Mathematical Problems in Engineering
Inputs
Inputs
Outputs
Activation
function
Output
Neuron
Model of a neuron
in layered RNN
Input
layer
Hidden
layer
Output
layer
W
b
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
z
1,k+1
z
2,k+1
z
N,k+1
y
1,k
y
2,k
y
l,k
W
a
W
c
z
1,k
z
2,k
z
N,k
u
1,k
u
2,k
u
m,k
i
N
2
1
Ψ
i
(·)
z
i,k+1
z
−1
z
−1
z
−1
Σ
Σ
Σ
Σ
(a)
WNCN
Actuator
Neuron
Plant
Model of a neuron
in WNCN
Sensor
Output
.
.
.
i
z
i,k+1
s
1
a
1
a
2
s
2
s
3
s
l
a
m
1
2
3
4
5
6
7
8
9
10
11
12
s
q
a
p
i
z
−1
z
i,k
(b)
F : (a) A three-layer fully connected RNN. (b) Example of WNCN with neurons consisting of a wireless mesh network (WMN)
where dashed lines represent radio connectivity.
2. Problem Formulation
2.1. Delayed Standard Neural Network Model. Consider the
following discrete-time DSNNM with input-output:
P
(
+1
)
=
(
)
+
𝑑
(
−
)
+
𝜙
(
(
))
+
𝜔
(
)
+
𝑢
(
)
,
(
)
=
𝜀
(
)
+
𝑑
(
−
)
+
𝜙
(
(
))
+
𝜔
(
)
+
𝑢
(
)
,
(
)
=
𝑦
(
)
,
()
with the initial condition function ()=(), ∀∈ [−,0],
where ()∈R
𝑛
is the state vector, ()∈R
𝑚
is the control
input vector, ()∈R
𝑙
isthemeasuredoutputvector,()∈
R
𝑟
is the disturbance that belongs to
2
[0,∞), (()) ∈
(R
𝐿
;R
𝐿
) is the activation function with the input vector
() ∈ R
𝐿
, ∈R is the number of nonlinear activation
functions, >0isthetimedelay,∈R
𝑛×𝑛
,
𝑑
∈ R
𝑛×𝑛
,
𝜙
∈ R
𝑛×𝐿
,
𝜔
∈ R
𝑛×𝑟
,
𝑢
∈ R
𝑛×𝑚
,
𝜀
∈ R
𝐿×𝑛
,
𝑑
∈ R
𝐿×𝑛
,
𝜙
∈R
𝐿×𝐿
,
𝜔
∈R
𝐿×𝑟
,
𝑢
∈R
𝐿×𝑚
,and
𝑦
∈R
𝑙×𝑛
. Assume
that the activation function satises (0) = 0and belongs
to a type of set ()as follows:
(
)
|0≤
𝑖
𝑖
(
)
𝑖
(
)
≤
𝑖
,=1,...,,
()
which means that is sector restricted to the interval [0,],
where =diag(
1
,...,
𝐿
)0.
2.2. Wireless Neural Control Network. e traditional design
approaches of dynamic controllers based on SNNMs are
centralized and the dimension of controller and plant must
remain consistent [–]. However, without losing system
stability, the WNCN can be structured as a distributed recur-
rent neurocontroller (RNC) with arbitrary dimension which
will be in favor of controlling the complex nonlinear sys-
tems with multiple geographically distributed sensors (multi-
output) and actuators (multi-input). So, the motivation for
introducing WNCN stems from the need for distributed
control approaches for WNCSs.
Assume that we use a WNCN consisting of neuron
nodes to control the aforementioned DSNNM P.ewire-
less network in the whole system can be described by a
directed graph as follows:
G A ∪V ∪S
vertex set
,E
I
∪E
C
∪E
O
edge set
,
()
where V ={V
1
,...,V
𝑁
}is the set of neuron nodes, A =
{
1
,...,
𝑚
}is the set of actuators which can execute the
input vector () = [
1
(),...,
𝑚
()]
𝑇
, S ={
1
,...,
𝑙
}
is the set of sensors used to measure the output vector
() = [
1
(),...,
𝑙
()]
𝑇
,andedgesetsE
I
={(V
𝑖
,
𝑝
)|
V
𝑖
∈ V,
𝑝
∈ A}, E
C
={(V
𝑖
,V
𝑗
)|V
𝑗
,V
𝑖
∈ V},and
E
O
={(
𝑞
,V
𝑖
)|
𝑞
∈ S,V
𝑖
∈ V}correspond to the physical
radio communication links in the wireless network. Dene
the following three sets:
the neighbor sensors of V
𝑖
, ∀∈{1,...,}
S
V
𝑖
𝑞
|
𝑞
∈S,∃
𝑞
,V
𝑖
∈E
O
=
𝑞
|
O
𝑖𝑞
=0,
()
the neighbor neurons of
𝑝
, ∀∈{1,...,}
V
𝑎
𝑝
V
𝑖
|V
𝑖
∈V,∃
V
𝑖
,
𝑝
∈E
I
=
V
𝑖
|
I
𝑝𝑖
=0
,
()
the neighbor neurons of V
𝑖
, ∀∈{1,...,}
V
V
𝑖
V
𝑗
|V
𝑗
∈V,∃
V
𝑗
,V
𝑖
∈E
C
=
V
𝑗
|
C
𝑖𝑗
=0
,
()