Dynamic Medium Access in Clustered NOMA IoT Networks based on Reinforcement Learning

The burgeoning growth of massive machine-type communication or massive Internet of things traffic in beyond fifth-generation communication systems calls for novel solutions for medium access. In this paper, we propose a reinforcement learning (RL)-based random access scheme for non-orthogonal multiple access (NOMA)-enabled uplink transmissions and investigate how RL can potentially bring benefits for medium access under various network configurations and traffic load conditions. Aiming to mitigate network congestion and enhance overall network performance, we introduce access control and compare network performance of the RL-based access scheme with two benchmark schemes. For performance assessment, we define system and cluster throughput as the key performance metrics. This study provides valuable insights on the feasibility and efficacy of RL-based access control mechanisms across low and high device density scenarios.