Cloud-RAN Fronthaul Rate Reduction via IBM-based Quantization for Multicarrier Systems

Abstract:
Industrial radio communication is identified as a new use case in the Industry 4.0 (I4.0) initiative as well as in the 3rd Generation Partnership Project (3GPP). 5th Generation (5G) Ultra Reliable Low Latency Communication (URLLC) requirements comprise high reliability and burst error resilience for short packets as well as low latency for I4.0 communication systems. We consider a Cloud Radio Access Network (Cloud RAN) architecture with distributed Radio Access Points (RAPs) that are connected via a rate limited fronthaul to a General Purpose Processor (GPP) cloud-platform. Thus, we can flexibly balance fronthaul data rates and joint processing gains to fully leverage spatial diversity. Here, we conduct an investigation on a functional split within the PHYsical layer (PHY) to harvest these benefits in the uplink while maintaining moderate data rates on the fronthaul for joint decoding. We investigate how data compression according to the Information Bottleneck Method (IBM) on the fronthaul link affects system performance for Generalized Frequency Division Multiplexing (GFDM) as well as OFDM. We show that 3 bit IBM quantization already achieves close to floating point performance in frequency-selective channels.