IEEE Systems, Man and Cybernetics Magazine - January 2023 - 33

Z
Y
U VS pS pS p
pp
x^h
=- -
-||
|
||
xz,
z 11 1
N
== =
M
y
m11
VV " , ,
}}nx
==
mx x
c
,, .
n
x
e
In (1), parameter ^h= 12 X ., f
$$
$$
xz
n
, ,,
l
yx x
d
lx
L
x
s
(1)
, denotes the set
xz,
S ,,yx
where
due to the diversity of service demand, each broker has an
opportunity to make a personalized decision to quote a
competitive price for the service product it is good at and,
thus, gains a certain share in the service product market.
of all the brokers; parameter S ,,xz where S ! " ,01,
denotes whether the secondary user z procures the product
from the broker x or not; parameter
yx,
spectrum status data to the broker x or not, while parameter
SS
,01, denotes whether the idle smart
lx lx
,,
^h! "
device l sells the rental intention information to the broker
x or not; and parameters
mx ,
nx ,
} and } denote the
computing resources rented by the broker x from the
cloud facility m and the edge device n, respectively, which
are positively correlated with the proportion of its own
bid in the total bid of all bidders. The meanings of the
parameters
p ,,xz
n
p , x
d
p , x
s
p , x
c
and px
e
have been explained
earlier, and they are quotes per unit of commodity.
If the brokers are cooperative, they will monopolize the
market; that is, the brokers will set the same price to maximize
their total revenue. In practice, they are usually selfish,
and each broker tries to gain as much revenue as
possible for itself. The revenue of broker xx V !^h
depends on not only the product prices it offers to the secondary
users and its costs of the purchased data and the
rented resources but also the product prices offered by the
other brokers V xand
the costs of the purchased data and
the rented resources paid by the other brokers V .xSince
a secondary user usually buys a product from the
lowest bidder and a primary user (or idle smart device) usually
sells its spectrum status data (or its rental intention
information) to the highest bidder, all the brokers form a
noncooperative game relationship [24]. Also, since a primary
user's spectrum status data (or an idle smart device's
rental intention information) are regarded as commodities,
a primary user (or an idle smart device) is allowed to sell its
data (or information) only to one buyer. Therefore, a broker
x must pay attention to the maximum quote from the other
brokers to determine its own quote px
d
(or px
s
(or px
s
). If px
d
) is lower than the maximum quote of the other brokers,
the broker x will lose the spectrum status data (or its
rental intention information) market. On the contrary, if
px
d
(or px
s
) is higher than the maximum quote of the other
brokers, the other brokers will compete to bid higher for the
corresponding market share.
Through the aforementioned competition for quotation,
the weak brokers will be eliminated by the market, while
equally powerful brokers will compromise with each other
to adopt the same data quote (or the same communication
resource quote). The ability to design high-quality machine
learning solutions is one of major factors in terms of measuring
the strength of brokers. For the same type of service
products, the quotation competition relationship of each
broker is also similar to that mentioned earlier. However,
S ! " ,01, denotes whether the primary user y sells the
Performance Analysis for Cognitive
Network Architecture Systems
Performance Analysis
In this section, we focus on the effects of the use of cognitive
engines on the performance of constructed cognitive
networks. Therefore, in the two schemes adopted in
the simulation for comparison, one uses the two cognitive
engines to analyze the collected spectrum status
data and the received user requirements for cognitive network
construction, respectively, while the other employs
the collected data directly. Apart from the aforementioned
differences, the two schemes are the same in all
other respects.
The performance metrics involve throughput, delay,
energy efficiency, and continuous service capacity.
Throughput is defined as the successfully transmitted
data delivered between a cognitive base station (or access
point) and its associated cognitive users during a given
unit time, while delay means the time it takes per unit of
data from the source node to the destination node. Energy
efficiency is defined as the data rate per unit transmitting
energy consumption, while continuous service capacity is
the amount of data successfully delivered between a cognitive
base station (or access point) and its associated
cognitive users before the transmission path is disconnected,
which can be used to roughly measure network
life. The number of brokers is set to 10. Moreover, the
numbers of edge devices and cloud facilities in the simulation
network are set to 20 and five, respectively, while the
numbers of primary small base stations (i.e., licensed
spectrum owners) and demanders of cognitive networks
in a macro cell are set to 30 and 10, respectively. The value
of p ,xz
n
p ,x
d
p , x
s
ranges from 100 to 1,000, while the value of each
ranges from 10 to 100. In addition,
mx ,
p , x
c
and px
e
the value of each of } and } ranges from one to 10.
nx ,
When each constructed cognitive network serves an average
of 100 cognitive users and each user has the same
number of IoT-based applications, the corresponding
results are shown in Figure 4.
From Figure 4(a) and (b), we can observe that the
scheme with cognitive engines can have larger throughput
and achieve lower delay. This is due to the fact that it is
easy for a network builder to obtain the better cognitive
spectrum and the more accurate networking requirements
through the cognitive engines. An aspect of better cognitive
spectrum quality is that a cognitive user uses it for a
longer time and is not frequently asked for channel switching.
Thus, transmission opportunities on the cognitive
spectrum can be fully utilized. From Figure 4(c) and 4(d),
we can also see that the scheme with the cognitive engines
January 2023 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE 33
IEEE Systems, Man and Cybernetics Magazine - January 2023

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - January 2023
