Machine-Learning Physical-Layer Receivers for Future 6G Networks

Improved sustainability and energy-efficiency is one of the decisive factors in future 6G systems. One way to improve the energy-efficiency of base station and UE transmitters is to operate close to the saturation level of the involved power amplifier (PA) systems. This, however, leads to excessive distortion and transmit waveform quality degradation that has been classically mitigated through digital predistortion. In this talk, as an alternative to digital predistortion, we discuss, present and demonstrate how advanced machine-learning (ML) methods can be applied at the receiver side to still efficiently demodulate and decode the information bits despite largely distorted transmit waveforms. We cover both CP-OFDM and DFT-s-OFDM based networks, and show concrete examples and corresponding results for the available energy savings as well as available coverage enhancements in the future networks through the ML-based physical-layer receiver technology.