Throughput maximization and latency optimization in fifth-generation networks using a multistage machine learning for early hybrid automatic repeat request

Abstract

The physical layer Hybrid Automatic Repeat Request (HARQ) protocol efficiently achieves low error-rate transmission and high network reliability in the fifth generation (5G) Ultra Reliable Low Latency Communication (URLLC) network. However, this retransmission protocol suffers from increased transmission latency resulting mainly from the delay caused by channel decoding. This problem is caused by the fact that the sender has to wait for acknowledgement of the transmission which is generated after the decoding process at the receiver, resulting in increased latency. To address the latency problem, this study proposed the multistage machine learning Early HARQ (E-HARQ) which uses machine learning algorithms for predicting the acknowledgement before the decoding process. Furthermore, the proposed scheme uses the multistage decision to mitigate the throughput loss resulting from incorrect predictions of the acknowledgement. The multistage decision controls the transmission bandwidth in a multilevel manner depending on channel conditions measured by the Channel State Information (CSI). The study used jupyter notebook and MATLAB for developing the proposed scheme and then evaluating its performance. Simulation results show that the proposed scheme improves the achievable trade-off between the transmission latency and throughput which contributes to the performance of 5G URLLC networks.