Reinforcement Learning based Resource Management for 6G-Enabled mIoT with Hypergraph Interference Model

For the future 6G-enabled massive Internet of Things (mIoT), how to effectively manage spectrum resources to support huge data traffic under the large-scale overlapping caused by the dense deployment of massive devices is the imperative challenge. In this paper, a novel hypergraph interference model is designed, and two reinforcement learning (RL)-based resource management algorithms in the 6G-enabled mIoT are proposed to enhance the network throughput and avoid overlapping interference. Then, based on the hypergraph interference model, the resource management problem of execution network throughput maximization is theoretically formulated under large-scale overlapping interference scenarios. To handle this problem, we convert it into a Markov decision process (MDP) model and then deal with this MDP model through the advantage actor-critic (A2C)-based resource management algorithm and asynchronous advantage actor-critic (A3C)-based resource management algorithm, which aim to maximize network throughput of the spectrum resource allocation among massive devices. The simulation results verify that the proposed algorithms can not only avoid large-scale overlapping interference but also improve the network throughput.