DDPG Algorithm with Elastic Replay Buffer Assisting in the Optimization of the NOMA System

Non-orthogonal multiple access networks inspired by cognitive radio technology can enhance energy efficiency and address the issue of spectrum resource scarcity. However, resource allocation strategies are still needed for optimization. This paper applies deep reinforcement learning (DRL) algorithms to optimize transmission strategies. We propose a DDPG algorithm integrated with elastic experience replay, which helps the agent strike a better balance between exploring new strategies and utilizing existing experiences. Simulation results demonstrate that the proposed algorithm achieves a higher transmission rate compared to the standard DDPG algorithm, random algorithm and greedy algorithm.