Predictive power allocation for real-time wireless applications in NOMA systems

Abstract

The escalating demand for high data throughput, seamless connectivity, and minimal latency in wireless communication necessitates advanced techniques such as non-orthogonal multiple access (NOMA), which offer superior efficiency and scalability compared to traditional methods. This study introduces a novel priority-fading weighted power allocation (PF-WPA) model for NOMA systems, which dynamically allocates transmission power by incorporating user-specific application priorities and varying channel conditions. The PF-WPA model is embedded within a long short-term memory (LSTM) architecture to enhance adaptability and predictive accuracy. The resulting hybrid model is evaluated across diverse fading environments, including Rayleigh, Rician, and Nakagami-m, to validate its predictive accuracy and power allocation performance. Analytical graphs illustrate optimal power distribution patterns for real-time wireless applications. Comparative analyses further confirm the superiority of the proposed scheme over existing ConvLSTM and MLP-LSTM approaches in terms of sum-rate efficiency and fairness. These findings underscore the model’s potential for deployment in latency-sensitive, resource-constrained wireless systems.