A. M. H. Alibraheemi et al.
: Spectrum Efficient Mode Selection and Resource Allocation Optimization
sharing cellular communication resources [3]. The coexis- based on ML algorithms has been employed in wireless
tence of D2D technology has many advantages including networks [13].
enhanced network connectivity, improved spectrum utiliza- The organization of this paper is as follows: In Section II,
tion, and reduced latency [4]. D2D communication has a comprehensive literature and research review is provided.
substantial applications in different fields. D2D technology In Section III, the motivation and contribution are listed. In
plays a crucial role in maintaining reliable communication Section IV, the system model is thoroughly described.
during critical situations, hence protecting public safety. Fur- In section V, the problem of mode selection, spectrum alloca-
thermore, D2D technology improves the Internet of Things tion, and power allocation is formally formulated. A descrip-
(IoT) by enabling direct communication, supporting real-time tion of the modified Q-learning algorithm is presented in
data exchange, and developing a cognitive environment, con- Section VI. The effectiveness of the proposed approach is
sequently improving existing communication networks [5]. assessed in Section VII via simulations. Section VIII con-
Furthermore, using several terminals as relays to provide cludes the papers.
data for other devices via D2D communication links has
become an efficient way to increase connectivity and quality II. RELATED WORK
of service (QoS) [6]. Mode selection is essential aspect for Efficient management of network resources allows for cel-
maximizing network performance in D2D communication. lular and D2D links to operate at a high-performance level.
Direct mode guarantees a high data rate for devices within Recently, several mathematical frameworks and performance
proximity, while relay-assisted mode improves network con- criteria have been developed to investigate resource manage-
nectivity and coverage for devices that are located beyond ment of D2D communication within cellular networks.
the range of direct communication. Effective mode selection In [14], the researchers conducted an investigation into the
optimizes network reliability and energy efficiency, which issue of interference control in D2D communication within
are crucial for addressing the demanding requirements of cellular networks, specifically focusing on an uplink scenario.
D2D applications [7]. Even though D2D communication has They presented a matching theory-based resource optimiza-
enhanced cellular networks, many obstacles need to be tack- tion strategy to maximize network performance with minimal
led. In particular, conventional cellular systems are severely interference from D2D communications on cellular commu-
impacted by the interferences from D2D communication nications. The obtained numerical results indicate that the
because of the resource sharing of cellular user resources [8]. proposed strategy has the potential to significantly enhance
These issues have a significant impact on the ability of other network performance. In [15], the researchers investigated
cellular network users to communicate and may potentially the mode selection issue in D2D networks. The proposed
influence the future of D2D technology. Consequently, it is approach considered the procedures for mode selection,
crucial to identify effective methods to allocate resources in power allocation, and channel allocation simultaneously. The
D2D communication while keeping the degree of tempo- suggested approach allows both Cellular Users (CUs) and
ral complexity low. Moreover, A suitable power allocation D2D users to fully share the spectrum resources. The mode
method needs to be developed to improve Energy Efficiency selection method is presented according to the level of inter-
(EE) and maintain QoS. Conventional optimization methods ference. Finally, the Hungarian algorithm is used to allocate
experience several challenges in resource management as spectrum resources according to communication modes. Sim-
a result of the growing variety and complexity of mobile ulation results illustrated the improved system throughput and
network designs [9]. Channel allocation and power alloca- EE of the proposed approach.
tion commonly depend on centralized methods, like heuristic In [16], the authors examined the power allocation and
techniques. The goals of these techniques are to guarantee mode selection problems for a relay-assisted D2D network.
a fair distribution of resources, optimize spectral efficiency, The optimal power distribution issue for relay-assisted modes
and reduce interference [10]. Nevertheless, since modern was formulated and addressed. A novel relay selection
wireless networks are both dynamic and ultra-dense, they method is introduced to identify the relay user that maximizes
frequently experience substantial obstacles, such as latency the channel capacity. In three common D2D deployment
and computational complexity. Furthermore, static allocation situations, the simulation results showed the improved EE,
algorithms fail to respond optimally to real-time variations channel capacity, and success probability of the proposed
in the network. There are still open issues, such as how to method. In [17], the researchers performed a study on chan-
guarantee energy efficiency in ultra-dense HetNets and what nel allocation based on a many-to-one matching algorithm
scalable solutions are necessary for adapting to the network to maximize the overall EE of D2D communication. Then,
circumstances variation. the Dinkelbach algorithm and Karush–Kuhn–Tucker (KKT)
The ability of Machine Learning (ML) to predict future sit- conditions are presented to overcome the power allocation
uations and solve complicated issues with massive volumes of problem. The performance enhancements and convergence
data has led to its effective use in several fields [11]. For com- of the iterative method are demonstrated by the simula-
plex non-convex optimization issues in wireless networks, tion results. In [18], the authors developed an approach
ML methods based on artificial neural networks have shown for allocating resources in D2D communication to enhance
to be an effective solution [12]. Reinforcement learning (RL) the network’s performance. They proposed a method for
131218 VOLUME 12, 2024