Online Learning-Induced Port Selection for Fluid Antenna in Dynamic Channel Environment
Jiaqi Zou, Songlin Sun, Chao Wang
Abstract
To expand the degrees of freedom and achieve additional gains beyond multiple-input multiple-output (MIMO), the fluid antenna system (FAS) has been regarded as a potential approach for exploiting spatial diversity by reconfigurable position-switchable antennas. In the operation of FAS, port selection is one of the key challenges, demanding substantial channel training overhead. In this letter, we propose a novel online learning framework to address the port selection issue without the instantaneous full channel state information (CSI) in time-varying channel conditions. Based on bandit learning, the proposed approach learns the optimal port selection by dynamically interacting with the varying channels. The proposed algorithm is proved to achieve guaranteed performance with a bounded deviation and simulation results confirm comparable performance by comparing with the ideal optimal selections.