Diseño y evaluación de un sistema de comunicación Dispositivo a Dispositivo (D2D) basado en redes cognitivas para entornos más allá de 5G (Beyond 5G (B5G))

Abstract

This work presents the design, implementation, and evaluation of a Device-to-Device (D2D) communication system based on cognitive networks for Beyond 5G (Beyond 5G (B5G)) en-vironments, using NS-3 with the 5G-LENA module and SUMO as simulation tools. The main objective is to implement a resource filtering method and to design and implement dynamic re-source allocation schemes in cellular vehicular networks (NR-V2X) that outperform the scheme defined in the standard. First, the theoretical foundations are reviewed to identify the limitations of the current scheme, and two filtering methods based on energy detectors are evaluated. Subsequently, six resource allocation schemes are developed and compared: two heuristic (Greedy and Proportional Fair) and four using Reinforcement Learning (RL) combined with the Deep Q-Learning (DQL) algorithm. The system validation was carried out using performance metrics such as throughput, delay, block error rate (BLER), and packet reception rate (PRR) in a vehicular platoon communication scenario located in the historic center of Cuenca, Ecuador. The results show that the proposed filtering method consistently improves all metrics compa-red to the standard; heuristic schemes provide slight improvements, while RL-based schemes achieve significant gains, with up to 40 % increases in throughput, 80 % reductions in delay, 30 % improvements in BLER, and a 16 % increase in PRR, especially in scenarios with high vehicular density.