1. Introduction
The Markov chains of M/G/1 type represent a basic model that explains the temporal evolution of population size. It is a two-component process, the first component of which is an integer-valued process describing the dynamics of population size, and the other represents the state (phase) of the external environment. The state space of Markov chains of M/G/1 type can be partitioned into non-empty disjoint subsets known as levels, such that one-step state transitions are limited to states at the same level, the adjacent lower level, or any higher level. The transition matrix of a Markov chain of M/G/1 type has a block upper Hessenberg form and all blocks along the main diagonal, except for the top ones, are identical.
The matrix geometric method proposed by M.F. Neuts in [
1] and his study of Markov chains of M/G/1 type in [
2] led to the rapid development of matrix-analytic methods in operations research. These methods provide a powerful framework for the unified analysis of large classes of Markov processes and, more importantly, for their numerical solution. Matrix-analytic methods and their applications as they stand today are outlined in [
3,
4,
5,
6,
7,
8,
9,
10,
11].
In this paper, we study discrete-time Markov chains on the state space , where is the set of nonnegative integers and . The number of elements in may be either finite or infinite. One-step transitions of the process from a state are limited to states such that where represents the vector of all 1s. We assume that the process is spatially homogeneous, meaning that for the one-step transition probability from a state to a state may depend on , and but not on the specific values of and We will refer to these processes as M-dimensional Markov chains of M/G/1 type . The processes are Markov chains of M/G/1 type, with the level consisting of states for which .
Multidimensional quasi-birth-and-death processes (
) are a specific type of multidimensional Markov chains of M/G/1 type. They are characterized by having one-step transitions from a state
restricted to states
such that
. The explicit analytical representation for the stationary distribution of
processes are unknown, and most work has been devoted to deriving asymptotic formulas of the stationary distribution (see [
12,
13,
14]). The conditions ensuring a positive recurrent or transient 2d-QBD process were analyzed in [
15]. Specific cases of
processes, where only one component of the vector
may change at a time, were studied in [
16].
In the theory of Markov chains of M/G/1 type, a crucial role is played by the so-called matrix
[
2]. If it is known, many relevant quantities may be efficiently computed. Different algorithms for the numerical computation of the matrix
have been proposed [
2,
17,
18,
19,
20,
21]. However, the structure of the levels of
processes can make it challenging to use these methods effectively.
This study aims to describe the matrix
of
processes in terms of matrices of order
.
Section 2 explores classical Markov chains of M/G/1 type.
Section 3 defines the multidimensional Markov chains of M/G/1 type and introduces the concept of the state sectors and the sector exit probabilities. We analyze the first passage probabilities and demonstrate that the family of matrices representing the sector exit probabilities satisfies a system of nonlinear matrix equations.
Section 4 illustrates how a solution to this system can be obtained through successive substitution. An example of the
Md-QBD process is discussed in
Section 5. Finally, we offer some concluding remarks in
Section 6.
We use bold capital letters to denote matrices and bold lowercase letters to denote vectors. Unless otherwise stated, all vectors in this paper have integer components and the length . For any vector we use the notation for the component of . For vectors and , means that for all , and means that for at least one value of . Notations and are defined similarly. Vector represents the vector of all 1s, and the vector indicates the vector with zero entries except the mth entry, which equals one. Given a vector and an integer , we define two sets and . Sets , , and are defined as , , and We refer to the set of states as the sector.
2. Markov Chains of M/G/1 Type
A Markov chain of M/G/1 type with a state space
is characterized by a block Hessenberg transition matrix of the form
where blocks
and
are nonnegative square matrices, such that
and
are stochastic matrices. Entries of the matrix
blocks are indexed by the elements of the phase space
. The subset of all states
, such that
, is called the level
.
A fundamental role in the theory of Markov chains of M/G/1 type is played by the matrix
, described by Neuts in [
2]. The element
of this matrix represents the probability that, starting from the state
, the chain will first appear at the level
in state
. Neuts demonstrated that
is the minimal nonnegative solution to the equation
There are several algorithms for calculating the invariant measures of the transition matrix
using the matrix
[
22,
23,
24].
3. Multidimensional Processes of M/G/1 Type
Consider an irreducible discrete-time Markov chain
on the state space
. Let us denote the probability of a one-step transition from
to
as
. We assume that the transition probability matrix
,
, partitioned into blocks
of order
, for all
has the following properties:
where
,
, are nonnegative square matrices such that
is a stochastic matrix. We refer to this process as an
M-dimensional Markov chain of M/G/1 type
3.1. Sectors of the Process States
Let us define the sequence of passage times as follows:
We say that at time
, the process
is in the sector
if
and if
. For a vector
, we define the sector
exit time as the moment
when the process first enters the set
,
Additionally, we define the number of sectors visited along a path to
as
If an initial state of the process belongs to , then we have and If with , then at the first hitting time of , the process exits the sector and enters the sector , which implies equality . The set is reached at the moment of transition from the set to the state
For vectors , , and we define matrices and as follows.
The element
of the matrix
is the conditional probability that the process
, starting in the state
, reaches the set
by hitting the state
after passing through exactly
sectors,
The element
of the matrix
is the conditional probability that the process
will eventually hit the set
in the state
, given that it starts in the state
,
It is clear that matrices
and
,
are related to each other by the equality
For
, any path of
leading from a state
to a state
must successively visit sets
, which will require visiting at least
sectors. Therefore, we have
Since the process
is spatially homogeneous, for any vector
and any vector
, the probabilities
may depend on
, and
but not on the specific values of
,
and
, i.e.,
This means that the matrix
may be expressed as
where the matrix
is defined as
independently of the vector
.
The element
of the matrix
is the conditional probability that the process
will hit the state
on the first visit to the set
, given that it starts in the state
Here, the vectors
and
satisfy conditions
and
. Therefore, the index
of a matrix
is a nonnegative vector and its index
belongs to the set
defined as
We refer to the matrices
as the matrices of the first passage probabilities.
The matrices
determine the transition probabilities of the embedded Markov chain
, since for
we have
We define matrices
as
and refer to these matrices as the matrices of the sector exit probabilities.
3.2. Matrices of the First Passage Probabilities
Theorem 1. The matrices of the first passage probabilities satisfy the system Proof of Theorem 1. We will initially demonstrate the validity of the following equality for all vectors
and
:
This formula adheres to the law of total probability, taking into account all possible states of the process following the first transition. Consider two states: and . The state can be reached from the state after a single transition. This contributes to (12) the term .The first transition may take the process to some state with . To reach the set from state , the process must necessarily cross one or more sectors and then hit the state . This contributes to the second term on the right-hand side of (12). Equation (10) for the matrices is derived from (12) using Formulas (2) and (7). □
For vectors and , let the set be defined as the set of all tuples , satisfying …, , and , and let the set be defined as .
Any nonnegative vector can be represented as , with , and the set being defined as . Since the inequality holds for all vectors , only elements of the set in quantity can satisfy the condition . Therefore, for any , and , the set is empty. If this set is not empty, then the following decomposition is valid in terms of the Cartesian products of sets of the form , where , , and ,
Lemma 1. For any vectors , and an integer the sets , , and the set can be represented as follows: Proof of Lemma 1. Since
and
, for any
tuple
of the length
, in the interval
there necessarily exist numbers
, such that
. Let us denote by
the minimum of such numbers. The number
of the elements
, satisfying the condition
, must be at least
. Therefore, the set
can be represented as follows:
Each element
of the set
can be written as
where
. Therefore, Equation (15) implies the following equalities:
Thus, for any integers
, the following formula is valid
Applying this formula for
, we can obtain the decomposition (13) for the set
.
Introducing a new variable
in (17) and changing the order of summation, we obtain the following result:
Applying this formula for
, we can achieve the decomposition (14) for the set
, which proves Lemma 1. □
Theorem 2. For and , the matrices are given by Proof of Theorem 2. Entries of matrices
are the one-step transition probabilities (9) of the embedded Markov chain
, while entries of the matrices
are the probabilities of reaching the set
after
steps of this Markov chain. Therefore, for all
and
, the matrices
are completely determined by the matrices
,
, as follows:
Here, the summation extends over the set
of all
tuples
, satisfying
,
,
,…,
,
.
From (5) and (19), it follows that for all
and
, we have
Using Formulas (7) and (20), we can obtain the matrix
as
Let us introduce new variables:
,
,
, …,
and
. From the definition of the set
, it follows that vectors
belong to the set
and satisfy the following conditions:
,
…, and
. Taking into account that
, we finally obtain (18). □
It follows from Lemma 1 and Theorem 2 that matrices with arbitrary nonnegative index can be expressed in terms of the matrices , with .
Corollary 1. For any vectors , and integers , the matrix can be represented as Proof of Corollary 1. It follows from (13) and (18) that the matrix
can be represented as follows:
After applying Formula (18) to each sum inside the square brackets in (23), we obtain Formula (22), which proves Corollary 1. □
3.3. Md-M/G/1 Processes as One-Dimensional Markov Chains of M/G/1 Type
Let
be a
Md-M/G/1 process and processes
and
be defined, respectively, as
and
. Consider a discrete-time process
on the state space
, where the set
is defined as
. The process
is the Markov chain of M/G/1 type with the phase space
. Mapping
, with
, is a bijection from
onto
. When the process
is in a state
, the process
is in the state
. Therefore, the probability of a one-step transition of
from
to
is given by
Element
of the matrix
of the process
represents the probability that, starting from the state
, the process
will first appear at the level
in state
. It may be expressed by elements
of the matrix
of the process
as
Let us partition the matrix
into blocks of order
:
,
As a result of the equality
, Equation (24) can be expressed in matrix form as
It implies that the matrices
are expressed through the matrices
as
From (26) and Theorem 2, it follows that
Hence, the matrices
uniquely define the matrix
and vice versa.
4. Matrices of the Sector Exit Probabilities
As a direct consequence of Theorem 1, we can derive the following representation for the matrices of the sector exit probabilities
:
By combining Formulas (11) and (12), we obtain the system of nonlinear equations for the matrices
of the sector exit probabilities:
This system may be solved by successive substitutions, starting with zero matrices.
Theorem 3. Let , , be families of matrices of order , recursively defined byThen, each sequence
is element-wise monotonically increasing and converges to a nonnegative matrix
. The family of matrices
,
, is the minimal solution of the systemin the set of families , , of nonnegative matrices of order .
Proof of Theorem 3. We first demonstrate that the sequences , , are monotonically increasing and satisfy for all and for all . We proceed using induction.
Since
and
, we know that
. Let us assume that
for some
and for all
. Then, it follows from (30) that
and
which proves the induction step. Thus, each entry in the sequence
,
is bounded and monotonically increasing for every
. This implies the existence of the limits
,
, satisfying the system (31).
Assume that
,
, is another nonnegative solution of (31). We show via induction that
for all
and all
. Since
, we know that
for all
. Now, let us assume that
for some
and all
. Then, we obtain inequalities
Therefore,
for all
and all
, which proves the induction step and the minimality property of the family
,
. □
Note that in one-dimensional cases, when , the sets and are singletons, and we have equality . In these cases, a system of matrix Equations (29) consists of a single Equation (1).
5. Example of the Md-QBD Process
Consider a
Md-QBD process
characterized by having one-step transitions from a state
restricted to states
such that
, where
. Matrices of transition probabilities
of such process have the following form:
where
,
,
are nonnegative square matrices such that
is a stochastic matrix. In this case, the representation (28) for the matrices of the sector exit probabilities has the following form:
We show that the matrices , , in (35) are nonzero only if the vector has a single negative component. For any vector , the element of the matrix is the conditional probability that the process will visit the state on the first visit to the set , given that it starts in the state The probability is nonzero if and only if there is some state , the probability of transition from which to the state is positive. For this, it is necessary that the matrix be nonzero. Since and , the vector has negative components. Since , we have either and for some , or and for some and . In both cases, the vector has a single negative component .
Let us denote by
the set of vectors
with a single negative component and by
the set of vectors with a large number of negative components. Given the above, the matrices
are zero for all
and
. In addition, for
matrices
are also zero. Therefore, as follows from equality (35), the matrices
,
are also zero. After removing the zero matrices, system (29) acquires the following form:
where the sets
are defined as
Specific cases of
processes, where only one component of the vector
can change at a time, were studied in [
16]. In these cases, the matrices
are zero for all indices
. Equation (18) in [
16] applies only when
as there is an error in the proof of Theorem 3. As a result, the findings presented in
Section 3 of the mentioned work that rely on this theorem are only valid in the one-dimensional case. For multidimensional
processes, Formula (27) of this article provides the correct matrix-multiplicative representation of the matrix
through matrices of order
, utilizing the solution of the system (36).
6. Conclusions and Future Work
This study presents several theoretical results that aim to simplify the analysis of multidimensional Markov chains of M/G/1 type. The novelty is that we have introduced new concepts of the state sectors and the sector exit probabilities. We have been able to demonstrate that Equation (1), which involves a matrix
, can be replaced by a system of Equation (29) that utilizes a family of matrices
of the sector exit probabilities. Entries of the matrix
are indexed by elements of the set
, where
and
, while the family of the matrices
of the order
is indexed by the set
In one-dimensional cases, the results of the article are reduced to the existing results by Neuts in [
2].
However, there are several challenges to overcome in practically implementing these results. In multidimensional cases, the family of the matrices is infinite. Since it is not feasible to compute infinite families of matrices, future research should focus on developing a method to select an appropriate truncation approximation for the proposed algorithm and conducting complexity and error analyses. It remains unclear whether in multi-dimensional cases the family of matrices is the minimal nonnegative solution to the system (29). Future research should also address this question.