AI-Empowered Fluid Antenna Systems: Opportunities, Challenges, and Future Directions

Abstract

Fluid antennas embrace all forms of flexible-position antennas, both movable and non-movable. Hence the concept of fluid antenna system (FAS) introduces a new dimension to enhance multiple-input multiple-output (MIMO) antenna systems, which is essential for achieving more ambitious goals in wireless communications. FAS fundamentally changes the way MIMO systems are optimized. In addition to optimizing precoding and decoding matrices, a flexible-position MIMO system, referred to as MIMO-FAS, needs to optimize the positions (i.e., ports) of the antennas to achieve the best performance. Unfortunately, due to the near-continuous nature of antenna position adjustment as well as the resulting high dimensionality, optimizing MIMO-FAS is NP-hard, complicated by the coupling between the optimization variables. Given the rapid advances in artificial intelligence (AI), it is fitting to harness its capabilities to alleviate the challenges of MIMO-FAS. This article explores a vision of how AI empowers FAS to overcome these hurdles using a learning-based approach, enabling FAS to excel. Furthermore, we use emerging integrated sensing and communication (ISAC) scenarios as a case study to illustrate the potential of enhancing FAS with AI capability.