The ever-growing demand for sustainable, efficient, and fair allocation in the next generation of wireless network applications is a serious challenge, especially in the context of high-speed communication networks that operate on Terahertz frequencies. This research work presents a novel approach to enhance queue management in 6G networks by integrating reinforcement learning, specifically Deep Q-Networks (DQN). We introduce an intelligent 6G Retrial Queueing System (RQS) that dynamically adjusts to varying traffic conditions, minimizes delays, reduces energy consumption, and guarantees equitable access to network resources. The system’s performance is examined under extensive simulations, taking into account multiple arrival rates, queue sizes, and reward scaling factors. The results show that the integration of RL in the 6G-RQS model successfully enhances queue management while maintaining the high performance of the system, and this is by increasing the number of mobile terminals served, even under different and higher traffic demands. Furthermore, singular value decomposition analysis reveals clusters and structured patterns, indicating the effective learning process and adaptation performed by the agent. Our research findings demonstrate that RL-based queue management is a promising solution for overcoming the challenges that 6G suffers from, particularly in the context of high-speed communication networks.

The dynamic behavior of the retrial queueing system following the incorporation of Deep Q-Network Reinforcement Learning in 6G mobile communication services is examined in this study. The proposed method lies in analyzing the DQN-RL agent’s learning convergence by using the first- and second-order Markov chain method. By simulating the temporal evolution of reward sequences as Markov and second-order Markov chains, we can quantify convergence characteristics through mixing time analysis. To capture a wide operational landscape, a thorough simulation framework with 120 independent parameter combinations is created. The obtained results indicate that Markov chain analysis confirms 10 training episodes are more than sufficient for policy convergence, and in some cases, as few as 5 episodes allow the agent to enhance the mobile network performance while maintaining low energy consumption. To assess learning stability and system responsiveness, the mixing time of DQN RL rewards is calculated for every episode and configuration. A deeper understanding of the temporal dependencies in the reward process can be gained by incorporating higher-order Markov models. This paper concentrates on studying the learning convergence using an analysis of the Markov model’s spectral gap properties as an indicator. The results provide a rigorous foundation for optimizing 6G queueing strategies under uncertainty by highlighting the sensitivity of DQN convergence to system parameters and retrial dynamics.

Moving toward new and higher frequencies would bring the 6G communication network into practice. Using a new MAC mechanism will enhance and overcome the THz challenges. Our paper focused on analyzing the entropy interdependence between two subsystems, normal and lucky mobile terminals. The two introduced entropy metrics play a crucial role in interpreting the interdependence of the two subsystems. However, the lack of data obliged us to use GAN and its methods to generate similar data. Therefore, we used four different GAN methods plus the standard GAN to obtain the most significant data, and we determined the similarity using six different similarity metrics. The results showed that cosine and correlation similarities are not appropriate to capture the similarity, meanwhile, the rest; dynamic time warping, Frechet inception, root mean square error, and peak signal-to-noise ratio agreed that DCGAN was the one who generated the most accurate data series compared to the rest.

Advancing towards higher frequencies will facilitate the implementation of the 6 G communication network. A novel MAC mechanism is proposed to address and mitigate THz-related challenges. This paper analyzes the entropy interdependence between two subsystems: normal and lucky mobile terminals grouped based on the physical position in the radio cell. Two entropy metrics were introduced to elucidate the interdependence of these subsystems. Due to insufficient data, Generative Adversarial Networks (GANs) were employed to generate analogous data. Five GAN variants, including the standard GAN, were utilized to produce the most relevant data, and eight similarity metrics were applied to assess data similarity. Results indicated that cosine and correlation similarities were inadequate. At the same time Dynamic Time Warping, Frechet inception distance, peak signal-to-noise ratio, relative Euclidean distance, root mean square error and Tanimoto coefficient metrics identified DCGAN as the most accurate generator. The Dynamic Time Warping metric demonstrated a moderate dependence on the minimum number of points parameter as determined by the OPTICS algorithm, whereas other metrics formed two distinct groups based on their dependence.