Robust downlink beamforming using covariance channel state information

The problem of multiuser downlink beamforming is studied under the assumption that the transmitter has erroneous covariance-based channel state information (CSI). The goal is to minimize the transmit power under the worst-case quality-of-service (QoS) constraints. Previous convex optimization-based solutions to this problem involve several coarse approximations of the original problem. In our proposed solution, such coarse approximations are avoided and an exact representation of the worst-case solution is obtained using Lagrange duality. The so-obtained problem is then converted to a convex form using semidefinite relaxation (SDR). Computer simulations show that the SDR step does not involve any approximation as the resulting solution is always rank-one. Simulation results demonstrate substantial performance improvements over earlier worst-case optimization-based downlink beamforming techniques.