Machine Learning Applied to 6G Radio over Fiber Systems Linearization

This work proposes the development and demon-strates the applicability of a machine learning (ML)-based digital pre-distortion (DPD) tool applied to the sixth-generation of mobile network (6G) radio over fiber (RoF) systems linearization. Particularly, our tool was idealized for linearizing electrically amplified analog RoF systems (A-RoF), which transport downlink signals from the central office, where the baseband unit might be implemented using a software-defined radio approach, to simplified remote radio units (RRUs). A memory recurrent neural network (RNN) DPD and its performance in terms of root mean square error vector magnitude (EVMRMS) and adjacent channel leakage ratio (ACLR) are investigated for different activation functions.