The Dual Quaternion Matrix Equation AXB=C with Applications

Preprint

Article

The Dual Quaternion Matrix Equation AXB=C with Applications

Altmetrics

Downloads

110

Views

Comments

A peer-reviewed article of this preprint also exists.

Yan Chen,

Qing-Wen Wang^*,Lv-Ming Xie

Yan Chen,

Qing-Wen Wang^*,Lv-Ming Xie

This version is not peer-reviewed

Submitted:

05 February 2024

Posted:

06 February 2024

You are already at the latest version

Alerts

Abstract

Dual quaternions have wide applications in automatic differentiation, computer graphics, mechanics and others. Due to its application in control theory, the matrix equation AXB=C has been extensively studied. However, there is currently limited information on matrix equation AXB=C over the dual quaternion algebra. In this paper, we provide the necessary and sufficient conditions for the solvability of the dual quaternion matrix equation AXB=C, and present the expression for the general solution when it is solvable. As an application, we derive the ϕ-Hermitian solutions for the dual quaternion matrix equation AXAϕ=C, where the ϕ-Hermitian extends the concepts of Hermiticity and η-Hermiticity. Lastly, we present a numerical example to verify the main research results of this paper.

Keywords:

Subject: Physical Sciences - Mathematical Physics

MSC: 15A03; 15A09; 15A24; 15B33; 15B57

1. Introduction

Let

R

denote the set of real numbers,

H^{m \times n}

stand for the space of all

m \times n

matrices over the real quaternion algebra

H = {u_{0} + u_{1} i + u_{2} j + u_{3} k | i^{2} = j^{2} = k^{2} = i j k = - 1, u_{0}, u_{1}, u_{2}, u_{3} \in R} .

The symbol of

r (A)

A^{*}

, I, and 0 are defined by the rank of a given quaternion matrix A, the conjugate transpose of A, identity matrix, zero matrix with appropriate sizes, respectively. The Moore-Penrose inverse of

A \in H^{l \times k}

is denoted as

A^{†}

, which is defined as the solution of

A Y A = A, Y A Y = Y, {(A Y)}^{*} = A Y

and

{(Y A)}^{*} = Y A .

Moreover, let

L_{A} = I - A^{†} A

and

R_{A} = I - A A^{†}

represent two projectors along A.

Since Hamilton’s discovery of quaternions in 1843, quaternions and quaternion matrices have found a huge amount of practical applications in fields such as computer science, statistics, quantum physics, signal and color image processing, flight mechanics, aerospace technology, and so on (see, e.g., [1,2,3,4]).

Matrix equations have garnered significant attention in recent years and have become one of the main topics in matrix theory(see, e.g., [5,6,7,8,9]). For the classical matrix equation

\begin{matrix} A X B = C, \end{matrix}

(1)

which has been studied by many authors. Ben-Israel and Greville [10] eatablished the necessary and sufficient condition for the solvability of (1). In 2003, Liao and Bai [11] investigated the least-squares solution of (1) over symmetric positive semidefinite matrices. Huang et al. [12] gave the skew-symmetric solution and the optimal approximate solution of (1). Peng [13] derived centro-symmetric solution of matrix equation (1). Deng et al. [14] studied the general expressions about the Hermitian solutions of (1). Xie and Wang [15] considered the reducible solution to (1) when it is solvable. As a special case of the matrix equation (1), the Hermitian solution X to the matrix equation

\begin{matrix} A X A^{*} = B \end{matrix}

(2)

has attracted extensive attention(see, e.g., [16,17]). Baksalary[18] and Größ[19] studied the nonnegative definite and positive definite solutions to the matrix equation (2), respectively. For

η \in {i, j, k}

, a quaternion matrix A is called

η

-Hermitian and

η

-skew-Hermitian if

A = A^{η^{*}}

and

A = - A^{η^{*}}

, respectively, where

A^{η^{*}} = - η A^{*} η

[20]. The utilization of

η

-Hermitian matrices in linear modeling is extensively recognized[21]. Kyrchei[22] derived the explicit determinantal representation formulas of

η

-Hermitian and

η

-skew-Hermitian solutions to the quaternion matrix equation

A X A^{η *} = B .

It is well known that dual numbers and dual quaternions have wide applications in computer graphics, automatic differentiation, geometry, mechanics, rigid body motions and robotics(see, e.g., [23,24,25,26]). For the related definitions of dual numbers and dual quaternions, please see Section 2.

So far, there has been little information on matrix equation (1) over the dual quaternion algebra. Motivated by the work mentioned above, we in this paper aim to investigate the general solution of dual quaternion matrix equation (1) by using Moore–Penrose inverses and ranks of matrices. Since the

ϕ

-Hermitian serves as an extended form of both Hermiticity and

η

-Hermiticity over the quaternions [27], we also provide the definition of

ϕ

-Hermiticity over the dual quaternions. As an application, we establish the

ϕ

-Hermitian solution of a special dual quaternion matrix equation

\begin{matrix} A X A^{ϕ} = C, C^{ϕ} = C . \end{matrix}

(3)

Further details regarding

ϕ

-Hermitian matrices will be illustrated in Section 2.

This paper is organized as follows. In Section 2, we provide an overview of essential definitions and lemmas that will be applied in the subsequent sections. In Section 3, we establish some necessary and sufficient conditions for solvability to the dual quaternion matrix equation (1) and consider some special cases of dual quaternion matrix equation (1). As an application, we investigate the

ϕ

-Hermitian solution of the dual quaternion matrix equation (3) in Section 4. In Section 5, we present a numerical example to illustrate the results of this paper. Finally a brief conclusion is provided in Section 6.

2. Preliminaries

In this section, we review some definitions of dual numbers, dual quaternions and related propositions. Moreover, we introduce the definitions of dual quaternion matrix and

ϕ

-Hermitian matrix which are fundamental for obtaining the main results.

Definition 1

([28]). Suppose that

x_{0}, x_{1} \in R

, we say x is a dual number if x has the form

\begin{matrix} x = x_{0} + x_{1} ϵ, \end{matrix}

where ϵ is the infinitesimal unit, satisfying

ϵ^{2} = 0

We call

x_{0}

as the real part or the standard part of x,

x_{1}

as the dual part or the infinitesimal part of x. The infinitesimal unit

ϵ

is commutative in multiplication with real numbers, complex numbers, and quaternions. The set of dual numbers is denoted by

D = {x = x_{0} + x_{1} ϵ | ϵ^{2} = 0, x_{0}, x_{1} \in R} .

Assume that

x = x_{0} + x_{1} ϵ, y = y_{0} + y_{1} ϵ \in D

, we have

x = y

x_{0} = y_{0}

and

x_{1} = y_{1}

, regarding addition and multiplication, there is

\begin{matrix} x + y = x_{0} + y_{0} + (x_{1} + y_{1}) ϵ, \\ x y = x_{0} y_{0} + (x_{0} y_{1} + x_{1} y_{0}) ϵ . \end{matrix}

Definition 2

([28]). Let

z_{0}, z_{1} \in H

. We say z is a dual quaternion if z has the form

\begin{matrix} z = z_{0} + z_{1} ϵ, \end{matrix}

where ϵ is the infinitesimal unit, satisfying

ϵ^{2} = 0

z_{0}, z_{1}

as the real part and the dual part of z, respectively.

The collection of dual quaternions is denoted by

DQ = {z = z_{0} + z_{1} ϵ | ϵ^{2} = 0, z_{0}, z_{1} \in H} .

Now, we introduce the definition of dual quaternion matrix. Let

X_{0}, X_{1} \in H^{m \times n}

. X is said to be a dual quaternion matrix if X has the form

X = X_{0} + X_{1} ϵ

, the set of dual quaternion matrices is denoted by

{DQ}^{m \times n} = {X = X_{0} + X_{1} ϵ | ϵ^{2} = 0, X_{0}, X_{1} \in H^{m \times n}} .

The conjugate transpose of X is defined as

X^{*} = X_{0}^{*} + X_{1}^{*} ϵ

. For

Y = Y_{0} + Y_{1} ϵ \in {DQ}^{m \times n}

, by analogy, we have

X = Y

X_{0} = Y_{0}

and

X_{1} = Y_{1}

, furthermore

\begin{matrix} X + Y = X_{0} + Y_{0} + (X_{1} + Y_{1}) ϵ, \\ X Y = X_{0} Y_{0} + (X_{0} Y_{1} + X_{1} Y_{0}) ϵ . \end{matrix}

To facilitate our study on the

ϕ

-Hermitian, we will first review the concept of nonstandard involution over quaternions, and then proceed to generalize it to dual quaternions.

Definition 3

([27]). A map

ϕ : H \to H

is called an antiendomorphism, if

ϕ (p q) = ϕ (q) ϕ (p)

and

ϕ (p + q) = ϕ (p) + ϕ (q)

for all

p, q \in H

. An antiendomorphism ϕ is called an involution if

ϕ (ϕ (p)) = p

for every

p \in H

Definition 4

([27]). Under the basis

(1, i, j, k)

, an involution ϕ is called nonstandard if and only if ϕ can be expressed as a real matrix

ϕ = (\begin{matrix} 1 & 0 \\ 0 & T \end{matrix}),

where T is a

3 \times 3

real orthoganal symmetric matrix with eigenvalues

1, 1, - 1

Proposition 1

([27]). Let

z \in H

. Then every nonstandard involution ϕ of

H

has the form

ϕ (z) = β^{- 1} z^{*} β

for some

β \in H

with

β^{2} = - 1

Definition 5

([27]). For a nonstandard involution ϕ, a quaternion matrix Z is said to be ϕ-Hermitian if

Z = Z^{ϕ}

, where

Z^{ϕ}

is obtained by applying ϕ entrywise to the transposed matrix of Z. In fact,

Z^{ϕ} = β^{- 1} Z^{*} β

Proposition 2.

Let

Z \in H^{m \times n}, β \in H

and

β^{2} = - 1

. Then

(1)

{(Z^{ϕ})}^{†} = {(Z^{†})}^{ϕ}

(2)

{(L_{Z})}^{ϕ} = R_{Z^{ϕ}}

(3)

{(R_{Z})}^{ϕ} = L_{Z^{ϕ}}

Proof.

It is obvious that

β^{- 1} = - β, β^{*} = - β

, by the definition of Moore-Penrose inverse, we obtain

\begin{matrix} Z^{ϕ} {(Z^{†})}^{ϕ} Z^{ϕ} & = (- β Z^{*} β) [- β {(Z^{†})}^{*} β] (- β Z^{*} B) \\ = - β Z^{*} {(Z^{†})}^{*} Z^{*} β = - β {(Z Z^{†} Z)}^{*} β \\ = - β Z^{*} β = Z^{ϕ}, \\ {(Z^{†})}^{ϕ} Z^{ϕ} {(Z^{†})}^{ϕ} & = [- β {(Z^{†})}^{*} β] (- β Z^{*} β) [- β {(Z^{†})}^{*} β] \\ = - β {(Z^{†})}^{*} Z^{*} {(Z^{†})}^{*} β = - β {(Z^{†} Z Z^{†})}^{*} β \\ = - β {(Z^{†})}^{*} β = {(Z^{†})}^{ϕ}, \\ {}^{ϕ}{[Z (Z^{†})}^{ϕ}]^{*} & = {[(- β Z^{*} β) (- β {(Z^{†})}^{*} β)]}^{*} \\ = {[- β Z^{*} {(Z^{†})}^{*} β]}^{*} = {(- β Z^{†} Z β)}^{*} \\ = - β {(Z^{†} Z)}^{*} β = (- β Z^{*} β) [- β {(Z^{†})}^{*} β] \\ = Z^{ϕ} {(Z^{†})}^{ϕ}, \\ [(Z^{†} {)^{ϕ} Z^{ϕ}]}^{*}^{†} & = {[- β {(Z^{†})}^{*} β (- β Z^{*} β)]}^{*} \\ = {[- β {(Z^{†})}^{*} Z^{*} β]}^{*} = {(- β Z Z^{†} β)}^{*} \\ = - β {(Z Z^{†})}^{*} β = [- β {(Z^{†})}^{*} β] (- β Z^{*} β) \\ = {(Z^{†})}^{ϕ} Z^{ϕ} . \end{matrix}

(4)

Based on this, we can deduce that

{(Z^{†})}^{ϕ}

is the Moore-Penrose inverse of

Z^{ϕ}

For

(2)

, we have

{(L_{Z})}^{ϕ} = {(I - Z^{†} Z)}^{ϕ} = I - Z^{ϕ} {(Z^{†})}^{ϕ} = I - Z^{ϕ} {(Z^{ϕ})}^{†} = R_{Z^{ϕ}} .

In a similar vein to

(2)

, we can offer a demonstration for

(3)

, therefore, we omit it here. □

By analogy, we propose the definition of

ϕ

-Hermiticity with respect to dual quaternion matrix, where

ϕ

is a nonstandard involution.

Definition 6.

For

X = X_{0} + X_{1} ϵ \in {DQ}^{m \times n},

X is called ϕ-Hermitian matrix if

X = X^{ϕ}

, where

X^{ϕ} : = β^{- 1} X^{*} β = β^{- 1} X_{0}^{*} β + β^{- 1} X_{1}^{*} β ϵ = X_{0}^{ϕ} + X_{1}^{ϕ} ϵ,

with

β \in H

and

β^{2} = - 1

Proposition 3.

Let

X, Y \in {DQ}^{n \times n}

. Then

(1)

{(X + Y)}^{ϕ} = X^{ϕ} + Y^{ϕ}

(2)

{(X Y)}^{ϕ} = Y^{ϕ} X^{ϕ}

(3)

{(X^{ϕ})}^{ϕ} = X

Proof.

By the algebraic properties of

ϕ

, we have

\begin{matrix} {(X + Y)}^{ϕ} & = {[(X_{0} + Y_{0}) + (X_{1} + Y_{1}) ϵ]}^{ϕ} \\ = {(X_{0} + Y_{0})}^{ϕ} + {(X_{1} + Y_{1})}^{ϕ} ϵ \\ = X_{0}^{ϕ} + Y_{0}^{ϕ} + X_{1}^{ϕ} ϵ + Y_{1}^{ϕ} ϵ \\ = X_{0}^{ϕ} + X_{1}^{ϕ} ϵ + Y_{0}^{ϕ} + Y_{1}^{ϕ} ϵ \\ = X^{ϕ} + Y^{ϕ} . \end{matrix}

In relation to

(2)

, we obtain

\begin{matrix} {(X Y)}^{ϕ} & = {[X_{0} Y_{0} + (X_{0} Y_{1} + X_{1} Y_{0}) ϵ]}^{ϕ} \\ = {(X_{0} Y_{0})}^{ϕ} + {(X_{0} Y_{1} + X_{1} Y_{0})}^{ϕ} ϵ \\ = Y_{0}^{ϕ} X_{0}^{ϕ} + Y_{1}^{ϕ} X_{0}^{ϕ} ϵ + Y_{0}^{ϕ} X_{1}^{ϕ} ϵ \\ = (Y_{0}^{ϕ} + Y_{1}^{ϕ} ϵ) (X_{0}^{ϕ} + X_{1}^{ϕ} ϵ) \\ = Y^{ϕ} X^{ϕ} . \end{matrix}

In terms of

(3)

, we have

\begin{matrix} {(X^{ϕ})}^{ϕ} & = {[{(X_{0} + X_{1} ϵ)}^{ϕ}]}^{ϕ} \\ = {(X_{0}^{ϕ} + X_{1}^{ϕ} ϵ)}^{ϕ} \\ = X_{0} + X_{1} ϵ \\ = X . \end{matrix}

□

Now, we provide a few lemmas which are basic tools for getting the key outcomes.

Lemma 1

([29]). Suppose that A, B and C are provided for matrices with the adequate dimensions over

H

, then the quaternion matrix equation (1) is consistent if and only if

R_{A} C = 0, C L_{B} = 0 .

In this case, the general solution can be expressed as

\begin{matrix} X = A^{†} C B^{†} + L_{A} U + V R_{B}, \end{matrix}

where

U, V

are any matrices over

H

with appropriate dimensions.

Lemma 2

([15]). Let

A_{1}, A_{2}, B_{1}, B_{2}

and

C_{1}

be given with appropriate sizes. Set

A = R_{A_{1}} C, B = B_{1} L_{B_{2}}, M = R_{A_{1}} A_{2}, C_{1} = C L_{B_{2}} .

Then the following descriptions are equivalent:

(1)

The quaternion matrix equation

A_{1} X_{1} B_{1} + A_{1} X_{2} B_{2} + A_{2} X_{3} B_{2} = C

(5)

is consistent.

(2)

R_{M} A = 0, R_{A_{1}} C L_{B_{2}} = 0, C_{1} L_{B} = 0 .

(3)

\begin{matrix} r (\begin{matrix} A_{1} & A_{2} & C \end{matrix}) = r (\begin{matrix} A_{1} & A_{2} \end{matrix}), \\ r (\begin{matrix} B_{2} & 0 \\ C & A_{1} \end{matrix}) = r (\begin{matrix} B_{2} \end{matrix}) + r (\begin{matrix} A_{1} \end{matrix}), \\ r (\begin{matrix} C_{1} \\ B_{1} \\ B_{2} \end{matrix}) = r (\begin{matrix} B_{1} \\ B_{2} \end{matrix}) . \end{matrix}

In this case, the general solution to (5) can be expressed as follows:

\begin{matrix} X_{1} = A_{1}^{†} C_{1} B^{†} + L_{A_{1}} V_{1} + V_{2} R_{B}, \\ X_{2} = A_{1}^{†} (C - A_{1} X_{1} B_{1} - A_{2} X_{3} B_{2}) B_{2}^{†} + T_{1} R_{B_{2}} + L_{A_{1}} T_{2}, \\ X_{3} = M^{†} A B_{2}^{†} + L_{M} U_{1} + U_{2} R_{B_{2}}, \end{matrix}

where

U_{1}, U_{2}, V_{1}, V_{2}, T_{1}

, and

T_{2}

are arbitrary matrices over

H

with appropriate sizes.

The following lemma, originally derived by Marsaglia and Styan[30], which can be extended to

H

Lemma 3.

Assume that

A \in H^{m \times n},

B \in H^{m \times k}, C \in H^{l \times n}

D \in H^{j \times k}

and

E \in H^{l \times i}

, then we have the following rank equality:

r (\begin{matrix} A & B L_{D} \\ R_{E} C & 0 \end{matrix}) = r (\begin{matrix} A & B & 0 \\ C & 0 & E \\ 0 & D & 0 \end{matrix}) - r (D) - r (E) .

3. The solution of the matrix equation (1)

In this section, we establish the necessary and sufficient conditions for the solvability of the dual quaternion matrix equation (1) and provide the expressions of its general solution. Additionally, we investigate some special cases of the dual quaternion matrix equation (1).

Theorem 1.

Let

A = A_{0} + A_{1} ϵ \in {DQ}^{m \times n}

B = B_{0} + B_{1} ϵ \in {DQ}^{k \times l}

C = C_{0} + C_{1} ϵ \in {DQ}^{m \times l}

be known. Put

A_{2} = A_{1} L_{A_{0}}, B_{2} = R_{B_{0}} B_{1}, C_{11} = A_{0} A_{0}^{†} C_{0} B_{0}^{†} B_{1},

(6)

C_{22} = A_{1} A_{0}^{†} C_{0} B_{0}^{†} B_{0}, C_{2} = C_{1} - C_{11} - C_{22},

(7)

M = R_{A_{0}} A_{2}, N = R_{A_{0}} C_{2}, E = B_{2} L_{B_{0}}, F = C_{2} L_{B_{0}} .

(8)

Then the following statements are equivalent:

(1)

The dual quaternion matrix equation (1) is consistent.

(2)

\begin{matrix} R_{A_{0}} C_{0} = 0, C_{0} L_{B_{0}} = 0, \end{matrix}

(9)

\begin{matrix} R_{M} N = 0, R_{A_{0}} C_{2} L_{B_{0}} = 0, F L_{E} = 0 . \end{matrix}

(10)

(3)

\begin{matrix} r (\begin{matrix} A_{0} & C_{0} \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}), r (\begin{matrix} B_{0} \\ C_{0} \end{matrix}) = r (\begin{matrix} B_{0} \end{matrix}), \end{matrix}

(11)

\begin{matrix} r (\begin{matrix} A_{1} & A_{0} & C_{1} \\ A_{0} & 0 & C_{0} \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}), \end{matrix}

(12)

\begin{matrix} r (\begin{matrix} C_{1} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}) + r (\begin{matrix} B_{0} \end{matrix}), \end{matrix}

(13)

\begin{matrix} r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{1} & C_{0} \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}) . \end{matrix}

(14)

In this case, the general solution X of dual quaternion matrix equation (1) can be expressed as

X = X_{0} + X_{1} ϵ

, where

\begin{matrix} X_{0} = A_{0}^{†} C_{0} B_{0}^{†} + L_{A_{0}} U + V R_{B_{0}}, \\ X_{1} = A_{0}^{†} (C_{2} - A_{0} V B_{2} - A_{2} U B_{0}) B_{0}^{†} + W_{1} R_{B_{0}} + L_{A_{0}} W_{2}, \\ U = M^{†} N B_{0}^{†} + L_{M} Q_{1} + Q_{2} R_{B_{0}}, \\ V = A_{0}^{†} F E^{†} + L_{A_{0}} Q_{3} + Q_{4} R_{E}, \end{matrix}

(15)

and

Q_{i} (i = \bar{1, 4})

W_{i} (i = \bar{1, 2})

are arbitrary matrices over

H

with appropriate dimensions.

Proof.

(1) \Leftrightarrow (2)

: Suppose the dual quaternion matrix equation (1) is solvable and its solution

X \in {DQ}^{m \times n}

which can be expressed as

X = X_{0} + X_{1} ϵ,

(16)

substituting (16) into (1), by the definition of equality of dual quaternion matrices, we can get that dual quaternion matrix equation (1) is equivalent to the system of quaternion matrix equations

\begin{matrix} \{\begin{matrix} A_{0} X_{0} B_{0} = C_{0}, \\ A_{0} X_{0} B_{1} + A_{0} X_{1} B_{0} + A_{1} X_{0} B_{0} = C_{1} . \end{matrix} \end{matrix}

(17)

The structure of the proof goes as follows. We first prove that

(1) \Leftrightarrow (2)

and illustrate the general solution of (17) have the form of (15), then we prove

(2) \Leftrightarrow (3)

We divide the system (17) into the following:

\begin{matrix} A_{0} X_{0} B_{0} = C_{0}, \end{matrix}

(18)

and

\begin{matrix} A_{0} X_{0} B_{1} + A_{0} X_{1} B_{0} + A_{1} X_{0} B_{0} = C_{1} . \end{matrix}

(19)

By Lemma 1, we obtain (18) is consistent if and only if

R_{A_{0}} C_{0}, C_{0} L_{B_{0}} = 0 .

In this case, the general solution of (18) can be written as

\begin{matrix} X_{0} = A_{0}^{†} C_{0} B_{0}^{†} + L_{A_{0}} U + V R_{B_{0}}, \end{matrix}

(20)

where

U, V

are any matrices over

H

with appropriate sizes.

By substituting equation (20) into equation (19) gives

A_{0} V B_{2} + A_{0} X_{1} B_{0} + A_{2} U B_{0} = C_{2},

(21)

where

A_{2}, B_{2},

and

C_{2}

are defined by (6)-(7). Using Lemma 2 to (21), we know that matrix equation (21) is solvable if and only if

\begin{matrix} R_{M} N = 0, R_{A_{0}} C_{2} L_{B_{0}} = 0, F L_{E} = 0, \end{matrix}

where

M, N, E

and F are given by (8). In this case, the general solution of (21) can be expressed as

\begin{matrix} X_{1} = A_{0}^{†} (C_{2} - A_{0} V B_{2} - A_{2} U B_{0}) B_{0}^{†} + W_{1} R_{B_{0}} + L_{A_{0}} W_{2}, \end{matrix}

(22)

\begin{matrix} U = M^{†} N B_{0}^{†} + L_{M} Q_{1} + Q_{2} R_{B_{0}}, \end{matrix}

(23)

\begin{matrix} V = A_{0}^{†} F E^{†} + L_{A_{0}} Q_{3} + Q_{4} R_{E}, \end{matrix}

(24)

where

Q_{1}, Q_{2}

Q_{3}

Q_{4}

W_{1}

and

W_{2}

are any matrices with the suitable dimensions over

H

. To sum up, we have shown that the matrix equation (1) has a dual solution

X \in {DQ}^{m \times n}

if and only if

(2)

holds.

(2) \Leftrightarrow (3)

: We devide it into two parts to prove its equivalence.

P a r t 1

. In this part, we prove that (9) hold if and only if (11) hold. According to Lemma 3, it is easy to verify that

(9) \Leftrightarrow (11)

P a r t 2

. In this part, we turn to prove that

(10) \Leftrightarrow (12) - (14)

. Let

X_{0} = A_{0}^{†} C_{0} B_{0}^{†} .

Then it is easy to verify that

X_{0}

is a particular solution to the matrix equation

A_{0} X_{0} B_{0} = C_{0} .

By Lemma 3 and block elementary operations, we obtain

\begin{matrix} R_{M} N = 0 & \Leftrightarrow r (R_{M} N) = 0 \Leftrightarrow r (\begin{matrix} R_{A_{0}} A_{2} & R_{A_{0}} C_{2} \end{matrix}) = r (R_{A_{0}} A_{2}), \\ \Leftrightarrow r (\begin{matrix} A_{0} & A_{2} & C_{2} \end{matrix}) = r (\begin{matrix} A_{0} & A_{2} \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} A_{0} & A_{1} L_{A_{0}} & C_{2} \end{matrix}) = r (\begin{matrix} A_{0} & A_{1} L_{A_{0}} \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} A_{1} & A_{0} & C_{2} \\ A_{0} & 0 & 0 \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} A_{1} & A_{0} & C_{1} - A_{0} A_{0}^{†} C_{0} B_{0}^{†} B_{1} - A_{1} A_{0}^{†} C_{0} B_{0}^{†} B_{0} \\ A_{0} & 0 & 0 \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} A_{1} & A_{0} & C_{1} \\ A_{0} & 0 & C_{0} \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}), \end{matrix}

\begin{matrix} R_{A_{0}} C_{2} L_{B_{0}} = 0 & \Leftrightarrow r (R_{A_{0}} C_{2} L_{B_{0}}) = 0 \Leftrightarrow r (\begin{matrix} C_{2} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (A_{0}) + r (B_{0}), \\ \Leftrightarrow r (\begin{matrix} C_{1} - C_{11} - C_{22} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (A_{0}) + r (B_{0}), \\ \Leftrightarrow r (\begin{matrix} C_{1} - A_{0} A_{0}^{†} C_{0} B_{0}^{†} B_{1} - A_{1} A_{0}^{†} C_{0} B_{0}^{†} B_{0} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (A_{0}) + r (B_{0}), \\ \Leftrightarrow r (\begin{matrix} C_{1} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (A_{0}) + r (B_{0}), \\ F L_{E} = 0 & \Leftrightarrow r (F L_{E}) = 0 \Leftrightarrow r (\begin{matrix} E \\ F \end{matrix}) = r (E) \Leftrightarrow r (\begin{matrix} B_{2} L_{B_{0}} \\ C_{2} L_{B_{0}} \end{matrix}) = r (B_{2} L_{B_{0}}), \\ \Leftrightarrow r (\begin{matrix} B_{0} \\ B_{2} \\ C_{2} \end{matrix}) = r (\begin{matrix} B_{0} \\ B_{2} \end{matrix}) \Leftrightarrow r (\begin{matrix} B_{0} \\ R_{B_{0}} B_{1} \\ C_{2} \end{matrix}) = r (\begin{matrix} B_{0} \\ R_{B_{0}} B_{1} \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{2} & 0 \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{1} - A_{0} A_{0}^{†} C_{0} B_{0}^{†} B_{1} - A_{1} A_{0}^{†} C_{0} B_{0}^{†} B_{0} & 0 \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}), \\ \Leftrightarrow r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{1} & C_{0} \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}) . \end{matrix}

□

Now, we consider some special cases of dual quarernion matrix equation(1).

Corollary 1

([31]). Assume that

A = A_{0} + A_{1} ϵ \in {DQ}^{m \times n}

C = C_{0} + C_{1} ϵ \in {DQ}^{m \times l}

are given. Put

A_{2} = A_{1} L_{A_{0}}, C_{22} = A_{1} A_{0}^{†} C_{0}, C_{2} = C_{1} - C_{22}, M = R_{A_{0}} A_{2}, N = R_{A_{0}} C_{2} .

(25)

Then the following statements are equivalent:

(1)

The dual quaternion matrix equation

A X = C

is consistent.

(2)

R_{A_{0}} C_{0} = 0, R_{M} N = 0 .

(26)

(3)

r (\begin{matrix} A_{0} & C_{0} \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}), r (\begin{matrix} A_{1} & A_{0} & C_{1} \\ A_{0} & 0 & C_{0} \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}) .

(27)

In this case, the general solution X of dual quaternion matrix equation

A X = C

can be expressed as

X = X_{0} + X_{1} ϵ

, where

\begin{matrix} X_{0} = A_{0}^{†} C_{0} + L_{A_{0}} U, \\ X_{1} = A_{0}^{†} (C_{2} - A_{2} U) + L_{A_{0}} W_{1}, \\ U = M^{†} N + L_{M} W_{2}, \end{matrix}

(28)

and

W_{1}, W_{2}

are arbitrary matrices over

H

with appropriate dimensions.

Corollary 2

([31]). Let

B = B_{0} + B_{1} ϵ \in {DQ}^{k \times l}

C = C_{0} + C_{1} ϵ \in {DQ}^{m \times l}

be known. Denote

B_{2} = R_{B_{0}} B_{1}, C_{11} = C_{0} B_{0}^{†} B_{1}, C_{2} = C_{1} - C_{11}, E = B_{2} L_{B_{0}}, F = C_{2} L_{B_{0}} .

(29)

Then the following statements are equivalent:

(1)

The dual quaternion matrix equation

X B = C

is consistent.

(2)

C_{0} L_{B_{0}} = 0, F L_{E} = 0 .

(30)

(3)

r (\begin{matrix} B_{0} \\ C_{0} \end{matrix}) = r (\begin{matrix} B_{0} \end{matrix}), r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{1} & C_{0} \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}) .

(31)

In this case, the general solution can be expressed as

X = X_{0} + X_{1} ϵ

, where

\begin{matrix} X_{0} = C_{0} B_{0}^{†} + V R_{B_{0}}, \\ X_{1} = (C_{2} - V B_{2}) B_{0}^{†} + W_{1} R_{B_{0}}, \\ V = F E^{†} + W_{2} R_{E}, \end{matrix}

(32)

and

W_{1}, W_{2}

are arbitrary matrices over

H

with appropriate dimensions.

4. Applications

As an application of Theorem 1, we can investigate the dual quaternion matrix equation (3).

Theorem 2.

Suppose that

A = A_{0} + A_{1} ϵ \in {DQ}^{m \times n}, C = C_{0} + C_{1} ϵ = C^{ϕ} \in {DQ}^{m \times m}

are given, denote

\begin{matrix} B_{2} = R_{A_{0}^{ϕ}} A_{1}^{ϕ}, C_{11} = A_{0} A_{0}^{†} C_{0} {(A_{0}^{ϕ})}^{†} A_{1}^{ϕ}, C_{22} = A_{1} A_{0}^{†} C_{0} {(A_{0}^{ϕ})}^{†} A_{0}^{ϕ}, \\ C_{2} = C_{1} - C_{11} - C_{22}, M = R_{A_{0}} B_{2}^{ϕ}, N = R_{A_{0}} C_{2} . \end{matrix}

Then the following statements are equivalent:

(1)

The dual quaternion matrix equation (3) is consistent.

(2)

The following equalities are satisfied:

\begin{matrix} R_{A_{0}} C_{0} = 0, R_{M} N = 0, R_{A_{0}} C_{2} L_{A_{0}^{ϕ}} = 0 . \end{matrix}

(3)

The following rank equalities hold:

\begin{matrix} r (\begin{matrix} A_{0} & C_{0} \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}), \\ r (\begin{matrix} A_{1} & A_{0} & C_{1} \\ A_{0} & 0 & C_{0} \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}), \\ r (\begin{matrix} C_{1} & A_{0} \\ A_{0}^{ϕ} & 0 \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}) + r (\begin{matrix} A_{0}^{ϕ} \end{matrix}) = 2 r (A_{0}) . \end{matrix}

In this case, the general solution X of (3) can be expressed as

X = X_{0} + X_{1} ϵ

, where

X_{0} = \frac{\tilde{X_{0}} + {\tilde{X_{0}}}^{ϕ}}{2}, X_{1} = \frac{\tilde{X_{1}} + {\tilde{X_{1}}}^{ϕ}}{2}

and

\begin{matrix} X_{0} = A_{0}^{†} C_{0} {(A_{0}^{ϕ})}^{†} + L_{A_{0}} U + V R_{A_{0}^{ϕ}}, \\ X_{1} = A_{0}^{†} (C_{2} - A_{0} V B_{2} - B_{2}^{ϕ} U A_{0}^{ϕ}) {(A_{0}^{ϕ})}^{†} + W_{1} R_{A_{0}^{ϕ}} + L_{A_{0}} W_{2}, \\ U = M^{†} N {(A_{0}^{ϕ})}^{†} + L_{M} Q_{1} + Q_{2} R_{A_{0}^{ϕ}}, \\ V = A_{0}^{†} N^{ϕ} {(M^{ϕ})}^{†} + L_{A_{0}} Q_{3} + Q_{4} R_{M^{ϕ}}, \end{matrix}

Q_{i} (i = \bar{1, 4})

W_{i} (i = \bar{1, 2})

are any matrices with appropriate dimensions over

H

Proof.

By using the definitions of equality of dual quaternion matrices and dual quaternion matrix multiplication, we can conclude that the consistency of the dual quaternion matrix equation (3) is contingent on the existence of the solutions to the system of quaternion matrix equations

\begin{matrix} \{\begin{matrix} A_{0} \tilde{X_{0}} A_{0}^{ϕ} = C_{0}, \\ A_{0} \tilde{X_{0}} A_{1}^{ϕ} + A_{0} \tilde{X_{1}} A_{0}^{ϕ} + A_{1} \tilde{X_{0}} A_{0}^{ϕ} = C_{1} . \end{matrix} \end{matrix}

(33)

In fact, if the matrix equation (3) has a

ϕ

-Hermitian solution

X = X_{0} + X_{1} ϵ

, it is obvious that

X_{0}

and

X_{1}

must be solutions to (33). Conversely, if the system (33) has solutions

\tilde{X_{0}}

and

\tilde{X_{1}}

, then the matrix equation (3) has solution

X = X_{0} + X_{1} ϵ

, where

X_{0} = \frac{\tilde{X_{0}} + {\tilde{X_{0}}}^{ϕ}}{2}, X_{1} = \frac{\tilde{X_{1}} + {\tilde{X_{1}}}^{ϕ}}{2} .

According to Theorem 1, we can present the necessary and sufficient conditions for the solvability of (33), along with the general expression for its solutions. □

5. Numerical example

Now, we give a numerical example to illustrate the main results of this paper.

Example 1.

Given the dual quaternion matrices:

\begin{matrix} A & = A_{0} + A_{1} ϵ = (\begin{matrix} 2 i + k & 0 \\ j & 0 \end{matrix}) + (\begin{matrix} 2 - 3 i + k & i - j \\ - 3 k & i \end{matrix}) ϵ, \\ B & = B_{0} + B_{1} ϵ = (\begin{matrix} 1 + i + j & - j \\ 0 & k + j \end{matrix}) + (\begin{matrix} - i - 2 k & j + 3 k \\ i & k \end{matrix}) ϵ, \\ C & = C_{0} + C_{1} ϵ \\ = (\begin{matrix} - 2 + i + j + 3 k & - 1 + 2 i - 2 j - 4 k \\ - 1 + j - k & 2 + i \end{matrix}) + (\begin{matrix} 11 - 5 i + 5 j - 3 k & - 3 - 8 j + 8 k \\ 1 + i - 5 j - k & 2 - 3 i - j + k \end{matrix}) ϵ . \end{matrix}

Computing directly yields

\begin{matrix} r (\begin{matrix} A_{0} & C_{0} \end{matrix}) = r (A_{0}) = 1, r (\begin{matrix} B_{0} \\ C_{0} \end{matrix}) = r (B_{0}) = 2, \\ r (\begin{matrix} A_{1} & A_{0} & C_{1} \\ A_{0} & 0 & C_{0} \end{matrix}) = r (\begin{matrix} A_{1} & A_{0} \\ A_{0} & 0 \end{matrix}) = 3, \\ r (\begin{matrix} C_{1} & A_{0} \\ B_{0} & 0 \end{matrix}) = r (\begin{matrix} A_{0} \end{matrix}) + r (\begin{matrix} B_{0} \end{matrix}) = 3, \\ r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \\ C_{1} & C_{0} \end{matrix}) = r (\begin{matrix} B_{1} & B_{0} \\ B_{0} & 0 \end{matrix}) = 4 . \end{matrix}

All rank equations are satisfied and the general solution of dual quaternion matrix equation (1) can be expressed as

\begin{matrix} X = X_{0} + X_{1} ϵ = (\begin{matrix} 1 & i \\ k & 0 \end{matrix}) + (\begin{matrix} 0 & k \\ i & 0 \end{matrix}) ϵ . \end{matrix}

6. Conclusions

In this paper, we have established the solvability conditions for the dual quaternion matrix equation (1) by using Moore-Penrose inverses and ranks of matrices, we also have derived the expressions of its general solution to (1) when the solvability conditions are met. As special cases, some dual quaternion matrix equations are also discussed. Moreover, we have investigated the

ϕ

-Hermitian matrix over dual quaternion algebra and provided its related properties, as an application of the aforementioned research, we consider a special case of (1) and give the

ϕ

-Hermitian solutions to (3). Finally, we have presented an example to illustrate the main results.

Author Contributions

Methodology, Y.C. and Q.-W.W.; software,Y.C. and L.-M.X.; writing—original draft preparation, Q.-W.W. and Y.C.; writing—review and editing, Q.-W.W., Y.C. and L.-M.X.; supervision, Q.-W.W.; project administration, Q.-W.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research is supported by the National Natural Science Foundation of China (No. 12371023).

Institutional Review Board Statement

Not applicable

Informed Consent Statement

Not applicable

Data Availability Statement

Not applicable

Conflicts of Interest

The authors declare no conflict of interest.

References

S. DeLeo, G. Scolarici. Right eigenvalue equation in quaternionic quantum mechanics. J. Phys. A. 2000, 33, 2971–2995. [Google Scholar] [CrossRef]
S. C. Pei, C.M. Cheng. Color image processing by using binary quaternion-moment-preserving thresholding techniqe. IEEE Trans. Image Process. 1999, 8, 614–628. [Google Scholar] [CrossRef]
N. LE. Bihan, J. Mars. Singular value decomposition of matrices of quaternions: A new tool for vector-sensor signal processing. IEEE Trans. Signal Process. 2004, 84, 1177–1199. [Google Scholar] [CrossRef]
L. Q. Qi, Z.Y. Luo, Q.W. Wang, X.Z. Zhang. Quaternion matrix optimization: motivation and analysis. J. Optim. Theory Appl. 2022, 193, 621–648. [Google Scholar] [CrossRef]
M. Dehghan, M. Hajarian. On the generalized reflexive and anti-reflexive solutions to a system of matrix equations. Linear Algebra Appl. 2012, 437, 2793–2812. [Google Scholar] [CrossRef]
A. P. Liao, Z.Z. Bai. Least-squares solution of AXB=D over symmetric positive semidefinite matrices X. J. Comput. Math. 2003, 21, 175–182. [Google Scholar] [CrossRef]
L. S. Liu, Q.W. Wang, S.M. Mahmoud. A Sylvester-type Hamilton quaternion matrix equation with an application. Mathematics. 2022, 10, 1758. [Google Scholar] [CrossRef]
Q. W. Wang, Z.H. He, Y. Zhang. Constrained two-side coupled Sylvester-type quaternion matrix equations. Automatica. 2019, 101, 207–213. [Google Scholar] [CrossRef]
L. S. Liu, Q.W. Wang, J.F. Chen. An exact solution to a quaternion matrix equation with an application. Symmetry. 2022, 14, 375. [Google Scholar] [CrossRef]
A. Ben-Israel, T.N.E. Greville. Generalized Inverses: theory and application. JohnWiley and Sons: New, York, USA, 1974. [Google Scholar] [CrossRef]
A. P. Liao, Z.Z. Bai. The constrained solutions of two matrix equations. Acta Math. Appl. Sin. Engl. Ser. 2002, 18, 671–678. [Google Scholar] [CrossRef]
G. X. Huang, F. Yin, K. Guo. An iterative method for the skew-symmetric solution and the optimal approximate solution of the matrix equation AXB=C. J. Comput Appl. Math. 2008, 212, 231–244. [Google Scholar] [CrossRef]
Z. Y. Peng. The centro-symmetric solutions of linear matrix equation AXB=C and its optimal approximation. Chinese J. Eng. Math. 2003, 20, 60–64. [Google Scholar] [CrossRef]
Y. B. Deng, Z.Z. Bai, Y.H. Gao. Iterative orthogonal direction methods for Hermitian minimum norm solutions of two consistent matrix equations. Numer. Linear Algebra Appl. 2006, 13, 801–823. [Google Scholar] [CrossRef]
M. Y. Xie, Q.W. Wang. The reducible solution to a quaternion tensor equation. Front. Math. China. 2020, 15, 1047–1070. [Google Scholar] [CrossRef]
D. S. Cvetković-IIić, A. Dajić , J.J. Koliha. Positive and real-positive solutions to the equation axa^*=c in C^*-algebras. Linear Multilinear Algebra. 2007, 55, 535–543. [Google Scholar] [CrossRef]
J. Größ. A note on the general Hermitian solution to AXA^*=B. Bull. Malays. Math. Sci. Soc. 1998, 21, 57–62. [Google Scholar]
J. K. Baksalary. Nonnegative definite and positive definite solutions to the matrix equation AXA^*=B. Linear Multilinear Algebra. 1984, 16, 133–139. [Google Scholar] [CrossRef]
J. Größ. Nonnegative-definite and positive-definite solutions to the matrix equation AXA^*=B revisited. Linear Algebra Appl. 2000, 321, 123–129. [Google Scholar] [CrossRef]
C. C.Took; D.P.Mandic. On the unitary diagonalisation of a special class of quaternion matrices. Appl. Math. Lett. 2011, 24, 1806–1809. [Google Scholar] [CrossRef]
C. C. Took, D.P. Mandic. Augmented second-order statistics of quaternion random signals. Signal Process. 2011, 91, 214–224. [Google Scholar] [CrossRef]
Ivan Kyrchei. Cramer’s rules of eta -(skew-) Hermitian solutions to the quaternion Sylvester-type matrix equations. Adv. Appl. Clifford Algebr. 2019, 29, 1–31. [Google Scholar] [CrossRef]
G. Brambley, J. Kim. Unit dual quaternion-based pose optimization for visual runway observations. Iet Cyber Syst. Robot. 2020, 2, 181–189. [Google Scholar] [CrossRef]
J. Cheng, J.Kim, Z. Jiang, W. Che. Dual quaternion-based graph SLAM. Robot. Auton. Syst. 2016, 77, 15–24. [Google Scholar] [CrossRef]
K. Daniilidis. Hand-eye calibration using dual quaternions. Int. J. Robot. Res. 1999, 18, 286–298. [Google Scholar] [CrossRef]
X. Wang, C. Yu, Z. Lin. A dual quaternion solution to attitude and position control for rigid body coordination. IEEE Trans. Rob. 2012, 28, 1162–1170. [Google Scholar] [CrossRef]
L. Rodman. Topics in quaternion linear algebra. Princeton University Press 2014. [Google Scholar]
L. Q. Qi, C. Ling, H. Yan. Dual quaternions and dual quaternion vectors. Commun. Appl. Math. Comput. 2022, 4, 1494–1508. [Google Scholar] [CrossRef]
Z. H. He , Q.W.Wang. The general solutions to some systems of matrix equations. Linear Multilinear Algebra. 2015, 63, 2017–2032. [Google Scholar] [CrossRef]
G. Marsaglia, G.P. Styan. Equalities and inequalities for ranks of matrices. Linear Multilinear Algebra. 1974, 2, 269–292. [Google Scholar] [CrossRef]
L.M. Xie, Q.W. Wang. A system of dual quaternion matrix equations with its applications. arXiv:2312.10037, arXiv:2312.10037. https://arxiv.org/abs/2312.10037.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.

MDPI Initiatives

Important Links

Choose an area of interest and we will send you notifications of new preprints at your preferred frequency.

Disclaimer

The Dual Quaternion Matrix Equation AXB=C with Applications

Abstract

1. Introduction

2. Preliminaries

3. The solution of the matrix equation (1)

4. Applications

5. Numerical example

6. Conclusions

Author Contributions

Funding

Institutional Review Board Statement

Informed Consent Statement

Data Availability Statement

Conflicts of Interest

References

MDPI Initiatives

Important Links

Subscribe