Location Prediction Using Bayesian Optimization LSTM for RIS-Assisted Wireless Communications

Xuejie Hu, Yue Tian, Yau Hee Kho, Baiyun Xiao, Qinying Li, Zheng Yang, Zhidu Li, Wenda Li

Research output: Contribution to journalArticlepeer-review

Abstract

Reconfigurable intelligent surface (RIS) represent a novel form of electromagnetic metamaterial that have been extensively studied for user equipment (UE) positioning by exploiting the multipath propagation of signals. A novel RIS-assisted localization prediction (RLP) method based on Bayesian optimization and long short-term memory (BO-LSTM) has been proposed in this paper. This method capitalizes on the predictive advantages of LSTM for data sequence and RIS's flexible and controllable multidimensional feature parameters, establishing a mobile UE localization model in an RIS-assisted wireless communications system based on the interplay between time slot transmission power and user location information. In order to provide a more stable communication environment for data collection during the localization process, a power allocation optimization (PAO) method is proposed for maximizing time slot channel capacity in the RLP system based on the number of RIS reflection elements. The study conducts a thorough comparison of simulation results of BO-LSTM, convolutional neural networks (CNN)-LSTM and improved bidirectional LSTM (BiLSTM) combined with Adaptive boost, employing adaptive moment estimation (Adam) and stochastic gradient descent with momentum (SGDM) optimizers. Experimental results demonstrate that the BO-LSTM-based RLP method exhibits improved prediction accuracy. These findings suggest the effectiveness of the proposed method and highlight its potential for further enhancement.
Original languageEnglish
Article number10549836
JournalIEEE Transactions on Vehicular Technology
Early online date5 Jun 2024
DOIs
Publication statusE-pub ahead of print - 5 Jun 2024

Keywords

  • Bayesian optimization LSTM
  • location prediction
  • reconfigurable intelligent surface
  • wireless communications

ASJC Scopus subject areas

  • Automotive Engineering
  • Aerospace Engineering
  • Computer Networks and Communications
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Location Prediction Using Bayesian Optimization LSTM for RIS-Assisted Wireless Communications'. Together they form a unique fingerprint.

Cite this