Modern wireless communication systems rely heavily on multiple antennas and their corresponding signal processing to achieve optimal performance. As 5G and 6G networks emerge, beamforming and beam management become increasingly complex due to factors such as user mobility, a higher number of antennas, and the adoption of elevated frequencies. Artificial intelligence, specifically machine learning, offers a valuable solution to mitigate this complexity and minimize the overhead associated with beam management and selection, all while maintaining system performance. Despite growing interest in AI-assisted beamforming, beam management, and selection, a comprehensive collection of datasets and benchmarks remains scarce. Furthermore, identifying the most-suitable algorithm for a given scenario remains an open question. This article aimed to provide an exhaustive survey of the subject, highlighting unresolved issues and potential directions for future developments. The discussion encompasses the architectural and signal processing aspects of contemporary beamforming, beam management, and selection. In addition, the article examines various communication challenges and their respective solutions, considering approaches such as centralized/decentralized, supervised/unsupervised, semi-supervised, active, federated, and reinforcement learning.

Network Slicing (NS) is an essential technique extensively used in 5G networks computing strategies, mobile edge computing, mobile cloud computing, and verticals like the Internet of Vehicles and industrial IoT, among others. NS is foreseen as one of the leading enablers for 6G futuristic and highly demanding applications since it allows the optimization and customization of scarce and disputed resources among dynamic, demanding clients with highly distinct application requirements. Various standardization organizations, like 3GPP’s proposal for new generation networks and state-of-the-art 5G/6G research projects, are proposing new NS architectures. However, new NS architectures have to deal with an extensive range of requirements that inherently result in having NS architecture proposals typically fulfilling the needs of specific sets of domains with commonalities. The Slicing Future Internet Infrastructures (SFI2) architecture proposal explores the gap resulting from the diversity of NS architectures target domains by proposing a new NS reference architecture with a defined focus on integrating experimental networks and enhancing the NS architecture with Machine Learning (ML) native optimizations, energy-efficient slicing, and slicing-tailored security functionalities. The SFI2 architectural main contribution includes the utilization of the slice-as-a-service paradigm for end-to-end orchestration of resources across multi-domains and multi-technology experimental networks. In addition, the SFI2 reference architecture instantiations will enhance the multi-domain and multi-technology integrated experimental network deployment with native ML optimization, energy-efficient aware slicing, and slicing-tailored security functionalities for the practical domain.

In the context of Fifth Generation mobile networks (5G), Search and Rescue (SAR) missions using Unmanned Aerial Vehicles (UAVs) can benefit from a dynamic, intelligent, and autonomous placement of both Network Functions (NFs) and Artificial Intelligence (AI) systems to quickly adapt in minimal human intervention scenarios. This article examines current 5G architectures and timely standardization efforts within this context. The contribution of this work is to identify associated 5G components and propose AI modules that enable efficient UAV-based SAR missions: the System Intelligence (SI) and Edge Intelligence (EI) concepts. SI is conceived as the entity responsible for defining and orchestrating the placement and processing tasks of NFs and AI systems, while EI is responsible for the optimization of AI-based end-user applications. The article also presents an open-source virtualized testbed that enables a concrete example of SI and EI roles in a SAR mission based on object detection with Deep Neural Networks (DNNs). In this proof-of-concept, the DNN layers are partitioned and the tradeoffs between communication and computational costs are highlighted. For instance, the results indicate that the latency can severely degrade the UAV trajectory and different DNN partitioning options can reduce the required bit rate to transmit DNN scores by more than three times.