Optimized Uplink Transmission in Multi-Antenna C-RAN With Spatial Compression and Forward

MIMO and cloud radio access network (C-RAN) are promising techniques for implementing future wireless communication systems, where a large number of antennas are deployed either being co-located at the base station or totally distributed at separate sites called remote radio heads (RRHs), both to achieve enormous spectrum efficiency and energy efficiency gains. Here, we consider a general antenna deployment method for wireless networks, termed multi-antenna C-RAN, where a flexible number of antennas can be equipped at each RRH to more effectively balance the performance and fronthaul complexity tradeoff beyond the conventional massive MIMO and single-antenna C-RAN. To coordinate and control the fronthaul traffic over multi-antenna RRHs, under the uplink communication setup, we propose a new “spatial-compression-and-forward (SCF)” scheme, where each RRH first performs a linear spatial filtering to denoise and maximally compress its received signals from multiple users to a reduced number of dimensions, then conducts uniform scalar quantization over each of the resulting dimensions in parallel, and finally sends the total quantized bits via a finite-rate fronthaul link to the baseband unit (BBU) for joint information decoding. Under this scheme, we maximize the minimum SINR of all users at the BBU by a joint resource allocation over the wireless transmission and fronthaul links. Specifically, each RRH determines its own spatial filtering solution in a distributed manner to reduce the signaling overhead with the BBU, while the BBU jointly optimizes the users´ transmit power, the RRHs´ fronthaul bits allocation, and the BBU´s receive beamforming with fixed spatial filters at individual RRHs. Numerical results show that, given a total number of antennas to be deployed, multi-antenna C-RAN with the proposed SCF and joint optimization significantly outperforms both massive MIMO and single-antenna C-RAN under practical fronthaul capacity constraints.