Abstract
In this paper, we present a sufficient condition for the exponential stability of a class of linear switched systems. As an application of this stability result, we establish an output-based adaptive distributed observer for a general linear leader system over a periodic jointly connected switching communication network, which extends the applicability of the output-based adaptive distributed observer from a marginally stable linear leader system to any linear leader system and from an undirected switching graph to a directed switching graph. This output-based adaptive distributed observer will be applied to solve the leader-following consensus problem for multiple double-integrator systems.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The distributed observer approach is an effective tool for dealing with a variety of cooperative control problems of multi-agent systems, such as the leader-following consensus problem [1–3], the containment control problem [4], the cooperative output regulation problem [5–8], the leader-following output synchronization problem [9–11], and the rendezvous problem [12, 13].
A distributed observer for the leader system is a distributed dynamic compensator that asymptotically estimates the state of the leader system over a communication network. The distributed observer for linear leader systems was first established over a connected static network in [7] and then over a jointly connected switching network in [8]. However, the distributed observers in [7, 8] have two drawbacks. First, they assume all followers know the system matrix of the leader system. Second, they assume the state of the leader system is available. The first drawback was overcome in [14] by proposing the so-called adaptive distributed observer, which can estimate both the state and the system matrix of a neutrally stable leader system over a jointly connected switching network. The adaptive distributed observer in [14] was strengthened to handle a general linear leader system provided that the switching signal is periodic in [15]. The second drawback was addressed in [16] by proposing the so-called output-based distributed observer which only makes use of the output of the leader system. By combining the ideas of the adaptive distributed observer in [14] and the output-based distributed observer in [16], reference [17] further developed the output-based adaptive distributed observer such that it can estimate the state of a leader system over a jointly connected switching network based on the output of the leader system only and without requiring all followers to know the system matrix of the leader system. Nevertheless, the existing output-based adaptive distributed observer over a jointly connected switching network still requires that the system matrix of the leader system be marginally stable [17]. This requirement is restrictive as it cannot even handle the frequently encountered double-integrator systems. For this reason, in this paper, we will explore the possibility of establishing an output-based adaptive distributed observer for a general linear leader system over a jointly connected switching network. Indeed, by establishing a key stability result of a class of linear periodic switched systems, we manage to ascertain the existence of an output-based adaptive distributed observer for a general linear leader system over a periodic jointly connected switching network.
The rest of this paper is organized as follows. We summarize some useful stability results for periodic linear switched systems and establish a key stability result of a class of linear switched systems in Sect. 2. In Sect. 3, we apply the stability result to develop an output-based adaptive distributed observer for a general linear leader system. In Sect. 4, we apply the output-based adaptive distributed observer to a leader-following consensus problem for multiple double-integrator systems. Section 5 closes this paper with some remarks.
Notation
\(\mathbb{R}\) denotes the set of real numbers. \(\Re (\cdot )\) denotes the real part of a complex number. \(\|x\|\) denotes the Euclidean norm of vector x and \(\|A\|\) denotes the induced norm of matrix A by the Euclidean norm. For a square matrix A, \(\bar{\lambda}_{\min}(A)\) denotes an eigenvalue of A with the smallest real part and \(\bar{\lambda}_{\max}(A)\) denotes an eigenvalue of A with the greatest real part. \(\boldsymbol{1}_{N}\) denotes the N-dimensional column vector with all elements being 1. For \(A_{i}\in \mathbb{R}^{p_{i}\times m}\), \(i=1,\ldots,n\), \(\operatorname{col}(A_{1},\ldots,A_{n})= \begin{bmatrix} A_{1}^{T} & \cdots & A_{n}^{T} \end{bmatrix} ^{T}\). ⊗ denotes the Kronecker product of matrices. block \(\operatorname{diag}(A_{1},\ldots , A_{n})\) denotes a block diagonal matrix whose diagonal block elements are \(A_{1},\ldots ,A_{n}\).
2 Preliminaries
Let \(\sigma (t) : [0,\infty ) \to \mathcal{P} = \{1, 2, \ldots , \rho \}\) be a piecewise constant switching signal with dwell time \(\tau _{0}>0\). Consider a switching digraphFootnote 1\(\bar{\mathcal{G}}_{\sigma (t)}=(\bar{\mathcal{V}}, \bar{\mathcal{E}}_{ \sigma (t)})\) where \(\bar{\mathcal{V}}=\{0, 1, \ldots , N\}\) and \(\bar{\mathcal{E}}_{\sigma (t)}\subseteq \bar{\mathcal{V}}\times \bar{\mathcal{V}}\) for \(t\ge 0\). Assume that the switching digraph \(\bar{\mathcal{G}}_{\sigma (t)}\) satisfies the following assumptions.
Assumption 1
There exists a subsequence \(\{i_{k}| k=0,1, \ldots \}\) of \(\{i| i=0,1,\ldots \}\) with \(t_{i_{k+1}}-t_{i_{k}}<\nu \) for some \(\nu >0\), such that the union graph \(\bigcup_{j=i_{k}}^{i_{k+1}-1}\bar{\mathcal{G}}_{\sigma (t_{j})}\) has the property that there is a path from node 0 to every other node.
Assumption 2
The switching signal \(\sigma (t)\) is periodic with period T.
Remark 2.1
Assumption 1 is called the jointly connected condition which allows the switching digraph to be disconnected at any time instant. As noted in [15], under Assumption 2, we can assume without loss of generality that the switching signal takes the form
where \(s=0, 1, \ldots \) , and \(\omega _{p}\), \(p=1,\ldots ,\rho \), are positive constants satisfying \(\sum_{p=1}^{\rho }\omega _{p}=1\). Under Assumptions 1 and 2, without loss of generality, we may assume that, for any \(k=0,1,\ldots \) , the union graph \(\bigcup_{j=i_{k}}^{i_{k+1}-1}\bar{\mathcal{G}}_{\sigma (t_{j})}= \bigcup_{p=1}^{\rho }\bar{\mathcal{G}}_{p}\). A switching digraph \(\bar{\mathcal{G}}_{\sigma (t)}\) satisfying Assumptions 1 and 2 is said to be periodic jointly connected.
Remark 2.2
Let \(H_{\sigma (t)} \in \mathbb{R}^{N\times N}\) be obtained from the Laplacian matrix \(\bar{\mathcal{L}}_{\sigma (t)}\) of \(\bar{\mathcal{G}}_{\sigma (t)}\) by removing the first row and the first column of \(\bar{\mathcal{L}}_{\sigma (t)}\). Under Assumption 1, by Corollary 4 of [8], the origin of the linear switched system
where μ is an arbitrary positive real number and \(\phi \in \mathbb{R}^{Nq}\), is exponentially stable. Moreover, under Assumptions 1 and 2, consider the switching signal \(\sigma (t)\) defined in (1). From the proof of Lemma 2.2 of [15], all the eigenvalues of the matrix \(\sum_{p=1}^{\rho} \omega _{p} H_{p}\) have positive real parts.
Let
The following lemma is rephrased from Lemma 2.2 of [15].
Lemma 2.1
Consider the linear switched system as follows:
where \(\sigma (t)\) is defined in (1); \(\vartheta \in \mathbb{R}^{Nn}\) is the state; \(G\in \mathbb{R}^{n\times n}\) is a constant matrix; and \(\mu \in \mathbb{R}\) is a positive real number. Under Assumptions 1and 2, there exists a positive constant \(\bar{T}_{0}\) such that, for any \(\mu >\bar{\mu}_{0}\) and any \(0< T<\bar{T}_{0}\), the origin of system (4) is exponentially stable.
In what follows, we establish a stability result of the following class of linear switched systems:
where N is a positive integer; μ is a positive real number; \(x\in \mathbb{R}^{Nq}\) is the state; \(S\in \mathbb{R}^{q \times q}\) and \(C\in \mathbb{R}^{p \times q}\) are constant matrices; and \(L \in \mathbb{R}^{q\times p}\) is a gain matrix to be designed.
Assumption 3
The pair \((C, S)\) is detectable.
Remark 2.3
Under Assumption 3 and the assumptions that S is marginally stable and the matrices \(H_{p}\in \mathbb{R}^{N \times N}\), \(p\in \mathcal{P}\) with \(\mathcal{P}\) being the switching index set, are positive semi-definite, the asymptotic stability of system (5) was established in [16] and the exponential stability of system (5) was established in [17].
Remark 2.4
Under Assumption 3, the following algebraic Riccati equation:
admits a unique positive definite solution \(P\in \mathbb{R}^{q \times q}\) [18].
Let us first establish the exponential stability of system (5) with S being a general square matrix and \(\sigma (t)\) being a periodic signal.
Lemma 2.2
Under Assumptions 1, 2, and 3, let \(\lambda _{1},\ldots ,\lambda _{N}\) be the N eigenvalues of the matrix \(\sum_{p=1}^{\rho} \omega _{p} H_{p}\), let \(\bar{\mu}_{1} = \frac{1}{2}\delta ^{-1}\) where \(\delta =\min_{i=1,\ldots ,N}\{\Re (\lambda _{i})\}\). Then, there exists a positive constant \(\bar{T}_{1}\) such that, for any \(\mu \ge \bar{\mu}_{1}\) and any \(0< T<\bar{T}_{1}\), the origin of system (5) with \(L=PC^{T}\) is exponentially stable.
Proof
Let
Under Assumptions 1 and 2, by Remark 2.2, the N eigenvalues \(\lambda _{1},\ldots ,\lambda _{N}\) of the matrix \(\sum_{p=1}^{\rho} \omega _{p} H_{p}\) have positive real parts. Let \(Q\in \mathbb{R}^{N\times N}\) be a nonsingular matrix such that \(J=Q (\sum_{p=1}^{\rho} \omega _{p} H_{p} )Q^{-1}\) where \(J\in \mathbb{R}^{N\times N}\) is the Jordan form of \(\sum_{p=1}^{\rho} \omega _{p} H_{p}\). Since \(\sum_{p=1}^{\rho }\omega _{p}=1\), \(M_{c}\) can be rewritten as
The block triangular structure of J implies that the eigenvalues of \(M_{c}\) coincide with the eigenvalues of \(S-\mu \lambda _{i} PC^{T}C\), \(i=1,\ldots ,N\). By Lemma 2.12 of [17], for any \(\mu \ge \frac{1}{2}\delta ^{-1}\) with \(\delta =\min_{i=1,\ldots ,N}\{\Re (\lambda _{i})\}\), the matrices \(S-\mu \lambda _{i} PC^{T}C\), \(i=1,\ldots ,N\), are Hurwitz. Hence, for any \(\mu \ge \frac{1}{2}\delta ^{-1}\), \(M_{c}\) is Hurwitz. Invoking Lemma 3.22 of [19] with \(\mu \ge \bar{\mu}_{1}= \frac{1}{2}\delta ^{-1}\) completes the proof. □
Remark 2.5
Since Lemma 2.2 imposes no restriction on the eigenvalues of the matrix S and does not require \(H_{p}\), \(p=1,\ldots ,\rho \), to be symmetric, it has partially extended the results in [16, 17] from a marginally stable matrix S to any matrix S and from an undirected graph to a directed graph provided that the switching signal \({\sigma (t)}\) satisfies Assumption 2 with the period T being smaller than some threshold.
3 Output-based adaptive distributed observer
In this section, we further apply Lemma 2.2 to establish the output-based adaptive distributed observer over a jointly connected graph for a general linear leader system of the following form:
where \(v\in \mathbb{R}^{q}\), \(y\in \mathbb{R}^{p}\), and \(S\in \mathbb{R}^{q\times q}\) and \(C\in \mathbb{R}^{p\times q}\) are constant matrices.
As in [17], for \(i=1,\ldots ,N\), consider a distributed dynamic compensator as follows:
where \(S_{0}=S\), \(C_{0}=C\), \(L_{0}=L\in \mathbb{R}^{q\times p}\) is a gain matrix to be designed, and \(\xi _{0}=v\); \(\mu _{1}\) to \(\mu _{4}\) are positive real numbers to be specified; and, for \(i=1,\ldots ,N\), \(S_{i} \in \mathbb{R}^{q \times q}\), \(C_{i} \in \mathbb{R}^{p \times q}\), \(L_{i} \in \mathbb{R}^{q \times p}\), \(\xi _{i} \in \mathbb{R}^{q}\), and \(a_{ij}(t)\) is the element of the weighted adjacency matrix.
Given the leader system (9) and a group of N dynamic compensators (10a)–(10d), we can define a switching digraph \(\bar{\mathcal{G}}_{\sigma (t)}=(\bar{\mathcal{V}}, \bar{\mathcal{E}}_{ \sigma (t)})\) where \(\bar{\mathcal{V}}=\{0, 1, \ldots , N\}\) and \(\bar{\mathcal{E}}_{\sigma (t)}\subseteq \bar{\mathcal{V}}\times \bar{\mathcal{V}}\) for \(t\ge 0\). In the set \(\bar{\mathcal{V}}\), node 0 is associated with the leader system (9) and node i, \(i=1, \ldots , N\), is associated with the ith dynamic compensator of (10a)–(10d). For \(i=1, \ldots , N\), \(j=0, 1, \ldots , N\), \(i \neq j\), \((j, i)\in \bar{\mathcal{E}}_{\sigma (t)}\) if and only if the ith dynamic compensator of (10a)–(10d) can make use of the information of agent j at time t. The weighted adjacency matrix of the digraph \(\bar{\mathcal{G}}_{\sigma (t)}\) is a nonnegative matrix \(\bar{\mathcal{A}}_{\sigma (t)}=[a_{ij}(t)]_{i,j=0}^{N}\in \mathbb{R}^{(N+1) \times (N+1)}\), where, for \(t\ge 0\), \(a_{ii}(t)=0\), and, for \(i\neq j\), \(a_{ij}(t)>0 \Leftrightarrow (j, i)\in \bar{\mathcal{E}}_{\sigma (t)}\). Let \(H_{\sigma (t)}=[h_{ij}(t)]_{i,j=1}^{N}\in \mathbb{R}^{N\times N}\), where \(h_{ii}(t)=\sum_{j=0}^{N}a_{ij}(t)\) and \(h_{ij}(t)=-a_{ij}(t)\) if \(i\neq j\), then \(H_{\sigma (t)}\) is the matrix obtained from the Laplacian matrix \(\bar{\mathcal{L}}_{\sigma (t)}\) of \(\bar{\mathcal{G}}_{\sigma (t)}\) by removing the first row and the first column of \(\bar{\mathcal{L}}_{\sigma (t)}\). Let \(\mathcal{G}_{\sigma (t)}=(\mathcal{V}, \mathcal{E}_{\sigma (t)})\) with \(\mathcal{V}=\{1, \ldots , N\}\) and \(\mathcal{E}_{\sigma (t)}\subseteq \mathcal{V}\times \mathcal{V}\) be the subgraph obtained from the switching digraph \(\bar{\mathcal{G}}_{\sigma (t)}\) by removing all the edges between node 0 and the nodes in \(\mathcal{V}\).
Assumption 4
The subgraph \(\mathcal{G}_{\sigma (t)}\) is undirected for all \(t\ge 0\).
Assumption 5
The matrix S is marginally stable, that is, there exists a positive definite matrix \(P_{0}\) satisfying \(P_{0}S^{T}+SP_{0}\le 0\).
The dynamic compensator (10a)–(10d) is a so-called output-based adaptive distributed observer for the leader system (9) due to the following lemma rephrased from Theorem 4.7 of [17].
Lemma 3.1
Consider systems (9) and (10a)–(10d). Under Assumptions 1, 3, 4, and 5, there exists a constant matrix \(L\in \mathbb{R}^{q\times p}\) such that, for any \(\mu _{1}, \mu _{2}, \mu _{3}, \mu _{4}>0\) and any initial conditions \(v(0)\), \(S_{i}(0)\), \(C_{i}(0)\), \(L_{i}(0)\), and \(\xi _{i}(0)\), \(i=1,\ldots ,N\), the solutions of systems (9) and (10a)–(10d) exist for all \(t\ge 0\) and satisfy
all exponentially.
However, Lemma 3.1 is quite restrictive since it only applies to a marginally stable leader system and requires the graph to be undirected. Thus, it is desirable to establish an output-based adaptive distributed observer for the leader system (9) without these restrictive assumptions, provided that the switching signal \({\sigma (t)}\) satisfies Assumption 2 with the period T being smaller than some threshold.
Theorem 3.1
Consider systems (9) and (10a)–(10d). Under Assumptions 1, 2, and 3, let \(\lambda > \Re (\bar{\lambda}_{\max}(S))\) be some positive real number and let
Then there exists a positive constant \(\bar{T}_{2}\), such that, if \(\mu _{1},\mu _{2}, \mu _{3}>\bar{\mu}_{2}\), \(\mu _{4}\ge \bar{\mu}_{3}\), and \(0< T<\bar{T}_{2}\), then, for any initial conditions \(v(0)\), \(S_{i}(0)\), \(C_{i}(0)\), \(L_{i}(0)\), and \(\xi _{i}(0)\), \(i=1,\ldots ,N\), the solutions of systems (9) and (10a)–(10d) with \(L=PC^{T}\) exist for all \(t\ge 0\) and satisfy
all exponentially.
Proof
Let \(\tilde{S}_{i}= S_{i}-S\), \(\tilde{C}_{i}= C_{i}-C\), \(\tilde{L}_{i}= L_{i}-L\), and \(\tilde{v}_{i}= \xi _{i}-v\), \(i=1,\ldots ,N\). Also, let \(\tilde{S}= \operatorname{col}(\tilde{S}_{1},\ldots ,\tilde{S}_{N})\), \(\tilde{C}= \operatorname{col}(\tilde{C}_{1},\ldots ,\tilde{C}_{N})\), \(\tilde{L}= \operatorname{col}(\tilde{L}_{1},\ldots ,\tilde{L}_{N})\), \(\tilde{v}= \operatorname{col}(\tilde{v}_{1},\ldots ,\tilde{v}_{N})\), \(\tilde{S}_{d}= \text{block diag}(\tilde{S}_{1}, \ldots ,\tilde{S}_{N})\), \(\tilde{C}_{d}= \text{block diag}(\tilde{C}_{1},\ldots ,\tilde{C}_{N})\), \(\tilde{L}_{d}= \text{block diag}(\tilde{L}_{1}, \ldots ,\tilde{L}_{N})\), and \(L_{d}= \text{block diag}(L_{1},\ldots ,L_{N})\). Then, from (9) and (10a)–(10d), as shown in Sect. 4.4.3 of [17], S̃, C̃, L̃, and ṽ are governed by
where
For \(i=1,\ldots ,q\), let \(\tilde{s}_{i}\in \mathbb{R}^{Nq}\) and \(\tilde{c}_{i}\in \mathbb{R}^{Np}\) be the ith column of S̃ and C̃, respectively. For \(j=1,\ldots ,p\), let \(\tilde{l}_{j}\in \mathbb{R}^{Nq}\) be the jth column of L̃. Then, (15a), (15b), and (15c) can be written as
Under Assumption 1, by Remark 2.2, the origin of system (17a)–(17c) is exponentially stable for any \(\mu _{1}, \mu _{2}, \mu _{3}>0\). Thus, we have \(\lim_{t\to \infty}\tilde{S}_{d}(t)=0\), \(\lim_{t\to \infty}\tilde{C}_{d}(t)=0\), and \(\lim_{t\to \infty}\tilde{L}_{d}(t)=0\) all exponentially, which implies (14a), (14b), and (14c). Since (14c) implies that \(L_{d}(t)\) is bounded over \([0, \infty )\), we have \(\lim_{t\to \infty}M_{d}(t)=0\) exponentially.
In what follows, we show (14d). For this purpose, we first show that there exists a positive constant \(\bar{T}_{3}\), such that, for any \(\mu _{1},\mu _{2}, \mu _{3}>\bar{\mu}_{2}\) and any \(0< T<\bar{T}_{3}\), \(\lim_{t\to \infty} F(t)=0\) exponentially, where \(F(t)\) is defined in (16). Let \(\bar{S}= I_{q}\otimes \lambda \), \(\hat{S}= I_{p}\otimes \lambda \), \(\zeta _{i}= (I_{N}\otimes e^{\bar{S}t})\tilde{s}_{i}\), and \(\gamma _{i}= (I_{N}\otimes e^{\hat{S}t})\tilde{c}_{i}\), \(i=1,\ldots ,q\). Then, \(\zeta _{i}\) is governed by
Similarly, \(\gamma _{i}\) is governed by
For all \(i=1,\ldots ,q\), system (18) (respectively, system (19)) is in the form of (4) with \(G=\bar{S}\), \(\mu =\mu _{1}\), and \(\vartheta =\zeta _{i}\) (respectively, \(G=\hat{S}\), \(\mu =\mu _{2}\), and \(\vartheta =\gamma _{i}\)). Since \(\Re (\bar{\lambda}_{\max}(\bar{S}))=\Re (\bar{\lambda}_{\max}( \hat{S}))=\lambda \), under Assumptions 1 and 2, by Lemma 2.1, there exists a positive constant \(\bar{T}_{3}\), such that, for any \(\mu _{1}, \mu _{2}>\bar{\mu}_{2}\) and any \(0< T<\bar{T}_{3}\), the origins of systems (18) and (19) are exponentially stable. Thus, for \(i=1,\ldots ,N\), we have
exponentially. Since \(\| e^{St} \| \le \beta e^{\lambda t}\) for some positive real number β, from (20) and (21), for \(i=1,\ldots ,N\), we have
exponentially. Thus, it holds that
exponentially. Since \(L_{d}(t)\) is bounded over \([0, \infty )\), from (16), (24), and (25), we have \(\lim_{t\to \infty} F(t)=0\) exponentially.
Now, we are ready to show (14d). For this purpose, consider the following system:
which is obtained from system (15d) by setting \(M_{d}(t)\) and \(F(t)\) to zero. Under Assumptions 1, 2, and 3, by Lemma 2.2, there exists a positive constant \(\bar{T}_{4}\), such that, for any \(\mu _{4}\ge \bar{\mu}_{3}\) and any \(0< T<\bar{T}_{4}\), the origin of system (26) is exponentially stable. Let \(\bar{T}_{2}= \min \{\bar{T}_{3}, \bar{T}_{4}\}\). Applying Lemma 1 of [20] to system (15d) completes the proof. □
Remark 3.1
The exponential stability of the linear switched system (26) with a general square matrix S is crucial in establishing the output-based adaptive distributed observer (10a)–(10d). This stability result of (26) cannot be obtained by any existing approach. Theorem 3.1 has made the output-based adaptive distributed observer (10a)–(10d) applicable to any linear leader system over a directed switching graph \(\bar{\mathcal{G}}_{\sigma (t)}\) that satisfies Assumption 2 with the period T being smaller than a threshold.
4 Example
In this section, we apply the output-based adaptive distributed observer (10a)–(10d) to solve the leader-following consensus and disturbance rejection problem of four double-integrator systems as follows:
where, for \(i=1,2,3,4\), \(x_{i1}, x_{i2}\in \mathbb{R}\); \(u_{i}\in \mathbb{R}\) is the control input to be designed; \(d_{i}=i\times v_{3}\in \mathbb{R}\) is the external disturbance; \(e_{i}\in \mathbb{R}^{2}\) is the tracking error; \(y_{mi}\in \mathbb{R}\) is the measurement output; and \(v=\operatorname{col}(v_{1},v_{2},v_{3})\in \mathbb{R}^{3}\) is the exogenous signal generated by the following leader system:
Hence, the four double-integrator systems are subject to exponentially growing disturbances. Since the pair \((C, S)\) is observable, Assumption 3 is satisfied.
We can interpret the above problem as a cooperative output regulation problem (see Definition 1 of [8]) by rewriting the four double-integrator systems in the following standard form:
where, for \(i=1,2,3,4\), \(x_{i}=\operatorname{col}(x_{i1}, x_{i2})\),
Then it can be verified that the pair \((A_{i}, B_{i})\) is controllable; the pair \((C_{mi}, A_{i})\) is observable; and the solution of the regulator equation
is given by and \(U_{i}= [ -1 \ 0 \ - i ] \).
Consider the following periodic switching signal:
where \(T=1\) and \(s=0, 1, \ldots \) . Thus, \(\sigma (t)\) is in the form of (1) with \(\omega _{1}=\frac{1}{4}\), \(\omega _{2}=\frac{1}{4}\), and \(\omega _{3}=\frac{1}{2}\). The three digraphs \(\bar{\mathcal{G}}_{i}\), \(i=1, 2, 3\), associated with \(\sigma (t)\) are shown in Fig. 1. Although the switching digraph \(\bar{\mathcal{G}}_{\sigma (t)}\) is disconnected at every time instant \(t\ge 0\), it can be seen that Assumptions 1 and 2 are satisfied. We let \(a_{ij}(t)=1\) whenever \((j,i)\in \bar{\mathcal{E}}_{\sigma (t)}\).
The three matrices associated with the three digraphs \(\bar{\mathcal{G}}_{i}\), \(i=1,2,3\), are given by
and
The four eigenvalues of the matrix \(\sum_{p=1}^{3} \omega _{p} H_{p}\) are \(\lambda _{1} = 0.5\), \(\lambda _{2} = 0.5\), \(\lambda _{3} = 0.25\), and \(\lambda _{4} = 0.25\), thus \(\Re (\bar{\lambda}_{\min} (\sum_{p=1}^{3} \omega _{p} H_{p} ) )=\min_{i=1,2,3,4}\{\Re (\lambda _{i})\}=0.25\). From (13), \(\bar{\mu}_{3}=2\). From (28), \(\Re (\bar{\lambda}_{\max}(S))=1\). Then we let \(\lambda = 1.2\) which satisfies \(\lambda > \Re (\bar{\lambda}_{\max}(S))\). From (12), we have \(\bar{\mu}_{2}=4.8\).
By Theorem 3.1, we can design an output-based adaptive distributed observer of the form (10a)–(10d) for (28) with \(\mu _{1}=10\), \(\mu _{2}=10\), \(\mu _{3}=10\), \(\mu _{4}=10\), and \(L=\operatorname{col}(-0.2, 1.4, 4.2)\). As shown in [8], the cooperative output regulation problem of the multi-agent system composed of (28) and (29) can be solved by the following dynamic measurement output feedback control law:
where, for \(i=1,2,3,4\), \(\xi _{i}\in \mathbb{R}^{3}\) is generated by (10a)–(10d), \(L_{mi}=\operatorname{col}(-22,-120)\), \(K_{1i} = [ -12 \ -7 ] \), and \(K_{2i} = U_{i}-K_{1i}X_{i}= [ 11 \ 7 \ -i ] \). In particular, for \(i=1,2,3,4\), \(K_{1i}\) and \(L_{mi}\) are such that \(A_{i}+B_{i}K_{1i}\) and \(A_{i}+L_{mi}C_{mi}\) are Hurwitz.
Our simulation is conducted with \(v(0)=\operatorname{col}(1,1,1)\) and other initial conditions randomly generated within the range \([-1.5, 1.5]\). The estimation errors of the distributed observer are shown in Figs. 2 to 5. The tracking errors of the followers are shown in Fig. 6. As expected, satisfactory performance is observed.
5 Conclusion
In this paper, having established an exponential stability property for a class of linear switched systems, we have further developed an output-based adaptive distributed observer for a general linear leader system over a periodic jointly connected switching communication network, which extends the applicability of the output-based adaptive distributed observer from a marginally stable linear leader system to a general linear leader system and from an undirected switching graph to a directed switching graph.
Availability of data and materials
Not applicable.
Notes
See Sect. 1.2 of [17] for an overview on digraphs.
References
A. Jadbabaie, J. Lin, A.S. Morse, Coordination of groups of mobile autonomous agents using nearest neighbor rules. IEEE Trans. Autom. Control 48(6), 988–1001 (2003)
W. Ni, D. Cheng, Leader-following consensus of multi-agent systems under fixed and switching topologies. Syst. Control Lett. 59(3), 209–217 (2010)
Y. Dong, Z. Lin, An event-triggered observer and its applications in cooperative control of multiagent systems. IEEE Trans. Autom. Control 67(7), 3647–3654 (2022)
S. Zuo, Y. Song, F.L. Lewis, A. Davoudi, Adaptive output containment control of heterogeneous multi-agent systems with unknown leaders. Automatica 92, 235–239 (2018)
H. Liu, C. De Persis, M. Cao, Robust decentralized output regulation with single or multiple reference signals for uncertain heterogeneous systems. Int. J. Robust Nonlinear Control 25(9), 1399–1422 (2015)
G.S. Seyboth, W. Ren, F. Allgöwer, Cooperative control of linear multi-agent systems via distributed output regulation and transient synchronization. Automatica 68, 132–139 (2016)
Y. Su, J. Huang, Cooperative output regulation of linear multi-agent systems. IEEE Trans. Autom. Control 57(4), 1062–1066 (2012)
Y. Su, J. Huang, Cooperative output regulation with application to multi-agent consensus under switching network. IEEE Trans. Syst. Man Cybern., Part B, Cybern. 42(3), 864–875 (2012)
Y. Yang, H. Modares, D.C. Wunsch, Y. Yin, Leader-follower output synchronization of linear heterogeneous systems with active leader using reinforcement learning. IEEE Trans. Neural Netw. Learn. Syst. 29(6), 2139–2153 (2018)
L. Zhu, Z. Chen, R.H. Middleton, A general framework for robust output synchronization of heterogeneous nonlinear networked systems. IEEE Trans. Autom. Control 61(8), 2092–2107 (2016)
P. Zhou, B.M. Chen, Semi-global leader-following output consensus of heterogeneous systems subject to actuator position and rate saturation. Auton. Intell. Syst. 1(1), 91–103 (2021)
Y. Dong, J. Chen, Adaptive control for rendezvous problem of networked uncertain Euler–Lagrange systems. IEEE Trans. Cybern. 49(6), 2190–2199 (2019)
Y. Dong, S. Xu, Rendezvous with connectivity preservation problem of linear multiagent systems via parallel event-triggered control strategies. IEEE Trans. Cybern. 52(5), 2725–2734 (2022)
H. Cai, J. Huang, The leader-following consensus for multiple uncertain Euler-Lagrange systems with an adaptive distributed observer. IEEE Trans. Autom. Control 61(10), 3152–3157 (2016)
C. He, J. Huang, Adaptive distributed observer for general linear leader systems over periodic switching digraphs. Automatica 137, 110021
Y. Su, J. Huang, Stability of a class of linear switching systems with applications to two consensus problems. IEEE Trans. Autom. Control 57(6), 1420–1430 (2012)
H. Cai, Y. Su, J. Huang, Cooperative Control of Multi-Agent Systems: Distributed-Observer and Distributed-Internal-Model Approaches (Springer, Berlin, 2022)
S.E. Tuna, LQR-based coupling gain for synchronization of linear systems. Available at. http://arxiv.org/abs/0801.3390
Z. Sun, S.S. Ge, Switched Linear Systems: Control and Design (Springer, London, 2005)
T. Liu, J. Huang, Leader-following attitude consensus of multiple rigid body systems subject to jointly connected switching networks. Automatica 92, 63–71 (2018)
Acknowledgements
The authors would like to thank the anonymous reviewers for their valuable comments and suggestions.
Funding
This work was supported in part by the Research Grants Council of the Hong Kong Special Administrative Region under Grant Nos. 14202619 and PDFS2223-4S02, and in part by the National Natural Science Foundation (NSFC) of China under Grant No 61973260.
Author information
Authors and Affiliations
Contributions
CH and JH participated in the framework design and manuscript writing. JH is the supervisor of CH at The Chinese University of Hong Kong. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
He, C., Huang, J. Output-based adaptive distributed observer for general linear leader systems over periodic switching digraphs. Auton. Intell. Syst. 3, 1 (2023). https://doi.org/10.1007/s43684-023-00046-6
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s43684-023-00046-6