A. M. H. Alibraheemi et al.

: Spectrum Efficient Mode Selection and Resource Allocation Optimization

obtaining optimal sub-channels using sequential geomet- acceptable, and the system performance is near-optimal in
ric programming approaches. The optimization of network terms of EE. In [25], the authors examined the resource
throughput is accomplished by considering the constraints allocation issue in D2D communication. The power and spec-
imposed by the transmission power, rate, and reuse fac- trum allocation problem is effectively tackled by the use
tor. Simulations indicated considerable network throughput of an algorithm that incorporates the graphic approach and
improvement. In [19], the author examined the problem of Newton’s method. The optimality of the proposed approach
resource allocation to improve the network throughput in in terms of EE was shown through simulations. In [26],
D2D-based HetNets. A graph coloring algorithm is presented for D2D communication within HetNets, the authors exam-
to assign channel resources through the use of priority- ined channel allocation and power control issues. Hence,
weighted resources. The proposed method minimized power a joint mode selection, power control, and resource alloca-
consumption, and improved system throughput according to tion algorithm is proposed to improve the total sum rate.
simulations. Simulations proved that the proposed method successfully
In [20], the authors developed a one-to-many D2D com- improves system throughput while decreasing computational
munication system underlaying cellular networks to enhance complexity.
spectrum utilization and reduce interferences. The optimal On the other hand, artificial intelligence algorithms are
power is obtained using the particle swarm optimization often used with conventional optimization methods to address
method that allows the system to attain its maximum rate. various communication resource optimization challenges.
Simulations showed the reduced computational complexity Many recent studies have addressed the issue of intelligent
of the proposed method while substantially improving system resource management by applying RL to wireless commu-
throughput and performance. In [21], the authors investi- nication. In [27], resource management for V2V-based D2D
gated channel allocation and power allocation problems in communication is investigated. A resource allocation scheme
D2D communication. First, a comprehensive framework is based on a dynamic Q-learning algorithm is proposed to
developed to address the resource allocation problem. Then, enhance the network performance in underlay mode. Simula-
to maximize the total sum rate, an algorithmic solution is tions illustrated that the method converged rapidly to the opti-
designed for a two-stage joined channel and power allocation. mal policies and optimized the network performance. In [28],
The channel allocation method is represented as a many- the authors investigated the problem of mode selection in
to-many matching game, which offers a reliable solution D2D communication within dense NOMA networks. They
while simultaneously considering interference issues. The formulated the Received Signal Strength (RSS)-threshold
power allocation problem is addressed via DC programming, selection issue as a decentralized Markov decision process to
which is iteratively optimized using the Frank-Wolfe method. optimize the network performance. To attain the maximum
Simulations illustrated enhanced total sum rate, interference sum rate through continuous environmental interaction, they
control, user fairness, and computational complexity. applied a multi-agent RL method in which each Small Base
In [22], channel assignment and power management for Station (SBS) works as an agent and picks the best RSS
multicast D2D communications are addressed. A coali- threshold. Extensive simulations showed that the proposed
tional game model is used to allocate near-optimal channels method enhanced coverage and system sum rate.
under MAX-power and MIN-rate constraints. A Dinkel- In [29], the researchers proposed a joint power allocation
bach algorithm is presented to overcome the power control and relay selection to enhance EE in D2D communications.
problem. Simulation results demonstrated that the presented Lagrange dual decomposition and Dinkelbach algorithm are
semi-distributed approach achieves performance that is very used to address the power allocation issue. A RL technique
similar to that achieved with an ideal centralized approach. called Q-learning is used to address the relay selection issue.
In [23], a joint power management and channel allocation Furthermore, they investigated the complexity and signaling
technique for D2D communication is proposed to improve overhead of the suggested approach along with providing
network SE. A two-phase pricing-based noncooperative a theoretical analysis. Simulations showed that the sug-
resource allocation game is presented for CUs and D2D pairs gested method enhanced EE. In [30], the resource allocation
using a novel utility function and price values for interfer- problem of D2D-based Internet-of-Things (IoT) networks is
ence links. Simulations showed that the proposed approach investigated to optimize EE. Q-learning algorithm is pro-
improved the total sum rate. In [24], the authors examined posed to solve the joint subchannel assignment and power
the joint channel assignment and power allocation problem splitting. Furthermore, the Dinkelbach algorithm is presented
in D2D within cellular networks. A two-step decomposi- to solve the power control problem. Simulations demon-
tion method is suggested to tackle resource allocation and strated the superiority of the suggested method in terms of
power control subproblems. The resource allocation problem EE. In [31], the authors investigated D2D communications in
is addressed in a suboptimal manner, using principles of HetNets from the EE perspective. A strategy using Q-learning
matching theory. The power allocation is determined using a and adaptive ε-greedy is proposed to optimize the association
conventional optimization approach. Simulation results indi- of users with a BS or access point to decrease the total trans-
cated that the computing complexity of the technique is mitted power. The presented ε-greedy method can effectively

VOLUME 12, 2024 131219