Port-LLM: A Port Prediction Method for Fluid Antenna based on Large Language Models

Abstract

The objective of this study is to address the mobility challenges faced by User Equipment (UE) through the implementation of fluid antenna (FA) on the UE side. This approach aims to maintain the time-varying channel in a relatively stable state by strategically relocating the FA to an appropriate port. To the best of our knowledge, this paper introduces, for the first time, the application of large language models (LLMs) in the prediction of FA ports, presenting a novel model termed Port-LLM. The proposed method consists of two primary steps for predicting the moving port of the FA: the first involves utilizing the channel tables that encompass historical channel state information from all movable ports of the FA to forecast the channel tables for subsequent time periods; the second step entails selecting the port of the FA for the forthcoming time based on the predicted channel tables and the known reference channels that require alignment. To enhance the learning capabilities of the LLM model in the context of FA port prediction, we incorporate the Low-Rank Adaptation (LoRA) fine-tuning technology. Furthermore, during the model training phase, we implement the warm-up-aided cosine learning rate (LR) technique to augment the accuracy of the predictions. The simulation results show that our model exhibits strong generalization ability and robustness under different numbers of base station antennas and medium-to-high mobility speeds of UE. In comparison to existing FA port calculation methods, the performance of the port predicted by our model demonstrates superior efficacy. Additionally, our model exhibits lower prediction costs and faster prediction and reasoning speeds.

Tasks

Reproductions