Abstract:
As the world becomes increasingly connected, it’s essential to maximize the functional ity of low-powered devices in order to fully capitalize on the next-generation Internet of
Things (IoT). This opens up unprecedented opportunities for energy efficiency, sustain ability, and ubiquitous connectivity.This research delves into the application of machine
learning to optimize transmission policies in cognitive radio-inspired non-orthogonal
multiple access (CR-NOMA) networks, central to the emerging landscape of IoT and 6G
communications. The investigation centers on a quality-of-service (QoS)-aware, energy harvesting (EH)-enabled passive IoT device in a CR-NOMA-assisted backscatter com munication network. The sum rate optimization problem is formulated, taking into
account RF power consumption. A non-convex problem is addressed to determine op timal solutions for key parameters, and the identified parameters are used to tackle
a one-dimensional optimization problem through a deep reinforcement learning (DRL)
technique known as deep deterministic policy gradient (DDPG). Using the DDPG algo rithm, dynamic adjustments can be made to the reflection coefficient and time-sharing
coefficient for the backscatter node, targeting peak performance. The goal is to en hance the sum rate of a secondary passive IoT device while meeting the QoS standards
of the primary device. Even with potential short-term setbacks, the overarching aim
remains to boost long-term throughput. Simulations indicate that the introduced DRL assisted NOMA transmission strategy surpasses traditional methods, underscoring the
importance of weaving advanced machine learning into IoT and wireless communication.
Addressing these significant challenges can uplift the performance of forthcoming IoT
networks.
