1st.
International Conference of Information Technology to enhance E-learning and other Application,
( IT-ELA 2020 ), Baghdad College of Economic Sciences University, Baghdad, Iraq
Distributed Cooperative and Noncooperative Joint
Power and Beamforming Adaptation Game for
MIMO Sensor Network
Ali Kamil Khairullah
Yildiz technical university Yildiz Technical University
Istanbul, Turkey
Ministry of Communication
Baghdad, Iraq
engineerali22@gmail.com
Abstract—Distributed Joint beamforming and power
adaptation algorithms are of interest for MIMO ad hoc
networks. Cooperative and non-cooperative games have been
presented to decrease interference (mutual) at each sensor node,
under receiver signal-to-interference and noise (SINR)
constraints. In reduced feedback algorithms, the optimum
transmitter node beamformer is selected from a predefined
codebook. This paper introduces a cooperative optimal
beamformer selection algorithm to minimize the total power
consumption for cluster-based network topology under
different minimum SINR constraints. The algorithm reduces
overhead incurred by 48% and 10% while increasing the
convergence rate to the steady-state allocation by 17% and 9%
for the cooperative and non-cooperative beamformer selection
games, respectively. Simulation results verify the proposed
theoretical analysis, and demonstrate the performance of the
Enhanced Co-Operative Power Minimization Algorithm
(ECOPMA), Reinforcement Learning based Power allocation
and Beamformer Algorithm (RLPBA) for the non-cooperative
game [1], with state of the art methods and centralized (optimal)
solutions as a fair benchmark.
Keywords—MIMO Ad hoc network, Beamforming, Overhead
Rate, SINR, Potential Game, Reinforcement Learning.
I. INTRODUCTION
The emergence of massive MIMO has fueled new
research into the design of future generation wireless
communications. Multi User-MIMO (MU-MIMO) supports
the implementation and expansion of Massive MIMO
(MIMO extension technique) system technologies [2]. Ad
hoc networks are formed by nodes of limited battery power,
under limited bandwidth channels across which nodes form
unstable capacity links due to random changes in topology
[3]. Beamforming algorithms in such networks need to be fast
converging, low overhead and computationally simple while
conserving battery life by reducing transmission power.
Quantized transmit beamforming (codebook) is used to
reduce the overhead incurred from multi-user beamforming
in a MIMO wireless network, which limits the messaging
overhead of the protocol to the transmission of the
beamforming index [4]. MIMO communication technology
efficiency largely depends on the intelligent adaptation of
transmitting and receiving parameters of the wireless node
like the selection of beamformer, transmit power,
modulation, etc. The mechanism design of the beamformer
978-1-7281-8233-9/20/$31.00 ©2020 IEEE
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on January 10,2021 at 19:27:45 UTC from IEEE Xplore. Restrictions apply.
Ufuk Tureli Didem Kivanc
Istanbul Okan university
Istanbul, Turkey Istanbul, Turkey
utureli@yildiz.edu.tr didem.kinvanc@okan.edu.tr
selection and power adaptation can significantly improve the
performance of adaptive wireless communication systems.
MIMO beamforming in wireless communication systems has
been designed for several wireless networks such as point-to-
point systems, mobile cellular networks, and MIMO ad hoc
technology [5].
The interference between wireless nodes is often modelled
using game theory [6]. Game theory has enabled optimal
efficiency and convergence proofs of the algorithms for
distributed power control, beamforming, and other
algorithms [7]. Cooperative and non-cooperative distributed
game methods for iterative joint channel allocation and power
control, which are selected from the "discrete" strategy space
profiles, are presented under wireless mesh networks
scenario [8]. The area of game theory presents some
advantages as one of the tools of the analysis of the
distributed protocols and algorithms for the ad hoc networks
[9].
On the other hand, Wireless Body Area Network (WBAN)
is a special purpose sensor network, short-range wireless
communications in the nearness of the human body or inside.
WBAN is an emerging solution to serve local and remote
healthcare-related facilities needed. Sensors usually utilize
the wireless medium for transferring data [10, 11], the
interruption caused by the human body or interference with
other coexist sensors resulting in certain radio propagation,
which should be considered in designing an algorithm for the
WBAN systems [12, 13]. Since most of the equipment in
WBAN is utilizing the wireless medium, free mobility. Such
devices are small and carry power source inside. Hence,
power is always limited, so an efficient algorithm to reduce
power consumption (interference incurred) is required.
The distributed cooperative and a non-cooperative based
learning schemes for joint power control, and transmit
beamforming selection in multiple antennae wireless multi-
user interference environment are presented in [9]. Directly
below the total network power reduction standard, a joint
iterative method is studied to minimize mutual interference
present at each sensor node with a constant received SINR at
each receiver node [18-20].
In this paper, the performance of ECOPMA in the
cooperative game and the RLPBA for the non-cooperative
one is further investigated, by considering the network
performance under different SINR constraints with minimum
system overhead, convergence computation and interference
127
 1st.International Conference of Information Technology to enhance E-learning and other Application,
( IT-ELA 2020 ), Baghdad College of Economic Sciences University, Baghdad, Iraq

burden, as compared to state-of-art algorithms. The novel denoted by ୯ ൌ ሼ‫ݐ‬௤ଵ ǡ ‫ݐ‬௤ଶ ǡ ǥ ǥ ǡ ‫ݐ‬௤ᙓ ሽ with cardinalityȺ. It is
contributions of this paper are the analysis of the ECOPMA assumed that such a sensor node can easily choose from Ⱥ
and RLPBA algorithms [1] and their comparison to COPMA transmit beamforming codebook (vectors). Let ‫ݐ‬௤ ‫ܵ א‬௤ be
and RMSG in [9] for various SINR constraints. Both the total the selected transmit beamformer for the q-th
transmission power and the communication overhead transmitting/receiving sensor node pair. The beamformer
incurred by the algorithms are compared. In simulations, codebook which is assumed to be initialized prior to the
ECOPMA and RLPBA algorithms are found to exhibit lower beginning of the algorithm. The optimization problem has
total power transmission while meeting the SINR been defined in [9]:
requirements with lower communication overhead.
The remaining issues of this work are arranged as follows: ே
In section II, the system model and parameters are presented. ‹ ෍ ‫ܧ‬௤ ൫‫ݐ‬௤ ǡ ‫ݓ‬௤ ǡ ‫ିܧ‬௤ ൯
The optimization and game approach solution description ሺ௪೜ ǡ௧೜ ሻ‫א‬஼ ಲ (3)
(the cooperative and non-cooperative game theory) with ா೘೔೙ ᦪா೜ ᦪ ா೘ೌೣ ௤ୀଵ
ష
overhead rates are studied in Section III. The system
performance evaluation has been presented in Section IV. Subject to: ߬௤ ൒ ߛ଴ (4)
Conclusion of the obtained results has been presented in ଶ ଶ
ฮ‫ݓ‬௤ ฮ ൌ ฮ‫ݐ‬௤ ฮ ൌ ͳǡ (5)
section V.
‫ܧ‬௠௜௡ ᦪ‫ܧ‬௤ ᦪ‫ܧ‬௠௔௫
ି (6)
II. MODEL OF FRAMEWORK AND PARAMETERS where ߛ଴ is the threshold SINR requirement and 
The network is comprised of N sensor node pairs. Each ‫ܧ‬௠௜௡ and ‫ܧ‬௠௔௫ are the maximum and minimum transmit
sensor ‫ א ݍ‬ሼͳǡʹǡ ǥ ǥ ǡ ܰሽ has a single transmitter/receiver powers achievable by nodes in the system. Let ܶ ൌ
஺ ஺
unit, each is equipped with A antennas. The complex symbol ൣ‫ݐ‬ଵǡ ‫ݐ‬ଶǡǥ ‫ݐ‬ே ൧ and Ƞ ൌ ൣ‫ܧ‬ଵǡ ‫ܧ‬ଶǡǥ ‫ܧ‬ே ൧ where T is the selected
stream ܾ௤  ‫  א‬has a constant average power of 1, transmit beamformer and Ε is the transmit power vectors for
ଶ N sensor nodes. For A × A matrix channels, the interference
‫ ܧ‬ቄหܾ௤ ห ቅ ൌ ͳ . Each sensor has a unit-norm (receive and
plus noise covariance at the q-th receiving sensor node can be
transmit) beamformer pair ሺ‫ݓ‬௤ ǡ ‫ݐ‬௤ ሻ ‫ ܥ א‬஺ . At any point in expressed as:
time, each node is assumed to function as either a transmitter
or a receiver; each pair of communicating nodes is assigned
a link number. At the q-th receiving node, the received signal ܴ௤ ൫ܶି௤ ǡ Ƞି௤ ൯ ൌ ෍ ‫ܧ‬௝ ‫ܪ‬୯ǡ୨ǡ ‫ݐ‬௝ ‫ݐ‬௝ு ‫ܪ‬௤ǡ௝
ு
൅  ߪ ଶ ‫ܫ‬ (7)
vector ‫ݎ‬௤  ‫ ܥ  א‬஺ is given by: ௝ஷ௤

ே where ܶି௤ and Ƞି௤ are the transmit beamform and powers of
nodes other than q. The optimal normalized receiver
‫ݎ‬௤ ൌ ට୯ ‫ܪ‬௤ǡ௤ ‫ݐ‬௤ ‫ݏ‬௤ ൅ ෎ ට୨ ‫ܪ‬୯ǡ୨ǡ ‫ݐ‬௝ ‫ݏ‬௝ ൅ ୯ (1) beamformer, given the known channel between the target
௝ୀଵ receiver q and all transmitting nodes, is [9]:
௝ஷ௤
‫ݓ‬
ෝ௤ (8)
where ‫ܪ‬௤ǡ௝ represents the A × A MIMO channel between the ‫ݓ‬௤ ൌ 
ฮ‫ݓ‬
ෝ௤ ฮ
transmitter of the Œ-th link and the receiver of the link in the
ad hoc network. ୨ is the transmit power of the Œ-th sensor where
node, and ݊௤ ‫ ܥ א‬஺ is the additive white Gaussian noise
covariance matrix ߪ ଶ I A×A. The quasi-static MIMO channel is ෝ௤ ൌ ܴ௤ିଵ ‫ܪ‬௤ǡ௤
‫ݓ‬ (9)
assumed in this work, and each receiver has an identical noise
variance ߪ ଶ . ‫ݐ‬௤  is the selected transmit beamformer for the
The proposed distributed algorithms achieve the target
“-th link. Since the first term in "(1)" contains the desired
SINR threshold by regulating each individual transmit power
signal, the second and third expressions contain interference
of the nodes after selecting the beamformer, which is closest
and noise for the signal of interest respectively, the SINR at
to the optimal beamformer above it.
receiver q is given by:

ଶ III. OPTIMIZATION METHOD AND GAME THEORY ANALYSIS

‫ܧ‬௤ ห‫ݓ‬௤ு ‫ܪ‬௤ǡ௤ ‫ݐ‬௤ ห
߬௤ ൌ
σ௝ஷ௤ ‫ܧ‬௝ ȁ ‫ݓ‬௤ு ‫ܪ‬௤ǡ௝ ‫ݐ‬௝ ȁଶ ൅ ߪ ଶ
In this scenario (limited feedback beamforming
communication system), the main contribution of the
receiving node that, it selects a suitable transmit beamformer
from the predefined codebook, known for both sensor node
pairs and feeds back the label of the chosen one to the
transmitter. The set of available codebook beamformers for
the q-th transmitting and receiving sensor node pair is
(2)
The optimization problem has been formulated as a game
with normal form style, and such a game can be practically
realized by the triplet G = [N, C, ሼܷ௤ ሽே
௤ୀଵ ] where G is a game,
ܰ ൌ ሼͳǡ ʹǡ ǥ ǥ ǡ ܰሽ is the finite number set of players in such
a game, ‫ ܥ‬ൌ  ‫ܥ‬ଵ  ൈ  ‫ܥ‬ଶ ൈǡ ǥ ǥ ǡ ‫ܥ‬ே represents the available
actions group for all the Nodes (players), and ሼܷ௤ ሽே ௤ୀଵ : C ֬
R is the set of pay off (utility function) by which the sensor
nodes can be linked with their actions (strategies) in the
game. Actions ܿ௤ ‫ܥ א‬௤ for a node q are the transmit powers

128

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on January 10,2021 at 19:27:45 UTC from IEEE Xplore. Restrictions apply.
 1st.International Conference of Information Technology to enhance E-learning and other Application,
( IT-ELA 2020 ), Baghdad College of Economic Sciences University, Baghdad, Iraq
‫ܧ‬௤  ‫  א‬ሾ‫ܧ‬௠௜௡ ǡ ‫ܧ‬௠௔௫ ሿ and the transmit beamformer selections
‫ݐ‬௤  ‫  א‬௤ .
The players usually select their acceptable performance
(actions) to maximize their achieved payoff functions (for
decreasing system power consumption and overhead incurred
by cooperative and non-cooperative manners). Any game
formulation which is shown to have a Nash Equilibrium (NE)
is guaranteed to converge to a stable state.
A. Centralized Method
The centralized procedure helps to minimize the total
network transmit power of all system transmitting antennas
A as shown in the following equation:
ே solution, the ‫ݍ‬௧௛ node pair concentrates on their own power
ሺȣ‫ כ‬ǡ ‫ כ ܧ‬ሻ ൌ ƒ”‰ ‹ ෍ ‫ܧ‬௤ ሺȣǡ ‫ିܧ‬௤ ሻ
஀ǡ୉
௤ୀଵ
where ȣ‫ כ‬ൌ ሺ‫ݐ‬ଵ‫ כ‬ǡ ‫ݐ‬ଶ
‫כ‬
ǡ ǥ ǥ ǡ ‫ݐ‬ே‫ כ‬ሻ and  ‫ כ‬ൌ ሺ‫ܧ‬ଵ‫ כ‬ǡ ‫ܧ‬ଶ
‫כ‬
ǡ ǥ ǥ ǡ ‫ܧ‬ே‫ כ‬ሻ
are defined as the best (optimal) node transmit beamformer
and its power vectors, respectively. The transmit node power
‫ܧ‬௤ of the q-th sensor, node pair can be defined as:
‫ݕ‬଴
‫ܧ‬௤ ൫ȣǡ ‫ିܧ‬௤ ൯ ൌ
‫ݐ‬௤ு ‫ܪ‬௤ǡ௤
ு
ܴ௤ିଵ ‫ܪ‬௤ǡ௤ ‫ݐ‬௤
Here ܴ௤ depends on the values of ሺȣି௤ ǡ ‫ିܧ‬௤ ሻ as presented in
"(7)". To determine the transmit node power ‫ܧ‬௤ , the
centralized operator can estimate the whole sensor network
power for Ⱥே available transmit beamforming index
collections ponderous in small-scale network topology (in our
case, the wireless MIMO network). For that, we will introduce
decentralized algorithms.
B. Enhanced Cooperative Power Minimization Algorithm
(ECOPMA):
According to clustering analysis [1, 14], let and ୧୑ǡ୫
represent the node transmission adjustment power and
transmit beamformer selected parameters (vectors) for ݉௧௛
sensors pair, in the ‫ܯ‬௧௛ cluster, respectively. Thus, the main
payoff for each node pair in ECOPMA is:
ெǡିெ
௜ ሺ‫ݐ‬ǡ ‫ܧ‬ሻ ൌ െ ෍ ‫ܧ‬୧୑ǡ୫ ൫–ǡ ‫ିܧ‬୧
୑ǡ୫
൯
௠ୀଵǡ௜ୀଵ
which is also equal to the global (utility) function [8] i.e.
ெǡିெ
௜ ሺ‫ݐ‬ǡ ‫ܧ‬ሻ ൌ ୬ୣ୲ ሺ‫ݐ‬ǡ ‫ܧ‬ሻ  ൌ െ ෍ ‫ܧ‬୧୑ǡ୫ ൫–ǡ ‫ିܧ‬୧
୑ǡ୫
൯
௠ୀଵǡ௜ୀଵ
Zeng et al. [10] show power-based non-orthogonal multiple
access systems. In this system users share the same
frequency, time and code and user separation are
accomplished through beamforming at the receiver achieve
higher throughput with better fairness than orthogonal
multiple access with minimum overhead rate and equal power
consumption. Similarly in ECOPMA the ݉௧௛ sensor node
pair in each of the M clusters can share common transmit
beamforming index, similar to the scheme described by Zeng
et al. [15]. Let ୯୑ǡ୫ and ‫ݐ‬௤ெǡ௠ be transmission power vectors
and transmit beamformer selection vectors for ݉௧௛ node pair
in cluster ‫ܯ‬. In a cooperative game, all nodes minimize total
network power. Which is an identical interest game [1] [9].
Given this, it will also be a special type of game known as a
potential game. There is at least one pure Nash equilibrium
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on January 10,2021 at 19:27:45 UTC from IEEE Xplore. Restrictions apply.
(NE) strategy profile produced by all the identical interest
games, and this profile is achieved when all players play an
action (decision) profile that enhances their utility payoff
୧ ሺܶǡ ‫ܧ‬ሻ [16].
C. Reinforcement Learning based Power allocation and
Beamformer Algorithm (RLPBA):
In [1], the second proposed technique is based on non-
cooperative game theory-based approach for beamforming
and adaptive power allocation in multi-user ad hoc MIMO
networks, along with using the reinforcement-learning
algorithm. For non-cooperative nodes, the utility function has
been employed in this scenario. In contrast to the cooperative
(10) minimizations instead of the entire power of the network. For
every node, the utility function is based on the selections of
transmitting beamformers and its power, and other node's
decisions for beamformers and transmit power during
recognizing incurred interference.
According to the selection of the node transmit beamformer
‫ݐ‬௤ ሺ݇ሻ , the current (new) node transmit power ‫ܧ‬௤ is
(11)
determined by applying "(11)" on the selected ‫ݐ‬௤ ሺ݇ሻ, until:
௨௣ௗ௔௧௘ௗ ௖௨௥௥௘௡௧
ߛ௢
‫ܧ‬௤ሺ௞ାଵሻ ൌ ‫ܧ‬௤ሺ௞ሻ ൈቆ ቇ (14)
߬௤
or a predefined number of iterations is reached. The utility
functions of a greedy user for choosing the specific
beamformer and the expected power consumption at the time
can be expressed as:
௤ ሺ‫ݐ‬௤ ǡ ‫ିݐ‬௤ ሺ݇ሻሻ ൌ Ž‘‰ሺ‫ݐ‬௤ு ‫ܪ‬௤ǡ௤
ு
ܴ௤ିଵ ‫ܪ‬௤ǡ௤ ‫ݐ‬௤ ሻǤ (15)
In Regret Matching (RM), the players minimize the regret
௧ҧ
‫ݐ‬௜ெǡ௠ ܴ௤೜ of not playing a certain action:
௧ҧ ͳ ௞ିଵ
ܴ௤೜ ሺ݇ሻ ൌ  ෍ ሺܷ௤ ሺ ‫ݐ‬௤ҧ ǡ ‫ିݐ‬௤ ሺ݅ሻሻ െ  ܷ௤ ሺ‫ݐ‬௤ ሺ݅ሻሻ (16)
݇െͳ ௜ୀଵ
The player is assumed to be able to evaluate the utility under
different actions at time k. The action is chosen according to
(12) the probability distribution found by normalizing a regret
vector.
௧ҧ
௧ҧ ሾܴ௤೜ ሺ݇ሻሿା
‫ݐ‬௤೜ ሺ݇ሻ ൌ ܲ‫ݎ‬൫‫ݐ‬௤ ሺ݇ሻ ൌ  ‫ݐ‬௤ҧ ൯ ൌ  ௧ҧ
(17)
σ௧೜ҧ ‫߲א‬ሾܴ௤೜ ሺ݇ሻሿା
(13) In Reinforcement Learning (RL), the player pursues a certain
behavioural rule if an action enhances payoff function in
point k is guaranteed with higher probabilities in the game
with k+1, further enhancement of utility can be achieved.
௞
݇
Ꮝ௜ ൌ  ‫݅ݎ‬൅ͳ ൅  ෍ ߚ ‫݅ݎ‬൅݇൅ͳ (18)
௜ୀଵ
where Ꮝ௜ represents the reinforcements that have been
received over 200 iterations, ‫ݎ‬௜ାଵ is the current reinforcement
signal with a discount factor denoted by symbol ߚ . The
achieved utility is bounded, and each player can periodically
supervise their payoff.
In RM [9], the players must be able to observe all actions of
other players that lead to large overhead, and this is where RL
has an important advantage, it only needs observation of the
utility by applying algebraic computations (for more details
see [6], [17]).
129
 1st.International Conference of Information Technology to enhance E-learning and other Application,
( IT-ELA 2020 ), Baghdad College of Economic Sciences University, Baghdad, Iraq

D. Overhead Rate:
The communication overhead required by the algorithms
depends on the time required for the beamforming and power
adaptation by using different methods. The communication
overhead rate is computed from the Matlab simulation code
as:

ܴ ൌ ሺܶ ௖ െ  ܵ ௖ ሻǡ (19)

where R is the communication overhead for the current

method, ܶ ௖ is the time at which the solution is generated, and
ܵ ௖ is the time at which the solution generation starts to solve
the problem. Accordingly, we have measured the R for each
iteration and each method.

IV. SIMULATION RESULTS Fig. 1. Total transmit powers with N =5 users [1].
In this section, we present the total network performance
results of the centralized (optimum) solution, COPMA, and
RMSG, ECOPMA and RLPBA. Three different values have
represented the target SINR, ‫ݕ‬଴ = 8, 10 [1] and, 12 dB for all
the node pairs. The total iterations number is κ = 200 for a
small-scale network to search for all possible profile spaces.
Table I in reference [1] illustrates the simulation parameters
for small-scale (5-Node Pairs) wireless ad hoc MIMO
topology when ‫ݕ‬଴ = 10.
The performance of the total power consumption is shown
in Figs. 1, 2 and 3 for target signal to noise ratios of 10dB,
12dB and 8dB respectively. It has been observed that
ECOPMA's performance settles to an optimum search result
as indicated by the COPMA algorithm for the cooperative
solution.
Fast convergence speed over 200 iterations can Fig. 2. Total transmit consumption powers versus iteration number for 5-
significantly decrease the overhead system rate, as well as the node pairs, SINR =12 dB.
whole, transmit power consumption of the ad hoc wireless
system. ECOPMA performance is enhanced because of the
clustering and the simple binary codebook [1]. Likewise, a
non-cooperative distributed game with the learning-based
algorithm is also examined for different SINR constraints.
The implementation of RLPBA technique shows a significant
reduction in terms of power consumption, convergence time,
and overhead rate for optimum and efficient allocation
solutions for sensor node pairs, over the regret-based learning
approach in RMSG as shown in Table I in this paper.
Figure 4 compares the overhead required for the algorithms,
measured through runtime. The proposed model RLPBA
shows a reduction in the overhead rate over the iterations to
be 200, which is significant when compared to the existing
methods, this reduction in communication overhead
compared to the previous RMSG technique is a result of the
reinforcement learning approach [6],[ 17]. Fig. 3. Total transmit consumption powers versus iteration number for 5-
node pairs, SINR =8 dB.

Cooperative game techniques have introduced significant

overhead in comparison with the non-cooperative game
procedures since the beamformer updates require less
overhead in non-cooperative game techniques. The presented
cooperative and non-cooperative solutions achieved much
better computational performance compared to the previous
methods.

130

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on January 10,2021 at 19:27:45 UTC from IEEE Xplore. Restrictions apply.
 1st.International Conference of Information Technology to enhance E-learning and other Application,
( IT-ELA 2020 ), Baghdad College of Economic Sciences University, Baghdad, Iraq
Fig. 4. Communication overhead analysis for N=5.
The global optimum solution obtained by enumerating all
feasible strategies ͳ͸ே profiles in this case, when the
maximum number of iterations is k = 200.
TABLE I. SIMULATION RESULT VALUES
Algorithm Total no. Convergence Total Power (mW) / NE
Types iteration
SINR SINR
8 dB 10 dB
Centralized 200 48/1-200 67/1-120 87/1-200
COPMA 200 55/53 78/95 125/198
ECOPMA 200 52/45 74/87 117/191
RMSG 200 112/171 200/120
RLBPA 200 95/166 170/110
The presented methods (ECOPMA and RLPBA) have
achieved a much better performance in terms of reduced
power consumption, compared to the previous algorithms
(COMPA and RMSG), with higher speed of convergence
(NE).
The simulation results of the proposed algorithms and
pervious work have shown to achieve the appropriate solution
based on the power consumption, convergence rate and,
overhead reduction (see Fig. 4 for overhead rate and Table I,
respectively).
V. CONCLUSION
This paper has investigated the performance of two novel
distributed cluster-based algorithms based on game-theory to
enhance the performance of wireless sensor networks. The
enhancement focus on joint transmit power and transmit
beamforming, with minimum total power and overhead
incurred under different SINR constraints. Results have
indicated that combining cooperative (ECOPMA) according
to cluster-based approach, binary codebook, and non-
cooperative (RLPBA) solution, the reinforcement learning-
based algorithm, can achieve good performance even with the
SINR value changes. The simulation outputs reflect the fast
convergence rate of our algorithms and their achievement
based on the whole power reduction, convergence, and
overhead rate in the system. The proposed techniques can
significantly improve performance as compared to the
previous methods. Future work includes investigating the
network performance for different propagation channels for
large network topologies (more than 10 node pairs) and by
using other game theoretical approaches.
Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on January 10,2021 at 19:27:45 UTC from IEEE Xplore. Restrictions apply.
REFERENCES
[1] A. K. Khiarullah, U. Tureli, and D. Kivanc. "Optimized Intelligent
Design for Smart Systems Hybrid Beamforming and Power Adaptation
Algorithms for Sensor Networks Decision-Making Approach."
International Journal of Computational Intelligence Systems Vol. 12,
no. 2, pp. 1436 – 1445, December 2019.
[2] A. J. Paulraj, D. A. Gore, R. U. Nabar, and H. Bolcskei, "An overview
of MIMO communications-a key to gigabit wireless," Proc. IEEE, vol.
92, no. 2, pp. 198–218, 2004.
[3] M. Devi and N. S. Gill, "Mobile Ad Hoc Networks and Routing
Protocols in IoT Enabled Smart Environment: A Review," Journal of
Engineering and Applied Sciences, vol. 14, pp. 802–811, 2019.
[4] D. J. Love, R. W. Heath, and T. Strohmer, "Grassmannian
beamforming for multiple-input multiple-output wireless systems,"
IEEE Trans. Inf. Theory, vol. 49, no. 10, pp. 2735–2747, 2003.
[5] D. Gesbert, S. Hanly, H. Huang, S. S. Shitz, O. Simeone, and W. Yu,
"Multi-cell MIMO cooperative networks: A new look at interference,"
IEEE J. Sel. Areas Commun., vol. 28, no. 9, pp. 1380–1408, 2010.
[6] W. Saad, M. Bennis, Game Theory for Future Wireless Networks:
Challenges and Opportunities, ICC, London, UK, 2015-Tutorial.
[7] M. Schubert and H. Boche, "Solution of the multi-user downlink
beamforming problem with individual SINR constraints," IEEE Trans.
Veh. Technol., vol. 53, no. 1, pp. 18–28, 2004.
[8] Y. Song, C. Zhang, and Y. Fang, "Joint channel and power allocation
in wireless mesh networks: A game-theoretical perspective," IEEE J.
SINR Sel. Areas Commun., vol. 26, no. 7, pp. 1149–1159, 2008.
12 dB [9] E. Zeydan, D. Kivanc, U. Tureli, and C. Comaniciu. "Joint iterative
beamforming and power adaptation for MIMO ad hoc networks."
EURASIP Journal on Wireless Communications and Networking
2011, no. 1 (2011): 79.
399/71 [10] A. Alkhayyat, S. f. Jawad and S. B. Sadkhan, "Cooperative
338/62 Communication based: Efficient Power Allocation for Wireless Body
Area Networks," 2019 1st AL-Noor International Conference for
Science and Technology (NICST), pp. 106-111, 2019.
[11] D. Hammood, A. Alkhayyat, A. Rahim, Hasliza and Ahmad,
R.Badlishah. Minimum Power Allocation Cooperative
Communication based on Health-Care WBAN. IOP Conference
Series: Materials Science and Engineering. 745. 012051.
10.1088/1757-899X/745/1/012051, 2020.
[12] D.A.Hammood, H.A. Rahim, A. Alkhayyat, B. Ahmad, R.,
Abdulmalek, M., Ali, S.Y. and Mahmoud, I.S, An energy-efficient
optimization-based scheme for low power devices in wireless body
area networks. Journal of Computational and Theoretical Nanoscience,
16(7), pp.2934-2940, 2019.
[13] H. Al-Mishmish, A. Alkhayyat, H. A. Rahim, D. A. Hammood, R. B.
Ahmad, and Q. H. Abbasi, "Critical data-based incremental
cooperative communication for wireless body area network," Sensors,
vol. 18, no. 11, p. 3661, 2018.
[14] F. M. S. Shah, H. Ilhan and U. Tureli, "CB-MAC: a novel cluster-based
MAC protocol for VANETs," in IET Intelligent transport systems, vol.
13, no. 4, pp. 587-595, 2019.
[15] M. Zeng, A. Yadav, O. A. Dobre, and H. V. Poor, "A fair individual
rate comparison between MIMO-NOMA and MIMO-OMA," in 2017
IEEE Globecom Workshops (GC Wkshps), 2017, pp. 1–5.
[16] M. Felegyhazi and J.-P. Hubaux, "Game theory in wireless networks:
A tutorial," 2006.
[17] R. S. Sutton and A. G. Barto "Reinforcement Learning: An
Introduction", 2014.
[18] S. B Sadkhan, “Performance Evaluation of Concatenated Codes
applied in Wireless Channels”, 2019 1st AL-Noor International
Conference for Science and Technology (NICST), PP: 89-93
[19] S. B Sadkhan, Dh. M Reda, “Best Strategies of Choosing Crypto-
System’s Key for Cryptographer and Attacker Based on Game
Theory”, 2018 Al-Mansour International Conference on New Trends
in Computing, Communication, and Information Technology
(NTCCIT), PP: 1-6
[20] S. B. Sadkhan, S. A. Mohammed, “Design of Wireless Network Based
on NS2”, Journal of Global Research in Computer Science 2012, Vol.
3, No. 12
131