Decentralizing the optimal multi-cell beamforming via large system analysis

A multi-cell minimum power beamforming problem is considered. It is known that the inter-cell interference (ICI) terms couple the base stations (BS) and inter-cell coordination is required for global optimal solution. The cooperation can be realized by exchange of instantaneous channel state information (CSI) or terms related to the ICI values via a backhaul link. However, the limited backhaul capacity and delay constraints put a limit on achievable performance when the number of antennas and users grow large or when dealing with a fast fading scenario. In this work, we demonstrate that the ICI terms coupling the coordinating BSs can be approximated using the random matrix theory (RMT) tools when the problem dimensions grow large, and that the approximated ICI values depend only on channels statistics, i.e, spatial load and user specific path loss values. This leads to a significant reduction in the information exchange rate among BSs. Furthermore, processing is simplified because with a fixed approximated ICI values the beamforming vectors can be obtained locally at each BS. The proposed solution guarantees the feasibility of the target signal-to-interference-plus-noise ratios (SINR) without any major loss of performance as compared to the optimal centralized design.