Abubakar Makarf et al. (2020) ^[7] explored combining the Radio Information System (RIS) and Monte Carlo concept within a network to maximize the potential benefits. Two different RIS-based network models were investigated, and many performance measures connected with the CR secondary user were implemented. The obtained equations were validated using Monte Carlo simulations. In the presence of an RIS-enhanced main network, the results showed the influence of key system parameters and a clear improvement in the CR network. According to Zhaoyuan Shi et al., the massive MIMO which is, underlying the cognitive radio user selection schema, is aware of the QoS requirements of the channel requested (2019) ^[8]. There are two major ways in which a CR may be implemented: The Channel State Indicator (CSI) of any cross-network is inaccessible at the secondary base station (SBS), but the SBS has access to the CSI of the cross-channel channel state. Low-complexity algorithms for increasing users while using the least amount of power (IUMP) and methods for decreasing users while using the most amount of power (DUMP) were developed to solve user selection via power allocation. The intractable challenge was addressed using a deep reinforcement-learning-based method, enabling the SBS to accomplish effective and intelligent user selection. In simulations, these algorithms dramatically outperform the current user selection approaches. Theneural network was able to rapidly learn the best user selection strategy in an unknown dynamic environment with a high success rate and fast convergence, as the findings also revealed. Kok-Lim Alvin Yau et al. ^[9] focused on how CR and the Cognition Cycle have been integrated into 5G to deliver spectrum efficiency, energy efficiency, enhanced quality of service and experience, and cost-efficiency (2018). Open research opportunities and platform implementation were made accessible to inspire new research in this area. Gianfranco Nencioni et al. (2018) ^[10] discussed the fifth-generation (5G) of cellular networks, which is expected to represent a significant advancement in wireless technology. A variety of new wireless technologies will be implemented to better serve 5G’s wide set of requirements, including upgrades to the radio access network. It was demonstrated that the convergence of many communication technologies has been facilitated by embedding softwarizations such as Software-Defined Networking (SDN) and Network Functions Virtualization (NFV). Through network slicing, 5G networks may be constructed at low cost. By using an SDN/NFV architecture, 5G radio access and core networks will be able to deliver network services more efficiently, flexibly, and in a more scalable manner. The researchers also discussed software-defined 5G radio access and core networks, as well as a wide range of future research topics in orchestration and control. Johana Hernández et al. proposed cognitive radio management (2018) ^[11].

The most available route for the opportunistic transmission of secondary user data is chosen throughout the decision-making process as a result of the main user characterization-model’s efficiency. It was claimed that an approach based on deep learning and long short-term memory might lessen the forecasting error now present in future significant user estimates in the Global System for Mobile Communication (GSM) and WiFi frequency bands. When compared to alternative approaches, such as multi-layer perceptron neural networks, Bayesian networks, and adaptive neuro-fuzzy inference systems, the results indicate that a lengthy short-term memory may significantly enhance the estimates of channel utilization (ANFIS−Grid). The neural structure has input, forget, and output gates, which complicates its implementation in cognitive radio networks (CRNs) based on core network topologies, despite the fact that a long short-term memory fared better at time series forecasting. Aaron Yi Ding et al. developed the criteria for assessing the restrictions of 5G-driven applications (2018) ^[12]. The usual hurdles and needs for various application domains were studied using 5G networks as a basis. The major objective was to create a network architecture that could adapt to changing traffic patterns, while also supporting diverse technologies, such as edge computing, blockchain-based distributed ledgers, software-defined networking, and virtualization. Researchers underlined the need to perform 5G application pilots to better understand how 5G networks are deployed and utilized in different vertical industries. Xingjian Li et al. investigated the issue of spectrum sharing in a cognitive radio system with a main and secondary user (2018) ^[13]. Secondary users are at odds with prime users. In particular, it was assumed that the primary user would alter its transmitted power in accordance with a pre-defined power management strategy. The secondary user has no idea what the main user’s transmission power and power control method are. For the secondary user, a power control system based on learning was developed to share the same spectrum with the main. To assist the secondary user, a network of sensor nodes was strategically placed across the wireless network to collect data on the received signal strength. The secondary user’s transmission power may be automatically adjusted using a deep reinforcement learning algorithm. This may be performed after a few rounds of interaction with the principal user. The results showed that secondary users might effectively connect with main users to achieve the desired state from any beginning situation in a few steps. When Maria Massaro et al. (2017) ^[14] examined the Licensed Shared Access (LSA) and Shared Access Spectrum (SAS) regimes in the EU, they discovered significant disparities between the LSA and SAS regimes. To acquire information on the technical and regulatory components of current and forthcoming spectrum-sharing regimes, policy documents, research publications, position papers, and analytical studies were studied. The LSA regime is notable for providing mobile operators with the regulatory certainty they need to invest in 5G, while also granting them access to an additional spectrum below 6 GHz. Other spectrum-sharing regimes will not safeguard cell operators from hazardous interference or ensure trustworthy Quality of Service (QoS) while utilizing sub-6 GHz airwaves. Due to its lower level of technical complexity, the LSA regime can be deployed more rapidly and with less effort when the two techniques are compared. However, as technology progresses, the LSA regime is projected to be surpassed in the long run by the SAS regime. More persons may share the same frequency channels under the SAS setup. A cognitive WLAN overlay over an OFDMA TDD main network was tested for saturation throughput by Parisa Rahimzadeh et al. (2017) ^[15], e.g., using LTE or WiMAX. The successful node delivers its data packet in the main network’s downlink and uplink subframes that have empty resource blocks (RBs). Unlike the OFDMA structure and time-scheduled resources in the primary network, the opportunity length in the secondary network does not follow a straightforward exponential on–off pattern. There is a mathematical model for the dynamic behavior of secondary node opportunities and contentions that incorporates a discrete-time Markov chain and two connected open multi-class queueing networks (QNs). As a random number is created when data are downloaded and uploaded, the research includes the random packet transmission time on WLANs, the dependency on the number of empty RBs in the subsequent frames, and aspects of the 802.11 MAC protocol. Multiple resource allotments were inserted into the main network for the purpose of conducting the analysis. Researchers were able to demonstrate the correctness of the method via simulations in a variety of different situations. Cheng Wu et al. (2016) ^[16] developed a multi-agent reinforcement learning-based spectrum management approach. For efficient spectrum and transmit power allocation, the approach employed value functions to evaluate the advantages of different transmission features, maximizing long-term return. Using a variety of learning factors, students were exposed to a variety of real-world circumstances, and their communication skills were evaluated. A Kanerva-based function approximation was utilized to enhance the management of large CRNs, and to examine the impact on communication performance in these networks. The proposed reinforcement learning-based spectrum management in a cognitive radio ad hoc network is shown to considerably minimize interference to licensed users, while preserving a high probability of successful transmissions in this network. The secondary users’ average sum rate was increased by Yang Yang et al. (2017) ^[17] by determining the best way to access and regulate their power in multiple bands (ASR). Researchers represented the random distributions of PUs and SUs using Poisson point processes (PPPs) based on stochastic geometry, from which researchers were able to compute the closed-form outage probability and estimate the ASR of SUs. On a number of bands, the ASR maximization problem included an outage probability. The optimal density of SUs with a given power was determined using closed-form convex optimization, and the optimal SU power was calculated and ASR convexity was checked. These findings prompted the creation of a spectrum access and power management strategy aimed at optimizing the ASR of SUs over several bands. The results of the simulations reveal that PUs and network interference limit SUs’ density and power, and that the proposed approach can achieve the maximum ASR for the SUs.