Programa del congreso

Multiple-Input Multiple-Output (MIMO) systems play a crucial role in fifth-generation (5G) mobile communications, primarily achieved through the utilization of precoding matrix techniques. This paper presents precoding techniques employing codebooks in downlink MIMO-5G wireless communications, aiming to enhance network performance to meet the overarching 5G objectives of increased capacity and reduced latency. We conduct a comparative analysis of various precoding techniques outlined by the 5G standard through diverse simulations across different scenarios. These simulations enable us to assess the performance of the different precoding techniques, ultimately revealing the strengths and weaknesses inherent in Type I and Type II codebooks.

Sustainability is a critical factor in the development of future beyond-5G (b5G)/6G networks. This work highlights the importance of energy efficient practices to reduce environmental impact and resource consumption, in line with the objectives of the International Mobile Telecommunications (IMT) 2030. The increasing energy consumption in telecommunications networks highlights the need for corrective action without compromising network performance or security. The paper presents an Open Radio Access Network (O-RAN) architecture for 5G networks that enables the monitoring and optimization of energy consumption. The study integrates xApps and rApps into various emulated network elements to provide a real streaming service while monitoring the power consumption of the different network elements.

As Open Radio Access Network (O-RAN) architecture gains prominence, it also introduces new security challenges particularly concerning the open interfaces, virtualization and the intelligence embedded within the network. This paper makes an overview of the security of this architecture and presents a testbed for the extraction of different metrics for an E2E video service under two situations: with and without attack. This testbed can be used to generate datasets and evaluate the impact of poisoning and evasion on network intelligence.

The evolution of mobile networks is currently going through a stage of opening up the infraestructure, known as O-RAN, a paradigm that also proposes providing more intelligence to the Radio Access Network (RAN) of the users. The key element that allows this change is the RAN Intelligent Control (RIC). Possible service improvements to customers are affected by new security breaches that may occur on the network. This paper analyses the impact of poisoning and evasion attacks, where training and testing data, respectively, are altered on Machine Learning (ML) algorithms. To this end, an E2E scenario has been analysed, in which the direct effects of the users' perception is studied.

With the advances in 5G, and beyond,

mobile communications, more focus is directed toward implementing

massive MIMO. While numerous solutions exist for

high-frequency bands exceeding 1GHz, exploration of increasing

MIMO order in low-frequency bands has not been yet explored,

despite their unique characteristics as wide coverage area and

penetration through obstacles that are essential for covering a

massive number of connected devices. The challenges of designing

antenna base station arrays that can support massive MIMO

at 5G New Radio (5G NR) 700 MHz bands are summarized.

Here, the goal is to move from the current non-massive 4T4R

solution to a massive 16T16R MIMO without the increase in the

overall size of the base station (restricted to standard sizes) or

sacrificing array performance, especially in terms of bandwidth, gain,

beamwidth, and isolation. Results based on an array version of

differential metallic cap-loaded multi-layer Dielectric Resonator

Antenna (DRA) are presented.

In beyond 5G cellular systems, Network Slicing (NS) functionality allows splitting a physical network into several logical slices tailored for a specific application. In sliced networks, a slice-aware automatic optimization of Network Functions (NFs) is key to guarantee Service Level Agreement (SLA) compliance while minimizing operation costs. For this purpose, with the advances in artificial intelligence, massive data collected in the operations and support system can be leveraged to develop Deep Reinforcement Learning (DRL) solutions deriving optimization policies automatically. This work proposes the first algorithm for slice-aware traffic steering based on DRL. The algorithm trains a slice-specific double deep Q-learning agent that learns the optimal traffic steering policy to improve SLA compliance per slice. Simulation results have shown that the proposed algorithm outperforms other slice-aware traffic steering approaches, increasing SLA compliance on 11\% in a realistic scenario with 3 slices serving uRLLC, uLBC and eMBB traffic.