Dimension Reduction-based Signal Compression for Uplink Distributed MIMO C-RAN with Limited Fronthaul Capacity

Abstract: This paper proposes a dimension reduction-based signal compression scheme for uplink distributed MIMO cloud radio access networks (C-RAN) with an overall excess of receive antennas, in which users are jointly served by distributed multi-antenna receivers connected to a central processor via individual finite-capacity fronthaul links. We first show that, under quantization noise-limited operation, applying linear dimension reduction at each receiver before compressing locally with a uniform quantization noise level results in a sum capacity that scales approximately linearly with fronthaul capacity, and can come within a fixed gap of the cut-set bound. The dimension reduction filters that maximize joint mutual information are then shown to be truncated forms of the conditional Karhunen-Loeve transform, with a block coordinate ascent algorithm for finding a stationary point given. Analysis and numerical results indicate that the signal dimension can be reduced without significant loss of information, particularly at high signal-to-noise ratio, preserving the benefits of using excess antennas. The method is then adapted for the case of imperfect channel state information at the receivers. The scheme significantly outperforms conventional local signal compression at all fronthaul rates, and with complexity linear in network size represents a scalable solution for distributed MIMO C-RAN systems.