Distributed Decoding Scheme for Uplink C-RAN System with Limited Backhaul Capacity

We propose a novel Machine Learning (ML) method to jointly learn the filtering, the quantization scheme, and the decoder, for an uplink Cloud Radio Access Network (C-RAN) with limited backhaul capacity. The proposed ML model employs a linear filter at each Remote Radio-Head (RRH), a uniform quantizer with non homogeneous bitwidth for each RRH output, and a decoder at the cloud. We formulate the learning problem aiming at correctly decoding the symbols sent while respecting a bit-budget constraint. We propose a Multi-Layer Perceptron (MLP) architecture and describe mathematically each of its layers. Numerical results show that the proposed method significantly outperforms well known benchmarks, especially for small capacity backhauls.