Uplink-Downlink Duality For Integer-Forcing in Cloud Radio Access Networks

Consider a cloud radio access network where neighboring basestations are able to jointly encode and decode their signals via rate-limited links to a central processor. One promising approach to such systems is for the basestations to simply quantize their observations and send them to the central processor (during the uplink phase) and to emit the quantized signals generated by the central processor (during the downlink phase). Several recent works have proposed compression-based architectures based on sequential source and channel coding. In prior work, we proposed an integer-forcing architecture for the uplink phase, and, in this paper, we propose an integer-forcing architecture for the downlink phase. As part of the achievability argument, we introduce a novel “reverse” integer-forcing source coding strategy that can be used to quantize sources so that their quantization noises are correlated, i.e., multivariate compression. We also establish uplink-downlink duality between our uplink and downlink integer-forcing architectures, and use this as the basis for optimizing the beamforming, equalization, and integer matrices.