1.定义
假设系统有一状态集 E = 0 , 1 , 2... K E={0,1,2...K} E=0,1,2...K,令 N ( t ) N(t) N(t)表示系统在 t t t时刻所处的状态,则有以下结论:
p
i
,
i
+
1
(
Δ
t
)
=
P
(
N
(
t
+
Δ
t
)
=
i
+
1
∣
N
(
t
)
=
i
)
=
λ
i
Δ
t
+
o
(
Δ
t
)
p_{i,i+1}(\Delta t)=P(N(t+\Delta t)=i+1|N(t)=i)=\lambda_i\Delta t+o(\Delta t)
pi,i+1(Δt)=P(N(t+Δt)=i+1∣N(t)=i)=λiΔt+o(Δt)
p
i
,
i
−
1
(
Δ
t
)
=
P
(
N
(
t
+
Δ
t
)
=
i
−
1
∣
N
(
t
)
=
i
)
=
μ
i
Δ
t
+
o
(
Δ
t
)
p_{i,i-1}(\Delta t)=P(N(t+\Delta t)=i-1|N(t)=i)=\mu_i\Delta t+o(\Delta t)
pi,i−1(Δt)=P(N(t+Δt)=i−1∣N(t)=i)=μiΔt+o(Δt)
p
i
,
j
(
Δ
t
)
=
P
(
N
(
t
+
Δ
t
)
=
j
∣
N
(
t
)
=
i
)
=
o
(
Δ
t
)
∣
i
−
j
∣
>
2
p_{i,j}(\Delta t)=P(N(t+\Delta t)=j|N(t)=i)=o(\Delta t) |i-j|>2
pi,j(Δt)=P(N(t+Δt)=j∣N(t)=i)=o(Δt)∣i−j∣>2
其中,
λ
i
>
0
,
i
=
0
,
1
,
2.....
K
−
1
,
μ
i
>
0
,
i
=
1
,
2
,
3....
K
\lambda_i>0,i=0,1,2.....K-1,\mu_i>0,i=1,2,3....K
λi>0,i=0,1,2.....K−1,μi>0,i=1,2,3....K,均为常数,则称随机过程
N
(
t
)
,
t
>
0
{N(t),t>0}
N(t),t>0为有限状态
E
=
0
,
1
,
2...
K
E={0,1,2...K}
E=0,1,2...K上的生灭过程。生灭过程是一个特殊的马尔科夫过程。
2.平稳分布
令
p
j
(
t
)
=
P
(
N
(
t
)
=
j
)
,
j
∈
E
p_j(t)=P(N(t)=j),j\in E
pj(t)=P(N(t)=j),j∈E,那么由全概率公式,有:
p
j
(
t
)
=
p
j
(
t
)
[
1
−
λ
i
Δ
t
−
μ
i
Δ
t
−
o
(
Δ
t
)
]
+
p
j
−
1
(
t
)
[
λ
i
Δ
t
+
o
(
Δ
t
)
]
+
p
j
+
1
(
t
)
[
μ
i
Δ
t
+
o
(
Δ
t
)
]
+
∑
i
−
j
≥
2
p
i
(
t
)
o
(
Δ
t
)
p_j(t)=p_j(t)[1-\lambda_i\Delta t-\mu_i\Delta t-o(\Delta t)]+p_{j-1}(t)[\lambda_i\Delta t+o(\Delta t)]+p_{j+1}(t)[\mu_i\Delta t+o(\Delta t)]+\sum_{i-j\ge 2} p_i(t)o(\Delta t)
pj(t)=pj(t)[1−λiΔt−μiΔt−o(Δt)]+pj−1(t)[λiΔt+o(Δt)]+pj+1(t)[μiΔt+o(Δt)]+∑i−j≥2pi(t)o(Δt)
=
p
j
(
t
)
[
1
−
λ
i
Δ
t
−
μ
i
Δ
t
]
+
p
j
−
1
(
t
)
[
λ
i
Δ
t
]
+
p
j
+
1
(
t
)
[
μ
i
Δ
t
]
=p_j(t)[1-\lambda_i\Delta t-\mu_i\Delta t]+p_{j-1}(t)[\lambda_i\Delta t]+p_{j+1}(t)[\mu_i\Delta t]
=pj(t)[1−λiΔt−μiΔt]+pj−1(t)[λiΔt]+pj+1(t)[μiΔt]
令
p
j
(
t
)
=
lim
t
→
+
∞
p
j
(
t
)
p_j(t)=\lim_{t\rightarrow+\infty}p_j(t)
pj(t)=limt→+∞pj(t), {
p
j
,
j
=
0
,
1
,
.
.
.
.
K
p_j,j=0,1,....K
pj,j=0,1,....K}存在,与初始条件无关,且
p
j
>
0
,
∑
j
=
0
j
=
K
p
j
=
1
p_j>0,\sum_{j=0}^{j=K}p_j=1
pj>0,∑j=0j=Kpj=1,即{
p
j
,
j
=
0
,
1
,
.
.
.
.
K
p_j,j=0,1,....K
pj,j=0,1,....K}为平稳分布。
3.性质
- p j P j , j + 1 = p j + 1 P j + 1 , j p_jP_{j,j+1}=p_{j+1}P_{j+1,j} pjPj,j+1=pj+1Pj+1,j
- 求各个状态的概率
结合 ∑ j = 0 j = K p j = 1 \sum_{j=0}^{j=K}p_j=1 ∑j=0j=Kpj=1可知, p j = λ 0 λ 1 . . . λ K − 1 μ 1 μ 2 . . . μ K p 0 , p_j=\frac{\lambda_0\lambda_1...\lambda_{K-1}}{\mu_1\mu_2...\mu_K}p_0, pj=μ1μ2...μKλ0λ1...λK−1p0,其中
p 0 = 1 1 + ∑ i = 0 i = K λ 0 λ 1 . . . λ i − 1 μ 1 μ 2 . . . μ i p_0=\frac{1}{1+\sum_{i=0}^{i=K}\frac{\lambda_0\lambda_1...\lambda_{i-1}}{\mu_1\mu_2...\mu_i}} p0=1+∑i=0i=Kμ1μ2...μiλ0λ1...λi−11
4.例子
- 2-state birth-death process
{ p 0 + p 1 = 1 p 0 α = p 1 β ⇒ p 0 = β α + β , p 1 = α α + β \begin{cases} & p_0+p_1=1\\ & p_0\alpha=p_1\beta \end{cases}\Rightarrow p_0=\frac{\beta}{\alpha+\beta},p_1=\frac{\alpha}{\alpha+\beta} {p0+p1=1p0α=p1β⇒p0=α+ββ,p1=α+βα - 例二
M是生成矩阵,将对应的向量相乘,我们可以得到
p 0 ∗ − N α + p 1 β = 0 , p 0 ∗ N α − p 1 [ β + ( N − 1 ) α ] + p 3 2 β = 0 p_0*-N\alpha+p_1\beta=0,p_0*N\alpha-p_1[\beta+(N-1)\alpha]+p_32\beta=0 p0∗−Nα+p1β=0,p0∗Nα−p1[β+(N−1)α]+p32β=0…
每个矩阵都是对应状态的平衡方程。
平衡方程。