Deep Deterministic Policy Gradient for Throughput Maximization in Energy Harvesting NOMA-Cognitive Radio Network

Non-Orthogonal Multiple Access (NOMA) is a promising technology for spectrum efficiency, and it is getting prominence in 5G cellular systems. In this work, we consider an uplink NOMA aided Cognitive Radio (CR) network. In the network, the Secondary Users (SUs) can also transmit their data signals to the Cognitive Base Station (CBS) through same frequency channel simultaneously. Now for the problem of energy consumption, we considered a wireless solar energy harvester to recharging the batteries of the SUs to protract their operations. Here, we proposed a Deep Deterministic Policy Gradient (DDPG) based approach of Deep Reinforcement Learning (DRL) for the overall throughput maximization in long-term scheme of an uplink NOMA aided CR network. The obtained simulation result provides an improvement in the throughput compare to the traditional Deep Q Network and actor critic algorithms.