Deep Reinforcement Learning for the Coexistence of eMBB and URLLC Services in NOMA Systems

This paper introduces a novel resource allocation framework for the coexistence of enhanced mobile broadband (eMBB) and ultra-reliable low latency communication (URLLC) services in a non-orthogonal multiple access (NOMA) system. The proposed solution utilizes a two-layer deep reinforcement learning (DRL) algorithm to optimize the dynamic allocation of subbands and power for both URLLC and eMBB users. In the top layer, a deep Q-network (DQN) is employed for discrete subband allocation while the bottom layer uses a multi-agent DRL approach to refine the continuous power allocation. This hierarchical approach ensures that low latency and reliability requirements of the URRLC services are met, without compromising the high throughput demands of eMBB. Extensive simulations validate the efficiency of the proposed algorithm, demonstrating improved system performance in terms of power and subband optimization for both user types.