Resource Allocation in Multi-cell NOMA Systems with Multi-Agent Deep Reinforcement Learning

Non-orthogonal multiple access (NOMA) technology can meet user access requirements and improve system capacity. In this paper, we investigate the joint subcarrier assignment and power allocation problem in an uplink multi-cell NOMA system to maximize the energy efficiency (EE) while ensuring the minimum data rate of all users. We propose a multi-agent deep reinforcement learning (MADRL) method with centralized training and distributed execution to solve this dynamic optimization problem. In our method, we design a deep q-network (DQN) with parameter sharing to generate the subcarrier assignment policy, and use multi-agent deep deterministic policy gradient (MADDPG) network for power allocation of NOMA user. Finally, we adjust the entire resource allocation policy by updating the parameters of neural networks according to the reward. The simulation shows that our method has better and more stable sum EE than centralized and distributed methods.