Spectral-energy efficient resource allocation for multi-user RIS-FD-MIMO systems using deep learning

Reconfigurable intelligent surfaces (RIS) and full-duplex (FD) communications have emerged as promising technologies to enhance the performance of multiple input multiple output (MIMO) systems. However, there are several methods for optimal resource allocation, accurate prediction of resource requirements, and efficient allocation that remain challenging. To overcome these issues, this paper introduces a resource allocation technique by considering spectral efficiency (SE) and energy efficiency (EE). The proposed double attention-based dilated LSTM is first used to predict resources. The proposed improved pufferfish optimisation (IPufO) algorithm is used to allocate resources optimally based on the predicted resource. The proposed IPufO is designed to integrate the chaotic Chebyshev mapping within the conventional pufferfish algorithm to enhance performance. Furthermore, the proposed DA_DiLSTM + IPufO model improves performance by taking SE and EE into account when allocating resources. The maximum EE and SE determined by the proposed technique are 218.531 bits/Joule and 98.8521 bits/Hz/cell, respectively.