IRS Element Selection Using LSTM-Based Deep Learning for UAV Communications

Sobia Jangsher, Arafat Al-Dweik, Emad Alsusa

Research output: Contribution to journalArticlepeer-review

Abstract

This letter proposes using deep learning (DL) for intelligent reflecting surface (IRS) element selection to reduce the bit error rate (BER) of unmanned aerial vehicle (UAV) communications affected by imperfect phase estimation and compensation. In the presence of phase errors, increasing the number of elements does not necessarily reduce the BER. In contrast, muting certain IRS elements can improve BER. However, solving the optimization problem heuristically is extremely complex because it requires evaluating BER expressions numerically for an enormous number of cases. Consequently, a long shortterm memory (LSTM)-based element selection (ES) technique is proposed to reduce the substantial complexity inherent in the conventional solution. A supervised learning approach with offline training is adopted where the decision of ES is made based on the phase estimation error parameter j. The obtained results show that the computation time of the proposed technique is 100 times less than that of state-of-the-art algorithms.

Original languageBritish English
Pages (from-to)1
Number of pages1
JournalIEEE Wireless Communications Letters
DOIs
StateAccepted/In press - 2023

Keywords

  • 6G mobile communication
  • 6th generation (6G)
  • Autonomous aerial vehicles
  • Bit error rate
  • Complexity theory
  • Deep learning
  • deep learning (DL)
  • intelligent reflecting surface (IRS)
  • long short-term memory (LSTM)
  • Phase estimation
  • Training
  • unmanned aerial vehicle (UAV)

Fingerprint

Dive into the research topics of 'IRS Element Selection Using LSTM-Based Deep Learning for UAV Communications'. Together they form a unique fingerprint.

Cite this