IEEE Systems, Man and Cybernetics Magazine - January 2023 - 32

A Game-Based Incentive Mechanism for
Cognitive Network Construction
The cognitive network brokers build a cognitive network
infrastructure platform by renting spectrum, communication
devices, and computing resources in future wireless
networks and then provide cognitive network construction
services to secondary users for gaining revenue through
the auction process. On the one hand, this platform needs
to design an incentive mechanism to encourage resource
owners to sell unused resources to it. On the other hand, it
wants to sell cognitive network construction services to
secondary users at a higher price. Such an interaction is
suitable for being modeled as a game model.
Therefore, we present a game-based incentive mechanism
for future cognitive network construction, which is
helpful to use various resources to improve the stability of
future cognitive network topology. We assume that there
are X cognitive network brokers, Y primary users (i.e., primary
wireless terminals and primary base stations), and Z
secondary users (i.e., cognitive base stations and cognitive
access points) in a given cellular region. The number of
cloud facilities that these brokers can rent is M, while N is
with respect to the edge computing devices. In addition, L
idle smart devices are willing to be scheduled by the platform
as cognitive network access points or relays on the
premise of getting the corresponding return. The interaction
relationship in the game-based incentive mechanism
is shown in Figure 3.
Each broker (e.g., x) offers a real-time data-acquisition
) in terms of spectrum status to each primaprice
(e.g., px
d
ry user (e.g., y), where each primary user usually sells its
data to the highest bidder. So px
d
is positively correlated
with the maximum value of the spectrum status data prices
offered by the other brokers. Also, each broker is always
ready to collect the networking requirements from any
secondary user (e.g., z). Based on the purchased spectrum
status data and the collected networking requirements,
each broker designs its own cognitive computing solutions
based on the typical machine learning schemes [22], [23]
and then evaluates the required computing resources. Similar
to providing a data price to each primary user, each
broker provides the corresponding computing resource
prices (e.g., px
c
for a cloud facility and px
e
for an edge computing
device) to each cloud facility (e.g., m) and each edge
computing device (e.g., n), respectively, where the higher
bidder usually gets more computing resources.
In addition, each broker provides the corresponding
) to each idle
communication resource prices (e.g., px
s
smart device (e.g., l), which is positively correlated with
the maximum value of the communication resource prices
offered by the other brokers. The performance of each
broker's cognitive engines mainly depends on the quality
of the designed machine learning solutions and the
amount of the rented computing resources, while the
quality of the products (i.e., cognitive networks) produced
by each broker depends on not only the cognitive engines
but also the quality and the amount of the purchased data
(i.e., the spectrum status information of primary users
and the rental intention information of the owners of idle
smart devices).
Therefore, the brokers that better understand secondary
users and meet their demands will get more orders
for cognitive network construction, which allows them
to afford and purchase more high-quality data and rent
more high-quality computing resources. For a secondary
user z, each broker x offers a cognitive network construction
price (e.g., p ,xz
n
) to it, where the secondary user
z usually purchases from the lowest bidder on the premise
that the product meets the performance requirements.
Each broker's profit is affected by not only its
own costs for the purchased data and the rented computing
resources and its own product price but also the
other brokers' corresponding costs and prices. Therefore,
p ,xz
n
designed machine learning solutions as well as the
amount of the purchased data and the rented resources.
For any broker x, its utility functionUx
the following formula.
is formulated as
is positively correlated with the quality of the
Primary
User 1
Smart
Device 1
. . .
Smart
Device l
. . .
Smart
Device L
Data Prices
Broker 1
. . .. . .
Broker x
Cognitive Networks
Secondary
User 1
Spectrum Status Data
Broker X
Network Requirements
. . .. . .
Secondary
User z
Secondary
User Z
Figure 3. The example of an interaction relationship in a game-based incentive mechanism.
32 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE January 2023
. . .. . .
Primary
User y
Primary
User Y
Cloud
Facility 1
. . .
Cloud
Facility m
. . .
Cloud
Facility M
Edge
Device 1
. . .
Edge
Device n
. . .
Edge
Device N
Communication
Resources
Resource
Prices
Computing
Resources
Resource
Prices
IEEE Systems, Man and Cybernetics Magazine - January 2023

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