A Deep Learning Approach to 5G Physical Layer Abstraction for Data Channels

The physical layer of fifth-generation (5G) technology is characterized by significant complexity, particularly in its high-consuming computation times during simulations. In this study, we propose a novel abstraction framework for the physical layer of 5G networks to reduce computation times while maintaining high accuracy in the results. This solution is grounded in a Deep Neural Network (DNN) capable of predicting the Block Error Rate (BLER) based on the instantaneous channel state, modulation, and coding rate used in the transmission. The paper details the structure of the framework, the architecture of the employed DNN, and presents results demonstrating the performance of the framework in comparison to classical abstraction mechanisms.