Deep Reinforcement Learning Based Energy Efficiency Maximization Scheme for Uplink NOMA Enabled D2D Users

Device-to-device communication (D2D-C) is an leading edge technique in 5G and forthcoming 6G networks due benefits for enhanced spectrum efficiency and energy-efficiency (EE). Despite these potential advantages, co-channel interference (CO-CI), cross-channel interference (CR-CI), and massive connectivity are the major issues in D2D-C. In order to handle these issues, an interference mitigation technique for D2D mobile groups (D2Gs) utilizing up-link Non Orthogonal Multiplexing (NOMA) is presented to improve the EE of the overall network. D2Gs boost the SE by sharing the sub-channels (SCs) to cellular users (CUs), and NOMA links a huge number of D2D users (DUs) to D2D transmitters (DT). The problem is formulated as a mixed-integer nonlinear programming (MINLP) problem with associated SCs and power restrictions of the CUs and DUs. A deep reinforcement learning (DRL) based distributed deep deterministic policy gradient (D3PG) approach is considered to enhance the EE by addressing the resource allocation and power control of DUs. Numerical outcomes showed that the suggested scheme overcomes state-of-the-art techniques in terms of results.