DNN Based Sum Rate Prediction for Time and Power Partitioning RSMA in 5G and Beyond

Cellular internet-of-things (IoT) domain has undergone rapid evolution, transitioning from an emerging technology to a swiftly expanding and dynamic field. Cellular IoT can serve several devices with various channel state information at transmitter (CSIT) features and requirements at the same time. This work examines multiple-input single-output (MISO) broadcast channel (BC) for an overloaded scenario under two distinct CSIT conditions: Users with high-end, premium devices, for which the transmitter has imperfect CSI, make up one group, while users with IoT smart devices, for which the transmitter has statistical CSI, make up the other. The rate splitting multiple access (RSMA) is a transmission scheme aided for the considered scenarios for Power or Temporal Partitioning and Time Partitioning based on their impact on the sum rate. From simulation results it is noted that in overloaded conditions, Power Partitioning within RSMA shows robustness in maintaining maximum-minimum user fairness and demonstrates durability in managing inaccuracies in CSIT. A deep neural network (DNN) model is used to forecast the system's sum rate in situations where the CSIT fluctuates quickly because of a high number of dynamic channels.