JOURNEY FROM 5G TO 6G: LARGE LANGUAGE MODELS TO SMALL LANGUAGE MODELS, ANTENNAS TO FLUID ANTENNA SYSTEMS—A COMPREHENSIVE SURVEY

Abstract

The introduction of fifth-generation (5G) wireless networks has been a major leap in the field of mobile communications and the networks promise enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC). However, a vision of the next decade, the sixth-generation (6G) era, envisions a seamlessly integrated cyber-physical world with such applications as holographic telepresence, pervasive intelligence and the Internet of Everything (IoE). These applications require unprecedented performance requirements that push 5G to its limits. This survey discusses the underlying paradigm shifts needed to get from 5G to 6G. We begin with examining the 5G limitations in terms of the emerging 6G requirements. We then position Artificial Intelligence (AI) not as an application but as a native part of the network architecture with a particular focus on the evolution of Large Language Models (LLMs) to agile and task-specific Small Language Models (SLMs) for network management. In the following we get into the radical change at the physical layer, where we explore the era of the traditional multiple antenna systems and the promising paradigm of Fluid Antenna Systems (FAS) for unprecedented spatial multiplexing and diversity gains. Additionally, we present survey on other key enablers, such as integrated sensing and communication (ISAC), reconfigurable intelligent surfaces (RIS) and terahertz (THz) communication. The aim of this article is to give an overall picture of the most important technological frontiers that characterize the path towards 6G, to point out the synergies, challenges and future research directions.