Decentralized linear transceiver design in coordinated multi-cell multiuser MIMO systems

This paper considers a downlink linear transmit and receive beamformer design in a coordinated multi-cell network where each multiantenna base station (BS) serves its own set of multiple antenna users. Optimization objective is to minimize sum transmission power among coordinated BSs while satisfying user specific minimum SINR targets. The problem is jointly non-convex in transmit and receive beamformers. Hence, even in a centralized case only a local optimal solution can be found by alternating optimization in which the transmit and receive beamformers are updated consecutively. The problem becomes even more complicated for a decentralized case since even if the channels from the BS to the neighboring cells' users are known, the receivers they are employing may not be. In order to obtain a decentralized implementation, each BS assumes worst case receivers for other cells' users when designing its own users' beamformers. Using this design approach, user specific SINR targets can be guaranteed using only local channel state information at each BS and some limited backhaul signaling among coordinated BSs. The proposed decentralized transceiver design algorithm is solved by repeating the following two optimization steps separately at each BS: transmit and receive beamformer optimization via alternating optimization and inter-cell interference power optimization via primal decomposition. Decentralized implementation comes at a cost of somewhat increased sum power compared to the centralized case. Numerical examples demonstrate fast convergence and significant gain over MISO system when SINR targets are low.