Equidistant Power Allocation in Downlink AE‐NOMA: A Deep Learning Approach

ABSTRACT Nonorthogonal multiple access (NOMA) is a key element of sixth‐generation (6G) wireless networks, designed to improve spectrum efficiency. The introduction of deep learning (DL) in wireless networks has advanced wireless system performance compared to previous generations. In a NOMA system, power is distributed to different users in a cluster or cell. This paper focuses on distributing power to different service demands of a single user in a cell with minimal error rate. We introduce the service‐based downlink autoencoder (AE)‐NOMA scheme for multiservice NOMA transmission for a single user. The power is distributed to different services through an equidistant power allocation scheme. Integrating AE, NOMA, and the power allocation scheme is developed to enable end‐to‐end (E2E) signal transmission with an enhanced signal‐to‐noise (SNR) ratio. It aims to improve the wireless network's block error rate (BLER) performance. The proposed scheme surpasses traditional standard state‐of‐the‐art (SOTA) techniques, the classical successive interference cancellation (SIC)‐NOMA scheme, and the AE‐based NOMA system. The simulation results highlight the effectiveness and adaptability of the proposed scheme across various scenarios.