Interference Mitigation and Energy Efficient Resource Allocation Scheme for D2d Communication In 5g and Beyond

Abstract

The proliferation of mobile users, smart gadgets, and multimedia applications generates an unprecedented growth of data traffic in 5G and beyond networks. This massive increase in data traffic puts a lot of burden on efficient spectrum utilization in the years to come. To address this problem, researchers recommended various D2D communication (D2D-C) techniques. In D2D-C technology, two neighboring devices can share the data directly without the base sta tion (BS). As a result, it enhances mobile users quality of service by reducing the transmission delay. Also, in D2D-C, the D2D pairs (DDPs) reuse the same resources as used by cellular users (CUs) to boost the spectral efficiency. Despite these advantages, key challenges such as cross-channel (CR-CI) and co-channel (CO-CI), as well as ultra-massive connectivity (UMC), need to be investigated. To overcome these challenges, academic and industry researchers sug gested the non-orthogonal multiple access (NOMA) approach. NOMA is a scheme in which more than one user shares the same spectrum resources at any instant but with different power levels, resulting in considerable improvements in the spec tral and UMC. Despite these benefits, NOMA brings additional challenges of intra-user in terference, and fairness. The receiver implements the SIC technique to reduce the effects of intra-user interference. However, it is necessary to explore new techniques to improve fairness in dynamic environments. To address this challenge, researchers used the deep reinforcement learning (DRL). DRL proves to be a highly proficient method for optimizing embedded systems, endowed with the remarkable ability to promptly respond in wireless communication networks (WCNs). In the realm of DRL approaches, it is customary to prepare neural networks (NNs) via offline training prior to their implementation on terminal devices or controllers. It employs the trained model to estimate the most optimized transmission power management while keeping a low level of computational complexity with regard