Pragmatic multi-cell MIMO beamforming with decentralized coordination

This paper considers a time division duplexing based coordinated multi-cell multiuser MIMO system. Optimization problem is sum power minimization among base stations (BSs) with minimum SINR constraint per each user. The problem is jointly non-convex with respect to transmit and receive beamformers. We propose a MIMO beamforming design where transmit and receive beamformers are updated consecutively using decentralized coordination. For receive beamforming, each user employs minimum mean squared error receiver which is then used as a precoder for uplink pilot signaling. Due to precoded pilot signaling, local effective channel state information is available at each BS. Using a primal decomposition method, originally centralized transmit beamforming optimization is decoupled between BSs and turned into a decentralized two-level optimization. At each BS, a lower BS-level optimization is cast as a second order cone program and solved using standard convex optimization tools, whereas a higher network-level optimization is solved via a projected subgradient method requiring limited backhaul information exchange among BSs. Alternating between receive and transmit beamforming optimization steps, algorithm converges to a locally optimal solution in a static channel scenario. Numerical results show similar performance as in the centralized case. Furthermore, significant performance gain over multi-cell MISO beamforming is demonstrated.