Abstract
The relation between passive and positive real systems has been extensively studied in the literature. In this paper, we study their connection to the more recently used notion of port-Hamiltonian descriptor systems. It is well-known that port-Hamiltonian systems are passive and that passive systems are positive real. Hence it is studied under which assumptions the converse implications hold. Furthermore, the relationship between passivity and KYP inequalities is investigated.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Port-Hamiltonian (pH) systems have been increasingly used in recent years as a unified structured framework for energy-based modeling of systems, see e.g. [1,2,3,4,5,6,7]. The pH formulation has gained interest from engineers and mathematicians due to its modeling flexibility and robustness properties [2, 8,9,10]. Specifically, pH systems are used in coupled networks of systems and multiphysics simulation and control. System coupling often imposes additional algebraic constraints on the system which naturally lead to linear time-invariant descriptor systems in the state-space form presented as
where \(u: \mathbb {R}\rightarrow \mathbb {K}^m\), \(x: \mathbb {R}\rightarrow \mathbb {K}^n\), \(y: \mathbb {R}\rightarrow \mathbb {K}^m\) are the input, state, and output of the system, and \(E,A\in \mathbb {K}^{n\times n}, B\in \mathbb {K}^{n\times m}\), \(C\in \mathbb {K}^{m\times n}\), \(D\in \mathbb {K}^{m\times m}\), and \(\mathbb {K}=\mathbb {R}\) or \(\mathbb {K}=\mathbb {C}\). The system (1) will be concisely denoted by \(\Sigma =(E,A,B,C,D)\) and throughout it is assumed that the pair (E, A) is regular which means that \(\lambda E-A\) is invertible for some \(\lambda \in \mathbb {C}\).
In addition, we consider the following notation: For a matrix \(A\in \mathbb {K}^{n \times m}\) let \(A^{\top },A^H,A^{-H}\) denote the transpose, conjugate transpose, inverse of \(A^H\), respectively. Note that in the real case \(A^\top = A^H\). The identity matrix of dimension n is denoted by \(I_n\). For a Hermitian matrix \(A \in \mathbb {K}^{n \times n}\), we use \(A > 0\) \((A \ge 0)\) if A is positive (semi-) definite. Furthermore, we denote the set of eigenvalues of a matrix pencil \(sE-A\) by
PH systems are then defined as follows:
-
(pH)
The system \(\Sigma \) is port-Hamiltonian if there exists \(J,R,Q\in \mathbb {K}^{n\times n}\), \(G,P\in \mathbb {K}^{n\times m}\), and \(S,N\in \mathbb {K}^{m\times m}\) such that
$$\begin{aligned} \begin{aligned} \begin{bmatrix} A&{}\quad B\\ C&{}\quad D \end{bmatrix}&=\begin{bmatrix}(J-R)Q&{}\quad G-P\\ (G+P)^HQ&{}\quad S+N\end{bmatrix},\quad Q^HE=E^HQ\ge 0,\\ \Gamma&:= \begin{bmatrix} J &{}\quad G \\ -G^H&{}\quad N\end{bmatrix} = - \Gamma ^H,\quad W := \begin{bmatrix} Q^HRQ &{}\quad Q^HP\\ P^HQ &{}\quad S \end{bmatrix} =W^H \ge 0. \end{aligned} \end{aligned}$$(2)
Here the quadratic function \(\mathcal {H}(x):=\frac{1}{2} x^HE^HQx\) is called the Hamiltonian which can often be interpreted as the energy of the system. Note that recently in [11,12,13] also a geometric pH framework was developed which is based on monotone, Dirac and Lagrangian subspaces and enlarges the class of pH systems. Furthermore, some references assume that the matrix Q in (2) is positive definite. In this case, the Q in the matrix W given by (2) is often replaced by the identity.
It is well-known that pH descriptor systems satisfy the following dissipation inequality which is referred to as passivity in the literature [14].
-
(Pa)
The system \(\Sigma \) is passive if there exists \(Q\in \mathbb {K}^{n\times n}\) such that \(Q^HE=E^HQ\) holds and if \(\mathcal {S}(x)=\tfrac{1}{2}x^H Q^HEx\), called storage function, satisfies for all \(T\ge 0\) the following inequality
$$\begin{aligned} \mathcal {S}(x(T))-\mathcal {S}(x(0))\le \int _{0}^{T} {{\,\textrm{Re}\,}}y(\tau )^Hu(\tau ) d \tau ,\quad \mathcal {S}(x(T))\ge 0, \end{aligned}$$(3)
for all consistent initial values \(x(0)=x_0\) and all functions x, u, y whose derivatives of arbitrary order \(k\in {\mathbb {N}}\) exist and fulfill (1).
Although we only consider in this paper smooth functions in the dissipation inequality (3), it can easily be extended to inputs u which are weakly differentiable up to some order by using the density of smooth functions in the Lebesgue space \(L^1([0,T],\mathbb {K}^m)\), see e.g. [15].
The property (Pa) is hard to verify in practice since one would have to consider all possible solution trajectories. It is more convenient to solve a linear matrix inequality called KYP (discovered independently by Kalman, Yakubovich and Popov) which is equivalent to (Pa) and can be obtained by differentiation of (3). An overview is presented in [16, p. 81] for standard systems, i.e. \(E=I_n\) and in [17, 18] for descriptor systems.
-
(KYP)
The system \(\Sigma \) has a solution \(Q\in \mathbb {K}^{n\times n}\) to the generalized KYP inequality if
$$\begin{aligned} \begin{bmatrix} -A^HQ-Q^HA&{}\quad C^H-Q^HB\\ C-B^HQ&{}\quad D+D^H \end{bmatrix}\ge 0,\quad E^HQ=Q^HE\ge 0. \end{aligned}$$(4)
In many applications, only input–output data is given and hence an important question is whether we can decide if a system is pH from this data and even more, we want to obtain a pH representation (2) of the system. The typical approach is to apply a Laplace transformation to (1) which leads to the transfer function
that describes the input–output behavior in the frequency domain. It is well-known for standard systems that the passivity implies that its transfer function is positive real, see [19, 20].
-
(PR)
The system \(\Sigma \) with transfer function \({\mathcal {T}}\) given by (5) is called positive real if \({\mathcal {T}}\) has no poles for all \(s\in \mathbb {C}\), \({{\,\textrm{Re}\,}}s>0\) and satisfies \({\mathcal {T}}(s)+{\mathcal {T}}(s)^H\ge 0\) for all \(s\in \mathbb {C}\) and \({{\,\textrm{Re}\,}}s>0\).
As mentioned above, it is well-known that pH descriptor systems are passive [14], denoted by (Pa), and that passive systems are positive real (PR). Moreover, the passivity is implied by the existence of solutions to KYP inequalities, see Proposition 1. The main goal of this note is to investigate under which assumptions also the converse implications hold. The study of these implications requires the use of controllability, observability and minimality notions.
Given a transfer function, \({\mathcal {T}}\), a realization is finding the matrices (E, A, B, C, D) in a descriptor state-space form (1) such that (5) is satisfied. In addition, the realization is called minimal if the number of states n in (1) needed to represent \({\mathcal {T}}\) is minimal.
A system is called controllable (observable) if and only if
In the case of standard systems, minimality is equivalent to the system being both controllable and observable.
Furthermore, one can define a weaker property stabilizability (detectability) of controllability (observability) such that
In [21, 22] the conditions on minimality were generalized to descriptor systems by showing that a realization (E, A, B, C, D) of a transfer function \({\mathcal {T}}\) is minimal if and only if it fulfills the following conditions
The first (second) property in (8) defines the behavioral controllability (behavioral observability) of the system. If the system fulfills, in addition to behavioral controllability, \({{\,\textrm{rk}\,}}[E,B]=n\), it is called completely controllable. If the system is behaviorally observable and \({{\,\textrm{rk}\,}}[E^\top ,C^\top ]=n\) holds, it is completely observable.
For standard systems, i.e. \(E=I_n\) in (1), which are controllable and observable, it is well-known that (pH), (Pa), (KYP), and (PR) are equivalent. For uncontrollable or unobservable standard systems a detailed study of the relation between (Pa), (KYP) and (PR) has been conducted in [16, Chapter 3] but without including (pH). The connection between (Pa) and (pH) was discussed in [23]. Another recent survey for standard systems was given in [24] see p. 59 therein for a discussion on unobservable and uncontrollable systems. An overview of standard systems is presented in Fig. 1.
For descriptor systems the relations between (pH), (Pa), (KYP) and (PR) were already studied in numerous works [17, 18, 25,26,27,28,29]. However, not all four properties have been investigated at the same time and often the minimality of the descriptor system is assumed.
As a first step, we combine the aforementioned results to obtain
which holds without observability or controllability assumptions and we provide examples showing that the converse implications do not hold.
Our aim is to provide sufficient conditions for the converse implications to hold. Hence the remaining questions which we will answer are
-
(Q1)
When do solutions to the KYP inequality lead to a pH formulation?
-
(Q2)
When does passivity lead to solutions of the KYP inequality and can we realize passive systems as pH systems?
-
(Q3)
Can every positive real transfer function be realized as a pH system?
The answer to question (Q1) is related to observability properties of the system. For standard systems it was shown in [23, p. 55] that only those solutions \(Q\in \mathbb {K}^{n\times n}\) to (KYP) which additionally satisfy
lead to a pH formulation. Conversely, the Q used in (pH) automatically satisfies (10). We show that the same condition is true for descriptor systems. If the system is behaviorally observable, then \(\ker A\cap \ker C=\{0\}\) and hence the existence of a (pH) is equivalent to the existence of invertible solutions to (KYP).
Due to the interesting properties of pH systems, ideally, we want to obtain (pH) for any passive system given in DAE form or given time domain data measurements [30]. In order to solve that problem we need to answer the question (Q2). This problem was already studied in [25, 27] where it was shown that (Pa) only guarantees (KYP) to hold on certain subspaces and as a consequence, we can only derive (pH) on a subspace. However, if the system has index at most one, we can derive a modified KYP inequality that is solvable for passive systems and which leads for behaviorally observable systems to a pH formulation of the system on the whole space.
The question (Q3) arises when one has to reconstruct a system from input–output data. If the system is expected to be passive, then the transfer function is positive real. If the data does not allow us to conclude (PR), e.g. due to measurement errors, then one can compute the nearest positive real transfer function [31]. Therefore, the remaining task is to find (pH) of this positive real transfer function, i.e. one has to compute the system matrices (E, A, B, C, D). We show that the computation of the system matrices is always possible for a given positive real transfer function and this is based on the well-known representation, see, e.g. [20, Section 5.1]
for some positive semidefinte \(M_1\in \mathbb {R}^{n\times n}\) and a proper positive real rational function \({\mathcal {T}}_p\). The summand \(M_1s\) can be realized as an index two subsystem which is combined with a pH realization based on a minimal realization of \({\mathcal {T}}_p\).
The paper is organized as follows: In Sect. 2 we summarize which relations between the basic notions (pH), (KYP), (Pa), (PR) are known for descriptor systems. In Sect. 3, it is shown that (10) can be used to define a pH realization which answers (Q1). Besides that, a possible generalization of the solutions of (KYP) is discussed. The answer to (Q1) will be used in Sect. 4 where we consider (Q2). In particular, we derive a pH formulation for passive index one systems that are (behaviorally) observable. Finally, in Sect. 5 we show how a pH formulation can be derived from a given real-valued positive real transfer function which answers (Q3). As a summary, an overview of the main results is presented in Fig. 2.
2 Literature review and combination of known results
Below, we give the first main result on the relationship between (pH), (Pa), (KYP) and (PR) for descriptor systems. Here we combine the results of [17], who studied (PR) and (KYP), and [31] who studied the relation between (PR), (pH) and (KYP) for invertible Q. In addition, we include the relation to passivity.
Proposition 1
Let (E, A, B, C, D) define a linear time-invariant descriptor system (1). Then the following holds
Furthermore, every \(Q\in \mathbb {K}^{n\times n}\) fulfilling (pH) is a solution to \((KYP )\) and every solution Q to \((KYP )\) leads to a storage function in (Pa). Moreover, the following holds
Proof
Step 1: We prove the implications (11). If (pH) holds for some Q, it solves the KYP inequality (4) since \(E^HQ=Q^HE\ge 0\) and
Hence Q fulfills (KYP). Next, we show that any \(Q\in \mathbb {K}^{n\times n}\) which fulfills (KYP) defines a storage function \(\mathcal {S}(x):=\tfrac{1}{2}x^HQ^HEx\) which fulfills (Pa). The basic idea goes back to [32] for standard systems. For sufficiently smooth u, consistent initial value \(x_0\) and for all \(t\ge 0\), it holds that
Integration of (16) leads to (Pa). Finally, it was shown in [17, Theorem 3.1] that (KYP) implies (PR). This completes the proof of (11).
Step 2: To prove (12), let Q be an invertible solution of (KYP). Then we can define a pH system via
Hence by definition it holds that \(-\Gamma =\Gamma ^H\), \(W =W^H\) and
Furthermore, \(W\ge 0\) follows from (15) and hence (pH) is satisfied. This proves (12).
Step 3: We continue with the proof of (13). Let \(Q\in \mathbb {K}^{n\times n}\) be such that \({\mathcal {S}}(x):=\tfrac{1}{2}x^HQ^HEx\) defines a storage function. To show (KYP) we verify the left inequality in (4) first. Since E is invertible the solutions of (1) are given by the solutions of the standard system \(\dot{x}(t)=E^{-1}Ax(t)+E^{-1}B u(t)\), \(x(0)=x_0\). In particular, for every choice of \((x_0,u(0))\in \mathbb {K}^{n}\times \mathbb {K}^m\) there exists a solution which fulfills (16) as a consequence of (Pa). Using now \(t=0\) in (16) this implies the left inequality in (KYP). Furthermore, by choosing \(T=0\) in (Pa) and the previous observation, that for every initial value \(x_0\in \mathbb {K}^n\) and \(u=0\) there exists a solution we conclude from the inequality on the right-hand side in (Pa) that \(Q^HE\ge 0\) holds. In summary, this proves that Q solves (KYP) and hence the implication (13). \(\square \)
However, we show with the following example that (KYP) does not necessarily imply (pH).
Example 2
Consider the real system given by \((E,A,B,C,D)=(1,-1,1,0,0)\). Then this system is asymptotically stable, i.e. all eigenvalues of matrix A have negative real parts, hence using (6) and (7), we conclude that it is detectable and controllable but not observable. Furthermore, (KYP) which is given by
has only the trivial solution \(Q=0\). Indeed, for \(Q\ne 0\) the above matrix is indefinite. To construct (pH) notice that \((J-R)Q=A=-1\) which implies \(R=Q^{-1}\) and
This yields \(G=\frac{1}{2}=-P\), \(N=S=0\) and hence
which is indefinite. Therefore the system does not fulfill (pH).
Next, we present an example that shows that \(\mathrm {(PR)}\wedge \mathrm {(Pa)}\nRightarrow \mathrm {(KYP)}\) for descriptor systems.
Example 3
Consider the real system \((E,A,B,C,D)=(0,1,0,1,0)\). Then the system dynamics is given by \(x(t)= 0\). Hence, \(x_0=0\) is the only consistent initial value. Therefore for all \(u\in L^1([t_0,t_1],\mathbb {R})\), \(0\le t_0\le t_1\) we have
which implies passivity. On the other hand, the corresponding LMI of this system for some \(x\in \mathbb {R}\) is given by
However, this cannot be valid since the matrix is indefinite for all \(x\in \mathbb {R}\). Furthermore, the system is positive real with transfer function \({\mathcal {T}}(s)=C(sE-A)^{-1}B+D=0\). Moreover, the system is behaviorally controllable but not minimal since \({{\,\textrm{rk}\,}}[E,B]=0\ne 1\) and \(A\ker E\nsubseteq {{\,\textrm{ran}\,}}E\). The proof that the system does not fulfill (pH) is similar to the calculation in Example 2.
In 3 we use that Note that it was important in the Examples 2 and 3 that the feedthrough term D is zero. It remains an open question if a similar construction is possible with a nonsingular feedthrough term.
Below, we briefly review known results:
-
(i)
In Theorem 2 of [18], the authors show that (KYP) with the condition \(E^HQ\ge 0\) is equivalent to the so-called extended strict positive realness and \(D+D^H>0\) under the condition that the systems are assumed to have index one and be asymptotically stable.
-
(ii)
In [17], it was shown that (KYP) implies (PR) and the converse is also true for certain minimal realizations of the system.
-
(iii)
In [26] it was shown that a nonnegative Popov function leads to a Hermitian solution to (KYP). It is assumed that there are no eigenvalues on the imaginary axis and that the system has index one.
-
(iv)
The author of [33] studies a more general behavioral approach and shows the equivalence of passivity and positive real pairs associated with the behavior. Although this approach contains descriptor systems, they were not investigated explicitly.
-
(v)
In [31], the authors have shown that (pH) \(\Rightarrow \) (KYP) \(\Rightarrow \) (PR) and that (KYP) \(\Rightarrow \) (pH) holds for invertible solutions Q of the KYP inequality. However, it remains open whether non-invertible solutions to the KYP inequality also lead to a pH formulation or if (PR) also implies (KYP).
-
(vi)
In [14, Section II], it is proven that the implication (pH) \(\Rightarrow \) (Pa) also holds for descriptor systems whose coefficients depend on the time t and the state x and not necessarily quadratic Hamiltonians.
-
(vii)
In [34], the authors consider stable systems, i.e. for all consistent initial values \(x_0\in \mathbb {K}^n\) the solution x to \(E\dot{x}(t)=Ax(t)\), \(x(0)=x_0\) fulfills \(\sup _{t\ge 0}\Vert x(t)\Vert <\infty \). It was shown that there exists \(Q\in \mathbb {K}^{n\times n}\) based on the solution of a generalized Lyapunov equation such that \(A=(J-R)Q\) holds on a certain subspace.
-
(viii)
In [35], the authors generalize the relationship between (KYP) and (PR) for standard systems. Moreover, for descriptor systems, our definition of (KYP) and (PR) can be seen as the special case where Q is replaced with QE. They show equivalence in the strict inequality case between (KYP) and (PR) if \(\det (s E -A) \ne 0\) for all \({{\,\textrm{Re}\,}}s \ge 0\), E nonsingular, and Q positive definite.
In the remainder of this section, we will show that passivity implies positive realness, i.e. (Pa) implies (PR). For descriptor systems, the state-control pair \((x,u)\in \mathbb {K}^{n}\times \mathbb {K}^m\) will in general not attain all values in \(\mathbb {K}^{n}\times \mathbb {K}^m\) which is the reason why we cannot deduce (KYP) from (Pa) for descriptor systems. Also, positive realness together with certain controllability assumptions will only lead to solutions of the KYP inequality on a subspace. In the following, we will use a more compact way of writing the KYP inequality for all \((x,u)\in \mathbb {K}^n\times \mathbb {K}^m\) as
Further, for Hermitian \(A\in \mathbb {K}^{n\times n}\) and a subspace \({\mathcal {V}}\) of \(\mathbb {K}^n\) the relation \(A\ge _{{\mathcal {V}}}0\) means that \(v^HAv\ge 0\) for all \(v\in {\mathcal {V}}\). Then we consider the following restricted versions of (4):
- (KYP|\({\mathcal {V}}\)):
-
There exists \(Q\in \mathbb {K}^{n\times n}\) with \(Q^HE=E^HQ\) and
$$\begin{aligned} {{\,\textrm{Re}\,}}\begin{bmatrix} Qx\\ u \end{bmatrix}^H\begin{bmatrix} A&{}\quad B\\ {}-C&{}\quad -D \end{bmatrix}\begin{bmatrix} x\\ u \end{bmatrix}\le 0,\quad x^HE^HQx\ge 0,\quad \text {for all }\begin{bmatrix} x\\ u \end{bmatrix}\in {\mathcal {V}}.\qquad \end{aligned}$$(18) - (KYP\(_{E}\)|\({\mathcal {V}}\)):
-
There exists \(Y=Y^H\in \mathbb {K}^{n\times n}\) such that (18) holds for some \(Q=YE\).
There are several authors who consider KYP inequalities restricted to suitable subspaces:
-
(i)
In [25], the authors use Ax instead of \(\dot{x}\) in the (KYP) inequality (4) which leads to a \(3\times 3\) block matrix where each block corresponds to one of the entries in \((\dot{x},x,u)\). To show the existence of (KYP) inequality solutions on a certain subspace for a positive real transfer function they assume minimality. Furthermore, the passivity notion is different since they allow for arbitrary nonnegative storage functions which must not vanish on \(\ker E\).
-
(ii)
The KYP inequality (KYP\(_{E}\)|\({\mathcal {V}}\)) is considered in [28] for the dual system restricted to the right deflating subspaces of the regular pair (E, A). The authors of [28] show that behavioral controllability and observability imply the existence of solutions to this inequality. However, the passivity notion used in [28] is only equivalent to (Pa) for behaviorally controllable and observable systems.
-
(iii)
In [27, 29], the authors consider KYP inequalities restricted to the so-called system space \(\mathcal {V}_{\textrm{sys}}\), which is the smallest subspace where the system trajectories evolve. It is shown in that case that behavioral controllability and (PR) imply (KYP|\(\mathcal {V}_{\textrm{sys}}\)).
-
(iv)
The authors in [15] consider the relation between storage functions, the feasibility of linear quadratic optimal control problems on an infinite time horizon and the existence of a solution to KYP inequalities on the system space \(\mathcal {V}_{\textrm{sys}}\). However, the definition of storage function is slightly different from the notion used in (Pa) and allows negative function values. Furthermore, in the definition of the system space \(\mathcal {V}_{\textrm{sys}}\) a slightly different solution concept is used.
Summarizing the literature review on restricted KYP inequalities given in (i)-(iv), we focus in the following on the most general approach presented in [27, 29].
The system space \(\mathcal {V}_{\textrm{sys}}\) can be characterized in terms of a limit of subspaces. It was shown in [27, Proposition 3.3] that the following sequence terminates after finitely many steps
where the pre-image of [A, B] is used. The resulting terminal subspace is the system space \(\mathcal {V}_{\textrm{sys}}\) and it fulfills \(\mathcal {V}_{\textrm{sys}}={\mathcal {V}}_{k}\) for all \(k\ge k_0\) and some \(k_0\ge 0\).
The restricted KYP inequalities (KYP|\(\mathcal {V}_{\textrm{sys}}\)) and (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) involve the semi-definiteness condition \(E^HQ\ge _{\mathcal {V}_{\textrm{sys}}^{x}} 0\), where \(\mathcal {V}_{\textrm{sys}}^{x}\) denotes the projection of \(\mathcal {V}_{\textrm{sys}}\subset \mathbb {K}^{n+m}\) to onto the first n components.
Compared to this, [27, 29] use the slightly different condition \(E^HQ\ge _{{\mathcal {V}}_{\textrm{diff}}} 0\), where the positive definiteness should hold on the subspace
Here a solution of (1) means that x is differentiable almost everywhere and fulfills (1) almost everywhere.
In the appendix, we show that (KYP|\(\mathcal {V}_{\textrm{sys}}\)) and (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)), are equivalent to the restricted KYP inequalities used in [27, 29].
We have the following results from [27, Theorem 4.1, Proposition 4.4] and [29, Theorem 4.3] on the relation between (PR), (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) and (KYP|\(\mathcal {V}_{\textrm{sys}}\)).
Proposition 4
Let \(\Sigma = (E,A,B,C,D)\) be a descriptor system of the form (1) and transfer function \({\mathcal {T}}(s)=C(sE-A)^{-1}B+D\). Then the following holds:
-
(a)
If there exists a solution \(Y=Y^H\) to (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) then \({\mathcal {T}}\) is positive real.
-
(b)
If the system is behaviorally controllable and \({\mathcal {T}}\) is positive real, then there exists a solution \(Y=Y^H\) to (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)).
-
(c)
If Y satisfies (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) then \(Q:=EY\) satisfies (KYP|\(\mathcal {V}_{\textrm{sys}}\)).
-
(d)
If Q satisfies (KYP|\(\mathcal {V}_{\textrm{sys}}\)) then there exists a solution Y to (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) with \(E^HYE=E^HQ\).
In Proposition 4 (b), we only consider the condition of behavioral controllability. The question that arises is whether the dual property, behavioral observability, can be used as a condition for the existence of solutions to the KYP inequality. To answer this question, we consider the following example.
Example 5
Consider \(\dot{x}=x\), \(y=cx+u\) for some \(c\ne 0\). Then this system is observable but not controllable. Furthermore, \({\mathcal {T}}(s)=1\) for all \(s\in \mathbb {C}\) which is positive real. However, (KYP) which is given by
is only solvable for sufficiently small \(Q<0\). Hence, positive realness and observability is not enough to guarantee a solution to (KYP).
Example 5 shows that behavioral observability of \(\Sigma =(E,A,B,C,D)\) together with a positive real transfer function does not imply (KYP\(_{E}\)|\(\mathcal {V}_{\textrm{sys}}\)) or (KYP|\(\mathcal {V}_{\textrm{sys}}\)). However, the behavioral observability of \(\Sigma \) is equivalent to the behavioral controllability of the dual system \(\Sigma '=(E^H,A^H,C^H,B^H,D^H)\) which leads us to the following result.
Corollary 6
Let \(\Sigma =(E,A,B,C,D)\) be a positive real and behaviorally observable descriptor system. Then the dual system \(\Sigma '=(E^H,A^H,C^H,B^H,D^H)\) is positive real and behaviorally controllable and there exists a solution to the generalized KYP inequality
Proof
The behavioral controllability of the dual system holds by definition and (8). Since for all \(\lambda \in \mathbb {C}\) with \({{\,\textrm{Re}\,}}\lambda \ge 0\) it holds that
the dual system is positive real and behaviorally controllable. Hence the existence of solutions to (20) follows from Proposition 4 (b),(c). \(\square \)
From Proposition 4 and (16) we immediately obtain the following corollary.
Corollary 7
Let (E, A, B, C, D) be a descriptor system of the form (1). Then the following holds:
Moreover, if the system is behaviorally controllable then (PR) implies (Pa).
It was shown in [29, Theorem 4.3] that (PR) together with the conditions
implies (KYP|\(\mathcal {V}_{\textrm{sys}}\)) and hence (Pa). In the case of standard systems the controllability assumptions on a positive real system to fulfill (Pa) or equivalently (KYP), can be relaxed to the sign-controllability, i.e.
see [36] and [16, Section A.4]. In particular, the first condition in (21) implies sign-controllability. For descriptor systems [27] shows that behavioral sign-controllability which means
together with a full rank assumption on the Popov function
where the rank is computed over the quotient field \(\mathbb {K}(s)\), implies that there is a Hermitian solution \(Q^HE=E^HQ\) to the generalized KYP inequality which is not necessarily nonnegative. The nonnegativity and hence (Pa) can be concluded from [29, Theorem 4.3 (c)] is the system is in addition behaviorally detectable meaning that
holds.
In addition to the presented results, the following example from [37] demonstrates that invertible solutions on the system space do not lead to (pH).
Example 8
Let \(E=\begin{bmatrix} 1&{}0\\ 0&{}0 \end{bmatrix}\), \(A=\begin{bmatrix} -1&{}\quad 0\\ 0&{}\quad 1 \end{bmatrix}\), \(B=\begin{bmatrix}1\\ 0 \end{bmatrix}\), \(C=\begin{bmatrix}1&1\end{bmatrix}\), \(D=0\). Then this system is behaviorally observable, behaviorally controllable, positive real with transfer function \({\mathcal {T}}(s)=\tfrac{1}{s+1}\) and the system space is \(\mathcal {V}_{\textrm{sys}}=\{(x_1,0,u)^\top \mid x_1,u\in \mathbb {R}\}\). However, it is not minimal, since \(A\ker E\nsubseteq {{\,\textrm{ran}\,}}E\) and \({{\,\textrm{rk}\,}}[E,B]=1\). Furthermore, the (KYP) has no solution but (KYP|\(\mathcal {V}_{\textrm{sys}}\)) holds since the system is behaviorally controllable.
3 When does (KYP) imply (pH)?
In order to study if (KYP) implies (pH), we first need to consider KYP solutions for DAEs. Below we show that for DAEs with invertible E the observability of the system guarantees invertible solutions of (KYP). This result was already obtained in [38] for positive semi-definite solutions of standard systems, but the proof can easily be extended to descriptor systems.
Proposition 9
Let \(\Sigma =(E,A,B,C,D)\) be a descriptor system of the form (1) with E invertible and (E, A, C) behaviorally observable. If \(Q\in \mathbb {K}^{n\times n}\) satisfies
then Q is invertible.
Proof
First, one shows that \(\ker Q\) is an (A, E)-invariant subspace, i.e. \(A\ker Q\subseteq E\ker Q\). Let \(z=Av\) with \(v\in \ker Q\) then
Hence \(v\in \ker (-A^H Q- Q^HA)\) which implies with \(v\in \ker Q\) that \(v\in \ker Q^HA\). Therefore \(Q^Hz= Q^HAv=0\), i.e. \(z\in \ker Q^H\). This shows \(A\ker Q\subseteq \ker Q^H\). Moreover, since E is invertible one has
Hence \(\ker Q\) is an (A, E)-invariant subspace. Next, we show that \(\ker Q\subseteq \ker C\). Let \(z\in \ker Q\). Then (KYP) implies for all \(w\in \mathbb {R}^m\) and \(\alpha \in \mathbb {R}\) that
Assume that \(Cz\ne 0\). Then all \(\alpha >0\) sufficiently small would violate (22). This contradiction leads to \(z\in \ker C\) which implies \(\ker Q\subseteq \ker C\). Hence Proposition 7.2 in [39] implies \(\ker Q=\{0\}\). \(\square \)
In Proposition 9 we do not require the positive semi-definiteness of \(Q^HE\) in (KYP). Hence as a special case, we obtain the following result.
Corollary 10
Let \(\Sigma =(E,A,B,C,D)\) be a descriptor system of the form (1) with E invertible and (E, A, C) behaviorally observable. If \(Q\in \mathbb {K}^{n\times n}\) satisfies (KYP) then Q is invertible.
Example 2 shows that the detectability of the system will in general not guarantee that the solutions of (KYP) are invertible.
Next, we extend the following result for standard systems [23, p. 55] which indicates that (pH) for a passive system can be established without the observability assumption.
Proposition 11
Let \(\Sigma =(E,A,B,C,D)\) define a descriptor system of the form (1). Then (pH) holds if and only if (KYP) holds for some \(Q\in \mathbb {C}^{n\times n}\) with \(\ker Q\subseteq \ker C\cap \ker A\).
Proof
If (pH) holds, then \(A=(J-R)Q\) and \(C=(G+P)^HQ\) for some \(J,R,Q\in \mathbb {K}^{n\times n}\) and \(G,P\in \mathbb {K}^{n\times m}\). Hence \(\ker Q\subseteq \ker A\cap \ker C\) and by Proposition 1, Q satisfies (KYP). To prove sufficiency assume that (KYP) holds for some \(Q\in \mathbb {K}^{n\times n}\) with \(\ker Q\subseteq \ker C\cap \ker A\). Then \(Q^HE\ge 0\) follows and we can introduce \(\Theta \in \mathbb {K}^{(n+m)\times (n+m)}\) via
Note that \(\Theta \) is well defined since \(\ker Q\subseteq \ker A\cap \ker C\) holds. To show that the system (E, A, B, C, D) fulfills (pH) we need to verify (2). To this end, we consider the trivial decomposition
and therefore the matrices \(J,R\in \mathbb {K}^{n\times n}\), \(P,G\in \mathbb {K}^{n\times m}\), \(S,N\in \mathbb {K}^{m\times m}\) in (2) are given by
Since N is skew-Hermitian, so is \(-N\) and therefore \(\Gamma \) given by (2) is skew-Hermitian.
Furthermore, (KYP) as it is rewritten in (17) yields
which finally proves (pH). \(\square \)
Remark 12
If Q is positive definite then (24) is equivalent to \(\Theta +\Theta ^H\le 0\) which also appears in some references as a definition of pH (descriptor) systems. If Q is singular, then \(\Theta +\Theta ^H\le 0\) cannot be concluded from (24) in general. Instead, we can redefine \(\Theta \) in such a way that it fulfills (23). To this end, we use the space decomposition \(\mathbb {K}^n={{\,\textrm{ran}\,}}Q\oplus \ker Q^H\) and letting
where \(P_{{{\,\textrm{ran}\,}}Q}\) is the orthogonal projector onto \({{\,\textrm{ran}\,}}Q\). We can redefine \(\Theta _1\) as
without changing the product on the right-hand side of (23). Hence the matrix \({{\hat{\Theta }}}\) which is obtained after replacing the block \(\Theta _1\) in \(\Theta \) with \({{\hat{\Theta }}}_1\) fulfills (23) and \({{\hat{\Theta }}}+{\hat{\Theta ^H\le }} 0\).
Remark 13
If \(\Sigma =(E,A,B,C,D)\) is behaviorally observable then the matrix Q in (pH) fulfills \(\ker Q\subseteq \ker A\cap \ker C=\{0\}\) by [39, Proposition 7.2, Theorem 7.3] and hence Q must be invertible.
Remark 14
Another way to use solutions Q to the KYP inequalities to obtain a representation that is quite similar to a pH formulation and does not require further assumptions is given by a left-multiplication of the state equations with \(Q^H\)
However, if Q is not invertible then this multiplication might enlarge the solution set of the descriptor system. The treatment of pH systems with singular Q is described in detail in [10, Section 6.3], see also [4].
4 When does (Pa) imply (pH)?
In Sect. 3, we discussed when (KYP) implies (pH). Although all solutions Q to (KYP) fulfill \(\ker Q\subseteq \ker C\) (proof of Proposition 9), one cannot guarantee that \(\ker Q\subseteq \ker A\cap \ker C\) holds, as Example 2 shows. Hence, a pH formulation of a given passive system is not always possible without further assumptions.
If E is invertible and (Pa) holds then Proposition 1 yields (KYP) and therefore the results of Sect. 3 can be used to characterize when (pH) holds. Hence we will focus in this section on the case where E is not invertible.
As a first result, we show that for passive systems one might restrict to systems that are controllable and observable with index at most two. First, we recall the notion of index of descriptor systems (1). Since (E, A) is assumed to be regular, there exist invertible \(S,T\in \mathbb {C}^{n\times n}\) and \(r\in {\mathbb {N}}\), see, e.g. [21], such that
where J and N are in Jordan canonical form and N is nilpotent. The block-diagonal form (25) is typically referred to as Kronecker–Weierstraß form. Based on this form, the index of the system (1) is defined as the smallest natural number \(\nu \) such that \(N^{\nu -1}\ne 0\) and \(N^{\nu }=0\).
If a given system in state-space form is not controllable or observable then one might consider the KYP inequality restricted to observable or controllable subspaces. This was proposed for standard systems in [40, 41], see also [16, Theorem 3.39].
For descriptor systems we recall a Kalman-like decomposition from [39, Theorem 8.1], see also [42] and [21, p. 51].
Proposition 15
For \(E,A\in \mathbb {K}^{l\times n}\), \(B\in \mathbb {K}^{l\times m}\) and \(C\in \mathbb {K}^{m\times n}\) there exists invertible \(S\in \mathbb {K}^{l\times l}\), \(T\in \mathbb {K}^{n\times n}\) such that
Furthermore, the subsystems
are completely controllable and observable, respectively.
Numerically, it is beneficial to use only unitary or orthogonal transformations S and T to obtain the Kalman-like form given in Proposition 15 as discussed in [43], see also [4, Section 7].
In the following we use the Kalman-like decomposition from Proposition 15 to obtain a completely controllable and observable realization of passive descriptor systems which has index at most two. Note that in [44] it was shown that the index of matrix pairs associated to port-Hamiltonian systems is at most two.
Corollary 16
Let \(\Sigma =(E,A,B,C,D)\) define a passive descriptor system of the form (1) with Kalman-like decomposition (26). Then the transfer function \({\mathcal {T}}(s)=C(sE-A)^{-1}B+D\) is positive real, \((E_{22},A_{22},B_2,C_2,D)\) is completely controllable and observable realization of \({\mathcal {T}}\) and the index of \((E_{22},A_{22})\) is at most two.
Proof
The matrices S and T used in (26) are invertible and therefore the following holds
Hence, \((E_{22},A_{22},B_2,C_2,D)\) is a realization of the transfer function \({\mathcal {T}}\). The passivity of the system together with Corollary 7 implies that \({\mathcal {T}}\) is positive real. Hence \(C_2(sE_{22}-A_{22})^{-1}B_2+D\) is positive real. Furthermore, Proposition 15 and the upper block-triangular structure implies that \((E_{22},A_{22},B_2,C_2,D)\) is completely controllable and observable. As a consequence, if \(n_2\) denotes the number of rows and columns of \(E_{22}\), then
It remains to show that the index of \((E_{22},A_{22})\) is at most two. By modifying the matrices S and T in (26) we can assume without restriction, that \((E_{22},A_{22})\) is already given in block diagonal form (25). We can further assume that \(E_{22}\) is a Jordan block at 0 of size \(n_2\) and that \(A_{22}\) is the identity. Using the nilpotency of \(E_{22}\) it follows that \((sE_{22}-A_{22})^{-1}=-\sum _{i=0}^{n_2-1}(sE_{22})^i\) holds. The rank condition (27) implies \(C_2E_{22}^{n_2-1}B_2\ne 0\). It is shown in [20, Section 5.1], see also Lemma 19, that the positive real function can be written as \({\mathcal {T}}(s)=C_2(sE_{22}-A_{22})^{-1}B_2+D=sM_1+{\mathcal {T}}_p(s)\) where \(\lim \limits _{s\rightarrow \infty }{\mathcal {T}}_p(s)\) exists. Therefore \(n_2\le 2\) which implies that the index of \((E_{22},A_{22})\) is at most two. \(\square \)
4.1 From (KYP|\(\mathcal {V}_{\textrm{sys}}\)) to (KYP)
It was shown in Corollary 7 that passive systems have a solution \(Q\in \mathbb {K}^{n\times n}\) to (KYP|\(\mathcal {V}_{\textrm{sys}}\)). In this section, we will rewrite this restricted KYP inequality and derive an equivalent KYP inequality that holds on the whole space. The idea is to choose a basis of \(\mathcal {V}_{\textrm{sys}}\) and redefine the system matrices in such a way that the system space of the new system will be the whole space.
Let \(M_{\mathcal {V}_{\textrm{sys}}}\in \mathbb {K}^{n\times \dim \mathcal {V}_{\textrm{sys}}}\) and \(M_{\hat{\mathcal {V}}_{\textrm{sys}}}\in \mathbb {K}^{n\times \dim \mathcal {V}_{\textrm{sys}}}\), \(\hat{\mathcal {V}}_{\textrm{sys}}=\left[ {\begin{matrix} A&{}B\\ {}-C&{}-D \end{matrix}}\right] \mathcal {V}_{\textrm{sys}}\) be matrices whose columns are a basis of \(\mathcal {V}_{\textrm{sys}}\) and a span of \(\hat{\mathcal {V}}_{\textrm{sys}}\), respectively. Using the Moore-Penrose pseudo-inverse \(M^\dagger \) of \(M\in \mathbb {K}^{k\times l}\), see, e.g. [45, Section 5.5.2], we introduce the system
If we consider the KYP inequality together with the projector formulas based on the pseudo-inverse
then for all \((x,u) \in \mathcal {V}_{\textrm{sys}}\) with \(w \in \mathbb {K}^{\dim \mathcal {V}_{\textrm{sys}}}\) such that \((x,u)= M_{\mathcal {V}_{\textrm{sys}}} w\) we obtain
and using \([A,B]\mathcal {V}_{\textrm{sys}}\subseteq [E,0]\mathcal {V}_{\textrm{sys}}\) yields
In a more simple way, this can be rewritten with
as
This is in fact (KYP) of a standard system which is obtained by using \({{\,\textrm{ran}\,}}E\) as state space and treating the remaining variables in \(\mathbb {K}^{\dim \mathcal {V}_{\textrm{sys}}}\) as input variables. However, the matrix \({{\hat{Q}}}\) which replaces \(\left[ {\begin{matrix} Q&{}0\\ 0&{}I_m \end{matrix}}\right] \) in (KYP) of the standard system must in general not admit this diagonal structure. This seems reasonable because the controls and states are not decoupled for general descriptor systems. For pH descriptor systems the interplay between state and control variables can be seen from a staircase form which was derived in [46] and can be achieved using only unitary transformations.
Analogously to Proposition 11, a pH formulation of the system \(\Sigma _{sys}\) is given if
holds. With this assumption we can define a matrix \({{\hat{\Theta }}}\in \mathbb {K}^{\dim \mathcal {V}_{\textrm{sys}}\times \dim \mathcal {V}_{\textrm{sys}}}\) as in Proposition 11 by setting
Motivated by the modified KYP inequality (28), we enlarge the class of solutions of (KYP) and show how in this case a pH formulation can be obtained. In the following we will restrict to standard systems \(\Sigma =(I_n,A,B,C,D)\) and study solutions \({{\hat{Q}}}\in \mathbb {K}^{(n+m)\times (n+m)}\)
Although every ordinary pH system fulfills (29), the converse is not necessarily true as the following example shows.
Example 17
We consider
This system does not fulfill (pH) since it is unstable. However the matrix \(\left[ {\begin{matrix} A&{}\quad B\\ {}-C&{}\quad -D \end{matrix}}\right] =\left[ {\begin{matrix} 1&{}\quad 3\\ {}-3&{}\quad -5 \end{matrix}}\right] \) used in (29) has the double eigenvalue \(\lambda _{1,2}=-2\). Hence the Lyapunov inequality (29) has a positive definite solution \({{\hat{Q}}}\in \mathbb {R}^{2\times 2}\).
In the following, we show different ways of obtaining (pH) if we additionally assume that \({{\hat{Q}}}\) in (29) is positive definite. The first way is to define
Since \({{\hat{Q}}}\) is not block diagonal, the new state and input variables might be given as a combination of the old state and input variables. Another way to obtain a pH formulation where either the new state variable or the new input variable can be chosen as a rescaling of the old variables is based on the following block Cholesky factorization, see, e.g. [45, Section 4.2.9],
This can be used to rewrite (29) as follows
Hence we could rewrite the system using the new variables
We obtain a pH system with Hamiltonian \({{\hat{Q}}}=I_n\), state \({{\hat{x}}}\) and input \({{\hat{u}}}\). If we would choose \({\mathcal {C}}\) to be lower triangular in the Cholesky factorization (30) then the new state \({{\hat{x}}}\) would be a rescaling of the old state x whereas \({{\hat{u}}}\) would be a linear combination of the state and input variables x and u.
Using the particular structure of \({\mathcal {C}}\), we further obtain
Furthermore, the matrix \({{\hat{A}}}\) satisfies \({{\hat{A}}}+{{\hat{A}}}^H\le 0\) and hence it is stable, i.e. it has only eigenvalues with a nonpositive real part and semi-simple eigenvalues on the imaginary axis (if there are any).
4.2 Passive systems with index at most one
As we have seen in Sect. 4, passive systems have always a realization that has index of at most two and that the index two blocks can be realized separately as pH systems. Hence the question remains whether passive systems with index at most one can be realized as (pH).
In the following, we construct a pH realization for descriptor systems \(\Sigma =(E,A,B,C,D)\) which are behaviorally observable and have index at most one.
If \(\Sigma \) has index at most one then, similar to the Kronecker–Weierstraß form (25), there exist invertible \(T_l,T_r\in \mathbb {C}^{n\times n}\) such that
with \(n = n_1 + n_2\), \(E_1\in \mathbb {K}^{n_1\times n_1}\) and \(A_2\in \mathbb {K}^{n_2\times n_2}\) invertible. Based on this transformation, we consider the following KYP inequality
Proposition 18
Let \(\Sigma =(E,A,B,C,D)\) be a descriptor system with index at most one which satisfies (32). Then the modified KYP inequality (33) has a solution \(Q_1\) if and only if \(\Sigma \) is passive. Moreover, if \(\Sigma \) is behaviorally observable then every solution \(Q_1\) of (33) is invertible.
Proof
We apply Corollary 7 where we showed that passivity is equivalent to (KYP|\(\mathcal {V}_{\textrm{sys}}\)). The system space of the block diagonal system (32) is given by
and therefore
Hence the KYP inequalities restricted to the system space are given by
which is equivalent to
Since \(E^HQ=Q^HE\), the following matrix is Hermitian
The invertibility of \(E_1\) implies \( Q_2=0\), i.e. \(T_l^{-H}QT_r\) is block lower-triangular. Hence the KYP inequalities on the system space are equivalent to
If the system \(\Sigma \) is behaviorally observable, then \((E_1,A_1,C_1)\) is observable and hence, using Proposition 9, we find that \(Q_1\) is invertible. \(\square \)
Proposition 18 can be used to obtain a pH representation of \((E_1,A_1,B_1,C_1,D-C_2A_2^{-1}B_2)\) as follows. If there exists a solution \(Q_1\) to (33) with \(\ker Q_1\subseteq \ker A_1\cap \ker C_1\) then a pH representation can be obtained by considering \(\Theta \) given by
and if \(Q_1\) is invertible then this simplifies to
5 When does (PR) imply (pH)?
In this section, we study whether a pH representation can be obtained from a positive real transfer function or not. This is also of particular interest when one wants to obtain a pH representation from frequency measurements of the transfer function. It was shown in [47] that if the interpolation points are chosen to be spectral zeros then one ends up with a pH realization. However, these spectral zeroes cannot be known in advance if only transfer function measurements are available. One solution proposed in [47] is to construct an intermediate realization from which the spectral zeroes can be computed. The question that arises then is what are the conditions on this intermediate system to end up in pH representation even for an index two pH descriptor system.
Given a transfer function \({\mathcal {T}}\) of a descriptor system (1) then \({\mathcal {T}}\) has a pole of finite order at \(\infty \) and if \(\nu \) is the index of the pair (E, A) then this order is at most \(\nu -1\). Hence, using the Laurent expansion of the entries of \({\mathcal {T}}\) there exists a sequence of matrices \((M_i)_{i=k}^{-\infty }\) with \(M_i\in \mathbb {C}^{n\times n}\) such that
If \({\mathcal {T}}(s)\) is assumed to be a real rational function, then \(M_i\in \mathbb {R}^{n\times n}\). In the following lemma, we show that for positive real rational functions, this representation can be simplified, see also [20, Section 5.1].
Lemma 19
Let \(\Sigma =(E,A,B,C,D)\) be a descriptor system with positive real transfer function \({\mathcal {T}}\) with Laurent expansion (35) then
holds for some rational function \({\mathcal {T}}_p(s)\) which fulfills \(M_0=\lim \nolimits _{s\rightarrow \infty }{\mathcal {T}}_p(s)\). Furthermore, it holds that \(M_0+M_0^H\ge 0\) and \(M_{1}=M_1^H\ge 0\). Moreover, if the system matrices E, A, B, C, D are real, then \(M_0\) and \(M_1\) are real and \({\mathcal {T}}_p(s)\) is positive real.
Proof
Since \({\mathcal {T}}\) is a rational function, it has no poles for all \(s=i\omega \) with \(|\omega |\) sufficiently large. Furthermore, the positive realness and analyticity imply that \({\mathcal {T}}(i\omega )+{\mathcal {T}}(i\omega )^H\ge 0\) holds for these values.
Using the positive realness of \({\mathcal {T}}\) and (35) for some \(k\ge 1\), we conclude for \(s=e^{i\varphi }r\) for all \(r>0\) and \(\varphi >0\) satisfying \(\tfrac{\varphi }{k}\le \tfrac{\pi }{2}\) that the following holds
We consider first the case \(k\ge 2\). Considering (37) for \(r>0\) sufficiently large and \(\varphi =\tfrac{\pi }{k}\) implies
Furthermore, choosing \(\varphi =0\) and \(r>0\) sufficiently large in (37) yields \(M_{k}+M_{k}^H\ge 0\). Therefore \(M_{k}=-M_{k}^H\) holds. If we consider (37) with \(\varphi =\tfrac{\pi }{2k}\) and \(r>0\) sufficiently large leads to \(2iM_{k}\le 0\) and choosing \(\varphi =-\tfrac{\pi }{2k}\) yields \(-2iM_{k}\le 0\). Hence we conclude \(M_k=0\) and by repeating this argument (if necessary) we obtain \(M_k=\ldots =M_2=0\).
Therefore (35) holds with \(k=1\). Then it remains to prove that \(M_1=M_1^H\ge 0\) is satisfied. We decompose \(M_1=M_1^++M_1^-\) where \(M_1^{\pm }=\frac{1}{2}(M_1\pm M_1^H)\). Then \(M_1^+\) is Hermitian and \(M_1^-\) is skew-Hermitian and we have \(sM_1+(sM_1)^H=s(M_1^++M_1^-)+{\overline{s}}(M_1^++M_1^-)^H=2({{\,\textrm{Re}\,}}s)M_1^++2({\textrm{Im}}\,s)iM_1^-\). Hence, if we consider \(s=i\omega \) and let \(\omega \rightarrow \infty \) then this contradicts \({\mathcal {T}}(s)+{\mathcal {T}}(s)^H\ge 0\). As a consequence, \(M_1^-=0\). Hence \(M_1=M_1^H\). If \(M_1\) would have a negative eigenvalue with eigenvector \(x\in \mathbb {C}^n\) we obtain a contradiction by considering \(x^H({\mathcal {T}}(s)+{\mathcal {T}}(s)^H)x\) for \(s\rightarrow \infty \). This shows that \(M_1=M_1^\top \ge 0\). Taking the limit \(\omega \rightarrow \infty \) in the positive realness condition
we further deduce \(M_0+M_0^\top \ge 0\).
If the system matrices are real then clearly \(M_0\) and \(M_1\) are real. Hence, it remains to conclude that \({\mathcal {T}}_p(s)\) is positive real. Since \({\mathcal {T}}(s)\) is real and positive real and \({\mathcal {T}}_p(i\omega )+{\mathcal {T}}_p(-i\omega )^\top ={\mathcal {T}}(i\omega )+{\mathcal {T}}(-i\omega )^\top \ge 0\) holds for all \(\omega \) such that \(i\omega \) is not a pole of \({\mathcal {T}}\), it follows from [20, Theorem 2.7.2] that \({\mathcal {T}}_p(s)\) is positive real. \(\square \)
Note that rational functions \({\mathcal {T}}\) for which \(\lim _{s\rightarrow \infty }{\mathcal {T}}(s)\) exists, are called proper. Hence we will refer to \({\mathcal {T}}_p\) in (36) as the proper part of a positive real transfer function.
The following example shows that positive realness of arbitrary real rational functions with polynomial growth cannot be concluded from considering the behavior on the imaginary axis alone.
Example 20
The function \({\mathcal {T}}(s)=s^3\) is analytic and hence it has no poles on the imaginary axis. Furthermore, it fulfills \({\mathcal {T}}(i\omega )+{\mathcal {T}}(-i\omega )^\top =0\). However, using the same arguments as in the proof of Lemma 19 we find that it is not positive real. Hence Theorem 2.7.2 in [20] cannot be extended to real rational functions that have a polynomial growth of order larger than two.
Furthermore, a nonnegative real part of the transfer function on the imaginary axis does not guarantee positive realness.
Example 21
Consider \({\mathcal {T}}(s)=-s^{-1}\), which satisfies \({\mathcal {T}}(s)+{\mathcal {T}}(s)^H=\tfrac{-2 {{\,\textrm{Re}\,}}s}{|s|^2}\le 0\) for all \(s\in \mathbb {C}\) with \({{\,\textrm{Re}\,}}s\ge 0\) and therefore does not fulfill (PR). Furthermore, a state-space realization is given by \((E,A,B,C,D)=(1,0,B,-B^{-1},0)\) for all scalar \(B\ne 0\). Hence the only solution to (KYP)
is given by \(Q=-\frac{1}{B^2}\). Observe that \({\mathcal {T}}\) satisfies \({\mathcal {T}}(i\omega )+{\mathcal {T}}(i\omega )^H\ge 0\) for all \(\omega \ne 0\), but \({\mathcal {T}}\) has a negative residue it the simple pole at \(\omega _0=0\), which is the reason why it is not positive real. Another consequence of this example is the existence of Hermitian solutions to (KYP) does not imply that the transfer function of the system is positive real.
Next, we study the relation of (PR) to (KYP) by recalling the following result presented in [17].
Proposition 22
If (E, A, B, C, D) is a real-valued minimal realization of a positive real transfer function \({\mathcal {T}}(s)\) with \(D+D^\top \ge M_0+M_0^\top \) then (KYP) holds.
First, observe that one can choose a minimal realization with \(D=M_0\). Hence for every positive real transfer function, there exists a realization that has a solution to (KYP). However, as we have seen in Sect. 4 this solution has to fulfill additional requirements if we want to define a pH realization. Furthermore, note that for minimal realizations of positive real transfer functions with index two it was shown in [17, Theorem 4.1] that (KYP\(_{E}\)|\(\mathbb {K}^n\times \mathbb {K}^m\)) is never fulfilled. In particular, we cannot obtain (pH) from the solution of this KYP inequality as the following example shows.
Example 23
Consider \(E=\begin{bmatrix} 1&{}0\\ 0&{}0 \end{bmatrix}\), \(A=\begin{bmatrix} 0&{}\quad -1\\ 1&{}\quad 0 \end{bmatrix}\), \(B=\begin{bmatrix} 0\\ 1 \end{bmatrix}=C^\top \), \(D=0\). Then this is a minimal realization of the positive real transfer function \({\mathcal {T}}(s)=s\). Here the (KYP) has the solution \(Q=I_2\). However, the inequalities in (KYP\(_{E}\)|\(\mathbb {K}^n\times \mathbb {K}^m\)) have no solution.
As a second example consider the descriptor system given by \((E,A,B,C,D)=(0,-1,1,-1,-1)\). Then \({\mathcal {T}}(s)=0\) which is positive real. However, (KYP) has no solution since D is negative. If we would replace D with \(M_1=0\) then (KYP) becomes solvable.
As an alternative, we can define a minimal pH realization directly from the transfer function (36). To this end, we consider a minimal realization of the proper part \({\mathcal {T}}_p(s)\) which is given by
where \(E_p\) is invertible which follows from the fact that minimal realizations of descriptor systems have no index one blocks in the Kronecker–Weierstraß form (25), see, e.g. [17, Theorem 6.3]. Then the minimality conditions (8) and (9) trivially hold which means that \((E_p,A_p,B_p)\) is behaviorally controllable and that \((E_p,A_p,C_p)\) is behaviorally observable. Hence, we know from Proposition 9 that there exists invertible \(Q_p\) such that \(Q_p^HE_p\ge 0\) and
Furthermore, a minimal pH realization of \(sM_{1}\) is given by
Indeed,
and the minimality conditions (8) and (9) can be verified easily. Furthermore, the system (38) is pH with \(Q_{\infty }=I_{2m}\).
In the following lemma, we show that we can combine the two pH systems from the proper and the non-proper part to obtain a minimal pH system realization of a positive real transfer function.
Lemma 24
Let \(\Sigma _i=(E_i,A_i,B_i,C_i,D_i)\), \(i=1,2\), be descriptor systems which fulfill (pH) for some \(Q=Q_1\) and \(Q=Q_2\) respectively. Then the system \(\Sigma _+\) given by
has the transfer function \({\mathcal {T}}_+={\mathcal {T}}_1+{\mathcal {T}}_2\) and fulfills (pH) with \(Q={{\,\textrm{diag}\,}}(Q_1,Q_2)\).
Proof
If systems \(\Sigma _i\), \(i=1,2\), fulfill (pH) then there exists \(J_i,R_i,Q_i\in \mathbb {K}^{n_i\times n_i}\), \(G_i,P_i\in \mathbb {K}^{n_i\times m_i}\), and \(S_i,N_i\in \mathbb {K}^{m_i\times m_i}\) such that
By setting \(J_+ = {{\,\textrm{diag}\,}}(J_1,J_2)\), \(R_+ ={{\,\textrm{diag}\,}}(R_1,R_2)\), \(Q_+={{\,\textrm{diag}\,}}(Q_1,Q_2)\), \(G_+={{\,\textrm{diag}\,}}(G_1,G_2)\), \(P_+={{\,\textrm{diag}\,}}(P_1,P_2)\), \(S_+={{\,\textrm{diag}\,}}(S_1,S_2)\) and \(N_+={{\,\textrm{diag}\,}}(N_1,N_2)\) it holds that
Hence \(\Sigma _+\) fulfills (pH). \(\square \)
In summary, this means that for behaviorally controllable and observable descriptor systems with positive real transfer functions, we can obtain a pH realization via
This is then already a pH descriptor system in the so-called staircase form which was studied recently in [46].
The connections of (pH), (KYP), (Pa) and (PR) for descriptor systems are summarized in Fig. 3.
6 Conclusion
In this paper, we studied conditions for the equivalence between passive, positive real and port-Hamiltonian descriptor systems, as well as their relation to the solvability of generalized KYP inequalities. The conditions on the equivalence already available in the literature were either validated or relaxed and counterexamples were also presented in the cases where the equivalence does not hold. In addition, we considered special cases: index one descriptor systems and KYP inequalities on a subspace which are shown to be equivalent to the finiteness of the available storage. Finally, we focused on conditions to obtain a port-Hamiltonian system from either a passive system, a positive real transfer function or the solution of the KYP inequality.
As future work, the analysis conducted in this note can be extended in several directions. Namely, one can study more general passivity properties such as cyclo passivity [48] which allows for possibly negative storage functions. In addition, we only focused on continuous-time systems. A similar study could be conducted for discrete-time systems. This could also help in finding a definition analogous to (pH) for this class of systems. Finally, pH systems are dissipative in the sense of [49] with respect to a specific supply rate \(w(x,u)={{\,\textrm{Re}\,}}u^H(Cx+Du)\). The analysis conducted in this note could also be extended to systems that are dissipative with respect to other quadratic supply rates such as the scattering supply rate which is given by \(w(x,u)=\Vert u\Vert ^2-\Vert Cx+Du\Vert ^2\) and closely related to the transfer functions being bounded real.
References
Beattie C, Mehrmann V, Xu H, Zwart H (2018) Linear port-Hamiltonian descriptor systems. Math Control Signals Syst 30. https://doi.org/10.1007/s00498-018-0223-3
Beattie C, Mehrmann V, Dooren PV (2019) Robust port-Hamiltonian representations of passive systems. Automatica 100:182–186
Jacob B, Zwart H (2012) Linear Port-Hamiltonian systems on Infinite-dimensional Spaces. Operator Theory: Advances and Applications, 223. Birkhäuser/Springer Basel AG, Basel
Mehrmann V, Unger B (2023) Control of port-Hamiltonian differential-algebraic systems and applications. Acta Numer 32:395–515. https://doi.org/10.1017/S0962492922000083
Ortega R, van der Schaft AJ, Mareels Y, Maschke BM (2001) Putting energy back in control. Control Syst Mag 21:18–33
van der Schaft AJ (2004) Port-Hamiltonian systems: network modeling and control of nonlinear physical systems. In: Advanced dynamics and control of structures and machines. CISM Courses and Lectures, vol 444. Springer, New York
van der Schaft AJ, Jeltsema D (2014) Port-Hamiltonian systems theory: an introductory overview. Found Trends Syst Control 1(2–3):173–378
Mehl C, Mehrmann V, Sharma P (2016) Stability radii for linear Hamiltonian systems with dissipation under structure-preserving perturbations. SIAM J Matrix Anal Appl 37(4):1625–1654
Mehrmann V, Van Dooren P (2020) Optimal robustness of port-Hamiltonian systems. SIAM J Matrix Anal Appl 41(1):134–151. https://doi.org/10.1137/19M1259092
Mehl C, Mehrmann V, Wojtylak M (2021) Distance problems for dissipative Hamiltonian systems and related matrix polynomials. Linear Algebra Appl 623:335–366. https://doi.org/10.1016/j.laa.2020.05.026
Mehrmann V, van der Schaft AJ (2022) Differential-algebraic systems with dissipative Hamiltonian structure. arXiv:2208.02737
van der Schaft AJ, Maschke B (2018) Generalized port-Hamiltonian DAE systems. Syst Control Lett 121:31–37
Gernandt H, Haller F, Reis T (2021) A linear relation approach to port-Hamiltonian differential-algebraic equations. SIAM J Matrix Anal Appl 42:1011–1044
Morandin R, Mehrmann V (2019) Structure-preserving discretization for port-Hamiltonian descriptor systems. In: IEEE 58th conference on decision and control (CDC), 6863–6868
Reis T, Voigt M (2019) Linear-quadratic optimal control of differential-algebraic systems: the infinite time horizon problem with zero terminal state. SIAM J Control Optim 57:1567–1596
Brogliato B, Lozano R, Maschke B, Egeland O (2007) Dissipative systems analysis and control. Springer, Cham
Freund R, Jarre F (2004) An extension of the positive real lemma to descriptor systems. Optim Methods Softw 19(1):69–87. https://doi.org/10.1080/10556780410001654232
Zhang L, Lam J, Xu S (2002) On positive realness of descriptor systems. IEEE Trans Circuits Syst-I: Fundam Theory Appl 49(3):401–407
Anderson BDO (1967) A system theory criterion for positive real matrices. J Control 5(2):171–182
Anderson BDO, Vongpanitlerd S (1973) Network analysis and synthesis. Prentice-Hall Inc, Englewood Cliffs
Dai L (1989) Singular Control Systems, vol 118. Lecture notes in control and information sciences. Springer, Berlin
Verghese G, Lévy B, Kailath T (1981) A generalized state-space for singular systems. IEEE Trans Autom Control 26(4):811–831
van der Schaft AJ (2009) Port-Hamiltonian systems. In: Modeling and control of complex physical systems: the port-Hamiltonian approach. Springer, Netherlands. https://doi.org/10.1007/978-3-642-03196-0
Hughes TH, Branford EH (2022) Dissipativity, reciprocity, and passive network synthesis: from the seminal dissipative dynamical systems articles of Jan Willems to the present day. IEEE Control Syst Mag 42(3):36–57. https://doi.org/10.1109/MCS.2022.3157135
Camlibel MK, Frasca R (2009) Extension of Kalman–Yakubovich–Popov lemma to descriptor systems. Syst Control Lett 58(12):795–803. https://doi.org/10.1016/j.sysconle.2009.08.010
Masubuchi I (2006) Dissipativity inequalities for continuous-time descriptor systems with applications to synthesis of control gains. Syst Control Lett 55(2):158–164. https://doi.org/10.1016/j.sysconle.2005.06.007
Reis T, Rendel O, Voigt M (2015) The Kalman–Yakubovich–Popov inequality for differential-algebraic systems. Linear Algebra Appl 485:153–193. https://doi.org/10.1016/j.laa.2015.06.021
Reis T, Stykel T (2010) Positive real and bounded real balancing for model reduction of descriptor systems. Int J Control 83(1):74–88
Reis T, Voigt M (2015) The Kalman–Yakubovich–Popov inequality for differential-algebraic systems: existence of nonpositive solutions. Syst Control Lett 86:1–8. https://doi.org/10.1016/j.sysconle.2015.09.003
Cherifi K, Mehrmann V, Hariche K (2019) Numerical methods to compute a minimal realization of a port-Hamiltonian system. arXiv:1903.07042
Gillis N, Sharma P (2018) Finding the nearest positive-real system. SIAM J Numer Anal 56(2):1022–1047
Willems JC (1971) Least squares stationary optimal control and the algebraic Riccati equation. IEEE Trans Autom Control AC-16(6):621–634
Hughes TH (2017) A theory of passive linear systems with no assumptions. Autom J IFAC 86:87–97
Gernandt H, Haller FE (2021) On the stability of port-Hamiltonian descriptor systems. IFAC-Papers OnLine 54(19):137–142. https://doi.org/10.1016/j.ifacol.2021.11.068
Iwasaki T, Hara S (2005) Generalized KYP lemma: unified frequency domain inequalities with design applications. IEEE Trans Autom Control 50:41–59
Ferrante A (2005) Positive real lemma: necessary and sufficient conditions for the existence of solutions under virtually no assumptions. IEEE Trans Autom Control 50(5):720–724. https://doi.org/10.1109/TAC.2005.847036
Breiten T, Schulze P (2021) Structure-preserving linear quadratic Gaussian balanced truncation for port-Hamiltonian descriptor systems. arXiv:2111.05065v1
Camlibel M, Iannelli L, Vasca F (2014) Passivity and complementarity. Math Program 145(1–2):531–563. https://doi.org/10.1007/s10107-013-0678-4
Berger T, Reis T, Trenn S (2017) Observability of linear differential-algebraic systems: a survey. In: Ilchmann A, Reis T (eds) Surveys in differential-algebraic equations IV. Differential-algebraic equations forum, pp 161–219. Springer, Berlin. https://doi.org/10.1007/978-3-319-46618-7_4
Scherer R, Wendler W (1994) A generalization of the positive real lemma. IEEE Trans Autom Control 39(4):882–886
Xiao C, Hill DJ (1999) Generalizations and new proof of the discrete-time positive real lemma and bounded real lemma. IEEE Trans Circuits Syst-I: Fundam Theory Appl 46(6):740–743
Banaszuk A, Kociȩcki M, Lewis FL (1992) Kalman decomposition for implicit linear systems. IEEE Trans Autom Control 37:1509–1514
Bunse-Gerstner A, Byers R, Mehrmann V, Nichols NK (1999) Feedback design for regularizing descriptor systems. Linear Algebra Appl 299:119–151
Mehl C, Mehrmann V, Wojtylak M (2018) Linear algebra properties of dissipative Hamiltonian descriptor systems. SIAM J Matrix Anal Appl 39:1489–1519
Golub GH, Van Loan CF (2013) Matrix computations, 4th edn. The Johns Hopkins University Press, Baltimore
Beattie C, Gugercin S, Mehrmann V (2019) Structure-preserving interpolatory model reduction for port-Hamiltonian differential-algebraic systems. arXiv:1910.05674
Benner P, Goyal P, Van Dooren P (2020) Identification of port-Hamiltonian systems from frequency response data. Syst Control Lett 143:104741. https://doi.org/10.1016/j.sysconle.2020.104741
van der Schaft AJ, Jeltsema D (2021) On energy conversion in port-Hamiltonian systems. In: 2021 60th IEEE conference on decision and control (CDC), pp 2421–2427. https://doi.org/10.1109/CDC45484.2021.9683292
Willems JC (1972) Dissipative dynamical systems—part 2: linear systems with quadratic supply rates. Soviet J Opt Technol (English translation of Optiko-Mekhanicheskaya Promyshlennost) 45(5):352–393
Ilchmann A, Reis T (2017) Outer transfer functions of differential-algebraic systems. ESAIM: COCV 23(2):391–425. https://doi.org/10.1051/cocv/2015051
Byers R, Geerts T, Mehrmann V (1997) Descriptor systems without controllability at infinity. SIAM J Control Optim 35(2):462–479. https://doi.org/10.1137/S0363012994269818
Acknowledgements
HG gratefully acknowledges the funding within the SPP1984 “Hybrid and multimodal energy systems” by the Deutsche Forschungsgemeinschaft (DFG). DH acknowledges funding from the DFG within SFB910 “Control of self-organizing nonlinear systems: Theoretical methods and concepts of application”. KC acknowledges funding from ProFIT (co-financed by the Europäischen Fonds für regionale Entwicklung (EFRE)) within the WvSC project: EA 2.0 - Elektrische Antriebstechnik. The authors would like to thank Volker Mehrmann for proposing this research topic and for his valuable suggestions on an earlier draft of this manuscript, as well as Arjan van der Schaft and Philipp Schulze for valuable discussions.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary results on the system space
Supplementary results on the system space
In this section, we provide basic properties of the system space \(\mathcal {V}_{\textrm{sys}}\) which mainly follow from [27].
First, a method for computing the system space from [27] is recalled. To this end, we consider for invertible \(S,T\in \mathbb {K}^{n\times n}\) and \(F\in \mathbb {K}^{m\times n}\) the feedback equivalent system
It is easy to see that the solution (x, u) of a system (1) and the solution \((x_F,u_F)\) of its feedback equivalent system (39) are connected via the following invertible linear transformation
If (E, A) is regular then it was shown, e.g. in [50, Proposition 2.12], see also [51, Theorem 3.2], that there exists \(F\in \mathbb {K}^{m\times n}\) and invertible \(S,T\in \mathbb {K}^{n\times n}\) such that \((E_F,A_F,B_F)\) given by (39) fulfills
where \(E_{33}\) is nilpotent. The matrices (41) lead to the differential algebraic equation
Since \(E_{33}\) is nilpotent, every solution of (42) fulfills \(x_3=0\). Therefore,
for some initial value \(x_{1}^0\in \mathbb {K}^{n_1}\).
It was shown in [27, Proposition 3.3] that the system (41) fulfills
Consequently, the system space of a general system (E, A, B) with corresponding feedback equivalent system \((E_F,A_F,B_F)\) given by (41) can be obtained by applying the invertible linear transformation (40).
Next, we consider the subspace of values attained by smooth trajectories
Here smooth solution means that the state and input functions x and u satisfy the equation (1) and that the derivatives \(x^{(k)}\) and \(u^{(k)}\) of arbitrary order \(k\in {\mathbb {N}}\) exist.
Using (43) we obtain smooth solution of (42) if and only if there exist \(x_{1}^0\in \mathbb {K}^{n_1}\) and \(u^0\in \mathbb {K}^{m}\) such that the following holds
Therefore, for the system (41) we obtain
Hence the system space can be interpreted as the smallest subspace in which the trajectories (x, u) of the descriptor system (1) evolve.
Finally, we argue why (KYP) is equivalent to the KYP inequalities used in [27, 29]. Recall the space of differential initial variables which is given by
If we denote by \(\mathcal {V}_{\textrm{sys}}^{x}\) the projection of \(\mathcal {V}_{\textrm{sys}}\) onto the first n components, then it follows from [29, Lemma 2.9 (b)] that the following relation holds
In [29] is was assumed that a solution \(Y=Y^H\) to (KYP\(_E\)|\(\mathcal {V}_{\textrm{sys}}\)) fulfills the condition
i.e. \(\mathcal {V}_{\textrm{sys}}^{x}\) is replaced by the larger set \({\mathcal {V}}_{\textrm{diff}}\). Therefore it is clear that the nonnegativity of solutions to the KYP inequalities used in [27, 29] are solutions to (KYP|\(\mathcal {V}_{\textrm{sys}}\)) and (KYP\(_E\)|\(\mathcal {V}_{\textrm{sys}}\)).
We continue by showing that the converse is valid. Let \(Y=Y^H\) solve (KYP\(_E\)|\(\mathcal {V}_{\textrm{sys}}\)) then \(Q:=YE=Y^HE\) fulfills \(Q^HE=E^HQ\) and for \(x=x_1+x_2\in {\mathcal {V}}_{\textrm{diff}}\) with \(x_1\in \mathcal {V}_{\textrm{sys}}^{x}\) and \(x_2\in \ker E\) it holds that
which implies (48).
Also, if (KYP|\(\mathcal {V}_{\textrm{sys}}\)) holds for some \(Q\in \mathbb {K}^{n\times n}\) satisfying \(Q^HE=E^HQ\) then Proposition 4.4 in [29] implies that there exists \(Y=Y^H\) such that \(Q^HE=E^HYE\) holds and therefore
and therefore (48) holds.
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
Cherifi, K., Gernandt, H. & Hinsen, D. The difference between port-Hamiltonian, passive and positive real descriptor systems. Math. Control Signals Syst. 36, 451–482 (2024). https://doi.org/10.1007/s00498-023-00373-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00498-023-00373-2
Keywords
- Port-Hamiltonian systems
- Differential-algebraic equations
- Minimal realizations
- Passive systems
- Positive real systems
- Kalman–Yakubovich–Popov inequality