design for the feedback stabilization of constrained BCNs. By Proposition 5 we consider the feedback stabilization of system (25) based on the reachability analysis.
Firstly, we give the definition of stabilization for system (10) with \(S_{x}\) and \(S_{u}\).
Definition 3
System (10) with \(S_{x}\) and \(S_{u}\) is said to be stabilizable to a given equilibrium \(x_{e}\in S_{x}\) if there exists a pinning control sequence \(\{(u_{1}(t), u_{2}(t),\ldots,u_{r}(t)):t\in\mathbb{N}\}\subset S_{u}\) under which the trajectory initialized at any \(x_{0}\in S_{x}\) converges to \(x_{e}\) and \(x(t;x_{0},(u_{1}(t), u_{2}(t),\ldots, u_{r}(t)))\in S_{x}\), \(\forall t \in\mathbb{N}\).
In this paper, we study the following two kinds of feedback pinning controls:

1.
State feedback pinning control:
$$ u_{i}(t)=K_{i}x(t), $$(28)where \(K_{i}\in\mathcal{L}_{2\times2^{n}}\), \(i=1,\ldots,r\).

2.
Output feedback pinning control:
$$ u_{i}(t)=G_{i}y(t), $$(29)where \(G_{i}\in\mathcal{L}_{2\times2^{p}}\), \(i=1,\ldots,r\).
In the following, we consider the pinning control design for the state feedback stabilization of system (10) with \(S_{x}\) and \(S_{u}\) based on the constrained algebraic form.
For system (25), let \(\widehat{x}_{e}=\Psi_{x}x_{e}=\delta _{n_{1}}^{\alpha}\in\widehat{S}_{x}\setminus\{\delta_{n_{1}}^{0}\} \). For any integer \(k>0\), define
Proposition 6
System (25) is stabilized to \(\widehat{x}_{e}=\delta_{n_{1}}^{\alpha}\) by a state feedback control if and only if there exists a positive integer \(\sigma\leq n_{1}\) such that

1.
\((\overline{Q})_{\alpha,\alpha}>0\),

2.
\(\operatorname{Row}_{\alpha}(\overline{Q}^{\sigma})>0\).
Proof
(Sufficiency) We can see from Condition 1 and Theorem 1 that \(\widehat{x}_{e}\in\Lambda_{1}(\widehat{x}_{e})\), which implies that \(\Lambda_{k}(\widehat{x}_{e})\neq\emptyset\), \(\forall k=1,\ldots,\sigma\).
Denote
where \(\Lambda_{0}(\widehat{x}_{e}):=\emptyset\). Then, we obtain \(\Lambda_{k_{1}}^{\circ}(\widehat{x}_{e})\cap\Lambda _{k_{2}}^{\circ}(\widehat{x}_{e})=\emptyset\), \(\forall k_{1},k_{2}\in\{1,\ldots,\sigma\}\), \(k_{1}\neq k_{2}\). Moreover, by Condition 2 and Theorem 1 we have \(\bigcup_{k=1}^{\sigma}\Lambda_{k}^{\circ}(\widehat{x}_{e})=\widehat {S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\). Thus, for any integer i satisfying \(1\leq i\leq n_{1}\), we can find the unique integer \(1\leq k_{i}\leq\sigma\) such that \(\delta _{n_{1}}^{i}\in\Lambda_{k_{i}}^{\circ}(\widehat{x}_{e})\).
For system (25), set \(\widehat{Q}=\delta_{n_{1}}[\mu _{1},\mu_{2},\ldots,\mu_{n_{1}r_{1}}]\). When \(k_{i}=1\), we can find an integer \(1\leq\omega_{i}\leq r_{1}\) such that \(\widehat{Q}\ltimes\delta_{r_{1}}^{\omega_{i}}\ltimes \delta_{n_{1}}^{i}=\delta_{n_{1}}^{\mu_{(\omega _{i}1)n_{1}+i}}=\widehat{x}_{e}\). When \(2\leq k_{i}\leq\sigma\), we can find an integer \(1\leq\omega _{i}\leq r_{1}\) such that \(\widehat{Q}\ltimes\delta_{r_{1}}^{\omega _{i}}\ltimes\delta_{n_{1}}^{i}=\delta_{n_{1}}^{\mu_{(\omega _{i}1)n_{1}+i}}\in\Lambda_{k_{i}1}(\widehat{x}_{e})\).
Let
Then, for any initial state \(\widehat{x}_{0}=\delta_{n_{1}}^{i}\in \widehat{S}_{x}\setminus\{\delta_{n_{1}}^{0}\}\), if \(k_{i}=1\), then we obtain
if \(2\leq k_{i}\leq\sigma\), we have
Hence, \(\widehat{x}(k_{i};\widehat{x}_{0},\widehat{u})=\widehat {x}_{e}\), \(\forall1\leq i\leq n_{1}\), and \(\widehat{x}(t;\widehat {x}_{0},\widehat{u})\in\widehat{S}_{x}\setminus\{\delta _{n_{1}}^{0}\}\), \(\forall0\leq t\leq k_{i}1\). Since \(\widehat {x}_{e}\in\Lambda_{1}(\widehat{x}_{e})\), we obtain
which implies that system (25) is stabilized to \(\widehat {x}_{e}=\delta_{n_{1}}^{\alpha}\) by the state feedback control \(\widehat{u}(t)=\widehat{K}\widehat{x}(t)\).
(Necessity) The proof of this part is based on a straightforward calculation, and thus we omit it here. □
Based on Propositions 5 and 6, we have the following result.
Theorem 2
System (6) with \(S_{x}\) and \(S_{u}\) is stabilized to \(x_{e}\) by a state feedback control if and only if there exists a positive integer \(\sigma\leq n_{1}\) such that \((\overline{Q})_{\alpha,\alpha}>0\) and \(\operatorname{Row}_{\alpha}(\overline {Q}^{\sigma})>0\).
From the proof of Proposition 6 we get the following procedure for the pinning control design of state feedback stabilization of constrained BCNs.
Remark 4
The procedure contains the following steps:

1.
Calculate \(\Lambda_{k}(\widehat{x}_{e})\) and \(\Lambda _{k}^{\circ}(\widehat{x}_{e})\), \(k=1,\ldots,\sigma\).

2.
For every integer \(1\leq i\leq n_{1}\), find the unique integer \(1\leq k_{i}\leq\sigma\) satisfying \(\delta_{n_{1}}^{i}\in\Lambda_{k_{i}}^{\circ}(\widehat{x}_{e})\).

3.
Find an integer \(1\leq\omega_{i}\leq r_{1}\) such that if \(k_{i}=1\), then \(\delta_{n_{1}}^{\mu_{(\omega _{i}1)n_{1}+i}}=\widehat{x}_{e}\); if \(k_{i}\geq2\), then \(\delta _{n_{1}}^{\mu_{(\omega_{i}1)n_{1}+i}}\in\Lambda_{k_{i}1}(\widehat {x}_{e})\).

4.
The state feedback pinning control can be designed as \(u_{i}(t)=K_{i}x(t)\), \(i=1,\ldots,r\), with \(K_{1}\ast K_{2}\ast\cdots\ast K_{r}=K\), where \(K=\delta_{2^{r}}[p_{1},\ldots,p_{2^{n}}]\), and
$$ \textstyle\begin{cases} p_{t}=j_{\omega_{\rho}} &\mbox{if }t=i_{\rho}, \rho\in\{1,\ldots,n_{1}\} , \\ p_{t}\in\{j_{1},\ldots,j_{r_{1}}\} &\mbox{otherwise}. \end{cases} $$(33)
Finally, we discuss the pinning control design for the output feedback stabilization of constrained BCNs. To this end, we recall the definition of a nilpotent matrix.
Definition 4
A nilpotent matrix N is a square matrix such that \(N^{k}=0\) for some positive integer k. The smallest such k is called the degree of N.
For system (25) with an output feedback control \(\widehat {u}(t)=\widehat{G}\widehat{y}(t)\), \(\widehat{G}\in\mathcal {L}_{r_{1}\times2^{p}}\), we have
where \(M_{r,n_{1}}=\operatorname{diag}\{\delta_{n_{1}}^{1},\delta_{n_{1}}^{2},\ldots ,\delta_{n_{1}}^{n_{1}}\}\). Then, we have the following result on the output feedback stabilization of system (25).
Theorem 3
System (25) is stabilizable to \(\widehat{x}_{e}=\delta_{n_{1}}^{\alpha}\) by an output feedback control if and only if there exist a logical matrix \(\widehat{G}\in \mathcal{L}_{r_{1}\times2^{p}}\) and an integer \(1\leq\tau\leq n_{1}\) such that
and
is a nilpotent matrix of degree τ, where \(A_{1}\in\mathcal {B}_{(\alpha1)\times(\alpha1)}\), \(A_{2}\in\mathcal{B}_{(\alpha 1)\times(n_{1}\alpha)}\), \(A_{3}\in \mathcal{B}_{(n_{1}\alpha )\times(\alpha1)}\), \(A_{4}\in\mathcal{B}_{(n_{1}\alpha)\times (n_{1}\alpha)}\), \(\zeta_{1}\) and \(\zeta_{2}\) are some proper Boolean row vectors, and \(\zeta_{3}\) and \(\zeta_{4}\) are zero column vectors.
Proof
(Sufficiency) Since
is a nilpotent matrix of degree τ, we see that
for any integer \(t\geq\tau\), which, together with a simple calculation, shows that
for any integer \(t\geq\tau\).
Hence, we obtain from (34) that
for any \(\widehat{x}(0)\in\widehat{S}_{x}\setminus\{\delta _{n_{1}}^{0}\}\) and any integer \(t\geq\tau\).
Therefore, system (25) is stabilizable to \(\widehat {x}_{e}=\delta_{n_{1}}^{\alpha}\) by the output feedback control \(\widehat{u}(t)=\widehat{G}\widehat{y}(t)\).
(Necessity) Suppose that system (25) is stabilizable to \(\widehat{x}_{e}=\delta_{n_{1}}^{\alpha}\) by an output feedback control, say, \(\widehat{u}(t)=\widehat{G}\widehat{y}(t)\). Then, we can find the smallest integer \(1\leq\tau\leq n_{1}\) such that (40) holds for any \(\widehat{x}(0)\in\widehat{S}_{x}\setminus\{ \delta_{n_{1}}^{0}\}\) and any integer \(t\geq\tau\). Hence, \((\widehat {Q}\widehat{G}\widehat{H}M_{r,n_{1}})^{\tau}=\delta_{n_{1}}[\alpha ,\alpha,\ldots,\alpha]\).
Split \(\widehat{Q}\widehat{G}\widehat{H}M_{r,n_{1}}\in\mathcal {L}_{n_{1}\times n_{1}}\) into the following blocks:
where \(\lambda\in\{0,1\}\), \(A_{1}\in\mathcal{B}_{(\alpha1)\times (\alpha1)}\), \(A_{2}\in\mathcal{B}_{(\alpha1)\times(n_{1}\alpha )}\), \(A_{3}\in\mathcal{B}_{(n_{1}\alpha)\times(\alpha1)}\), \(A_{4}\in\mathcal{B}_{(n_{1}\alpha)\times(n_{1}\alpha)}\), \(\zeta_{1}\) and \(\zeta_{2}\) are some proper Boolean row vectors, and \(\zeta_{3}\) and \(\zeta_{4}\) are some proper Boolean column vectors.
It is easy to see from \(\widehat{Q}\widehat{G}\widehat {H}M_{r,n_{1}}\widehat{x}_{e}=\widehat{x}_{e}\) that \(\lambda=1\), and \(\zeta_{3}\) and \(\zeta_{4}\) are zero column vectors.
In the following, we prove that
is a nilpotent matrix of degree τ.
We can easily see from \((\widehat{Q}\widehat{G}\widehat {H}M_{r,n_{1}})^{\tau}=\delta_{n_{1}}[\alpha,\alpha,\ldots,\alpha ]\) that \(\widehat{A}^{\tau}=0\), which shows that Â is a nilpotent matrix. If its degree is less than τ, then there exists a positive integer \(\tau'<\tau\) such that \(\widehat{A}^{\tau '}=0\), and thus
which is a contradiction to the minimality of τ. This completes the proof. □
Remark 5
Based on Theorem 3 and Algorithm 1 presented in [32], we can design an output feedback gain matrix, say, \(\widehat{G}=\delta _{r_{1}}[w_{1},w_{2},\ldots,w_{2^{p}}]\), under which system (25) is stabilizable to \(\widehat{x}_{e}=\delta _{n_{1}}^{\alpha}\). Then, the output feedback pinning control can be designed as \(u_{i}(t)=G_{i}y(t)\), \(i=1,\ldots,r\), with \(G_{1}\ast G_{2}\ast \cdots\ast G_{r}=G\), where \(G=\delta_{2^{r}}[j_{w_{1}},\ldots,j_{w_{2^{p}}}]\).
Remark 6
It should be pointed out that the pinning control design for the state feedback stabilization of Boolean networks was studied in [34], and a novel design procedure was established. Compared with [34], our main results have the following advantages: (i) we established some necessary and sufficient conditions for the pinning control design of both state feedback and output feedback stabilization problems, whereas [34] only considered the state feedback stabilization problem; (ii) our results are applicable to the pinning control design for the state feedback stabilization of constrained BCNs.
Illustrative examples
In the section, we give two illustrative examples to show how to use the obtained results to design pinning control for feedback stabilization of constrained BCNs.
Example 1
Consider the following Boolean model to simulate the λ phage, which is a virus growing on a bacterium. The virus can only follow one of two different pathways: lysogeny or lysis, after injecting chromosome into the bacterium cell. The molecular mechanism responsible for the lysogeny/lysis decision is known as λ switch [36]. For example, two genes, cI and cro, directly affect the decision. When cI is active (inactive) and cro is inactive (active), the phage is in the lysogenic (lytic) state, and whether lysogenic state will be established or not depends on five phage genes, cI, cro, cII, cIII, N, and the environmental state. More details can be found in [37]. Its dynamic can be described in the form
where \(u(t)\) is a binary input that represents whether environmental condition is favorable or not.
Letting \(x(t)=\mathit{cII}(t)\mathit{cIII}(t)N(t)\mathit{cI}(t)\mathit{cro}(t)\), we obtain the following algebraic form:
where
In this example, several environmental conditions including concentration of nutrition, growth rate, temperature, and multiplicity of infection can influence the cII and cIII genes. If the environmental conditions are favorable, then the cII and cIII genes are highly active, and the cII gene product turns the cI gene on. The cI gene inhibits all other genes including cro, and the lysogenic state is established. If the environmental conditions are not favorable, then the genes cII and cIII are not activated, the cro gene remains active, and its product represses the cI gene. Thus, the lytic state is established.
Due to the limitation of the environmental conditions, we constrain the state in \(S_{x}=\{\delta_{32}^{19},\delta_{32}^{23},\delta _{32}^{27},\delta_{32}^{31}\}\).
By Theorem 2 and Remark 4 we can design 2^{30} state feedback pinning controls that stabilize system (41) with \(S_{x}\) to the equilibrium \(\delta _{32}^{31}\) (the lytic state), and one of them is
Example 2
Consider the following BCN:
Setting \(x(t)=\ltimes_{i=1}^{3}x_{i}(t)\), \(u(t)=\ltimes _{i=1}^{2}u_{i}(t)\), and \(y(t)=\ltimes_{i=1}^{2}y_{i}(t)\), we obtain the following algebraic form:
where
and
We assume that \(S_{x}=\{\delta_{8}^{1},\delta_{8}^{2},\delta _{8}^{3},\delta_{8}^{4},\delta_{8}^{5},\delta_{8}^{8}\}\), and \(S_{u}=\{\delta_{4}^{1},\delta_{4}^{2},\delta_{4}^{3},\delta _{4}^{4}\}\). We aim to design an output feedback pinning control such that system (43) with \(S_{x}\) and \(S_{u}\) is stabilized to \(x_{e}=\delta_{8}^{2}\).
Based on Theorem 3 and Algorithm 1 presented in [32], we can obtain eight output feedback pinning control gain matrices; one of them is \(K_{1}=\delta_{2}[1\ 1\ 2\ 1]\), \(K_{2}=\delta_{2}[1\ 1\ 1\ 1]\).
Conclusion
In this paper, we have studied the pinning control design for feedback stabilization of constrained BCNs. We have obtained the constrained algebraic form for constrained BCNs with pinning control via the semitensor product of matrices. We have given a necessary and sufficient condition for the reachability of constrained BCNs. We have proposed a procedure for the pinning control design of state feedback stabilization of constrained BCNs. Moreover, we have presented a necessary and sufficient condition for the output feedback stabilization of constrained BCNs with pinning control. The study of two illustrative examples has shown that the new results obtained in this paper are very effective.
References
Karlebach, G, Shamir, R: Modelling and analysis of gene regulatory networks. Nat. Rev. Mol. Cell Biol. 9, 770780 (2008)
Wang, Z, Lam, J, Wei, G, Fraser, K, Liu, X: Filtering for nonlinear genetic regulatory networks with stochastic disturbances. IEEE Trans. Autom. Control 53, 24482457 (2008)
Kauffman, S: Metabolic stability and epigenesis in randomly constructed genetic nets. J. Theor. Biol. 22, 437467 (1969)
Akutsu, T, Hayashida, M, Ching, WK, Ng, MK: Control of Boolean networks: hardness results and algorithms for tree structured networks. J. Theor. Biol. 244, 670679 (2007)
Tournier, L, Chaves, M: Uncovering operational interactions in genetic networks using asynchronous Boolean dynamics. J. Theor. Biol. 260, 196209 (2009)
Pal, R, Datta, A, Bittner, ML, Dougherty, ER: Optimal infinite horizon control for probabilistic Boolean networks. IEEE Trans. Signal Process. 54, 23752387 (2006)
Xiao, Y: A tutorial on analysis and simulation of Boolean gene regulatory network models. Curr. Genomics 10, 511525 (2009)
Cheng, D, Qi, H, Li, Z: Analysis and Control of Boolean Networks: A SemiTensor Product Approach. Springer, London (2011)
Cheng, D, Qi, H, Zhao, Y: An Introduction to SemiTensor Product of Matrices and Its Applications. World Scientific, Singapore (2012)
Cheng, D, Qi, H: A linear representation of dynamics of Boolean networks. IEEE Trans. Autom. Control 55, 22512258 (2010)
Chen, H, Sun, J: Output controllability and optimal output control of statedependent switched Boolean control networks. Automatica 50, 19291934 (2014)
Fornasini, E, Valcher, ME: Observability, reconstructibility and state observers of Boolean control networks. IEEE Trans. Autom. Control 58, 13901401 (2013)
Feng, J, Yao, J, Cui, P: Singular Boolean networks: semitensor product approach. Sci. China Inf. Sci. (2012). doi:10.1007/s1143201246668
Laschov, D, Margaliot, M: Controllability of Boolean control networks via the PerronFrobenius theory. Automatica 48, 12181223 (2012)
Li, F: Global stability at a limit cycle of switched Boolean networks under arbitrary switching signals. Neurocomputing 133, 6366 (2014)
Li, F, Lu, X: Minimum energy control and optimalsatisfactory control of Boolean control network. Phys. Lett. A 377, 31123118 (2013)
Lu, J, Zhong, J, Ho, DWC, Tang, Y, Cao, J: On controllability of delayed Boolean control networks. SIAM J. Control Optim. 54, 475494 (2016)
Lu, J, Zhong, J, Tang, Y, Huang, T, Cao, J: Synchronization in outputcoupled temporal Boolean networks. Sci. Rep. 4, 6292 (2014)
Liu, Z, Wang, Y: Disturbance decoupling of mixvalued logical networks via the semitensor product method. Automatica 48, 18391844 (2012)
Zhang, L, Zhang, K: Controllability and observability of Boolean control networks with timevariant delays in states. IEEE Trans. Neural Netw. Learn. Syst. 24, 14781484 (2013)
Zhao, Y, Cheng, D, Qi, H: Inputstate incidence matrix of Boolean control networks and its applications. Syst. Control Lett. 59, 767774 (2010)
Zhong, J, Lu, J, Liu, Y, Cao, J: Synchronization in an array of outputcoupled Boolean networks with time delay. IEEE Trans. Neural Netw. Learn. Syst. 25, 22882294 (2014)
Li, H, Wang, Y, Xie, L: Output tracking control of Boolean control networks via state feedback: constant reference signal case. Automatica 59, 5459 (2015)
Guo, P, Wang, Y, Li, H: Algebraic formulation and strategy optimization for a class of evolutionary networked games via semitensor product method. Automatica 49, 33843389 (2013)
Cheng, D, He, F, Qi, H, Xu, T, He, F: Modeling, analysis and control of networked evolutionary games. IEEE Trans. Autom. Control 60, 24022415 (2015)
Liu, Z, Wang, Y, Cheng, D: Nonsingularity of feedback shift registers. Automatica 55, 247253 (2015)
Ay, F, Xu, F, Kahveci, T: Scalable steady state analysis of Boolean biological regulatory networks. PLoS ONE 4, e7992 (2009)
Li, R, Yang, M, Chu, T: State feedback stabilization for Boolean control networks. IEEE Trans. Autom. Control 58, 18531857 (2013)
Fornasini, E, Valcher, ME: On the periodic trajectories of Boolean control networks. Automatica 49, 15061509 (2013)
Li, H, Wang, Y: Output feedback stabilization control design for Boolean control networks. Automatica 49, 36413645 (2013)
Li, F, Yu, Z: Feedback control and output feedback control for the stabilization of switched Boolean networks. Int. J. Control 89, 19 (2015)
Bof, N, Fornasini, E, Valcher, ME: Output feedback stabilization of Boolean control networks. Automatica 57, 2128 (2015)
Lu, J, Zhong, J, Huang, C, Cao, J: On pinning controllability of Boolean control networks. IEEE Trans. Autom. Control (2015). doi:10.1109/TAC.2015.2478123
Li, F: Pinning control design for the stabilization of Boolean networks. IEEE Trans. Neural Netw. Learn. Syst. (2015). doi:10.1109/TNNLS.2015.2449274
Li, H, Wang, Y: Minimumtime state feedback stabilization of constrained Boolean control networks. Asian J. Control (2015). doi:10.1002/asjc.1234
Thieffry, D, Thomas, R: Dynamical behaviour of biological regulatory networks  II. Immunity control in bacteriophage lambda. Bull. Math. Biol. 57, 277297 (1995)
Laschov, D, Margaliot, M: Minimumtime control of Boolean networks. SIAM J. Control Optim. 51, 28692892 (2013)
Acknowledgements
The authors would like to thank the Associate Editor and the anonymous reviewers for their constructive comments and suggestions, which improved the quality of this paper. The research was supported by the National Natural Science Foundation of China under Grants 61374065 and 61503225, the Major International (Regional) Joint Research Project of the National Natural Science Foundation of China under Grant 61320106011, and the Natural Science Foundation of Shandong Province under Grant ZR2015FQ003.
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
QY and HL carried out the main part of this article, and YL corrected the manuscript and brought forward many suggestions on this article. All authors have read and approved the final manuscript.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Yang, Q., Li, H. & Liu, Y. Pinning control design for feedback stabilization of constrained Boolean control networks. Adv Differ Equ 2016, 182 (2016). https://doi.org/10.1186/s1366201609090
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s1366201609090
MSC
 93C55
 93B52
Keywords
 Boolean control network
 pinning control
 feedback stabilization
 constrained control
 semitensor product of matrices