Robust multi-group multicast beamforming design and antenna selection for massive MIMO systems

Abstract

In this dissertation, we use an Alternating Direction Method of Multipliers (ADMM) algorithm to design robust multi-group multicast beamforming scheme for massive multiple-input multiple-output (MIMO) systems for two scenarios; 1) all antennas are available at the base station (BS), 2) only subset of antennas are available due limitation on the number of RF chains (joint antenna selection and beamforming design). In both scenarios we assume only estimates of the channel covariance matrices are available at the base station with a bounded error. In the first scenario, we formulate the robust multicast beamforming optimization problem to minimize the transmit power while the minimum required quality-of service (QoS) is met. We directly solve the formulated optimization problem. We develop a two-layered ADMM-based fast algorithm to directly tackle the non-convex problem, where we obtain closed-form or semi-closed-form solutions to each subproblem. Simulation results show that our proposed algorithm provides a favorable performance compared with existing alternative methods with considerable computational complexity reduction. In the second scenario, we investigate the problem of joint antenna selection and robust multi-group multicast beamforming scheme for massive MIMO systems. We aim to obtain binary antenna selection and multicast beamforming vectors in order to minimize power consumption at the BS subject to per antenna transmit power limit, and the minimum worst case SINR requirements while limited number of antennas can be exploited for beamforming. To overcome the acquired mixed-Boolean problem, we replace the binary constraints with a continuous equivalent form. We use exact penalty function to transfer the obtained continuous problem into a more tractable problem. We develop a two phases solution, with phase one focusing on antenna selection, and second phase to obtain multicast beamforming vectors using the selected antennas in the first phase. In phase one, we propose two different approaches, SINR-based and SLR-based, to acquire antenna selection vector. We exploit a fast ADMM algorithm to directly tackle the problems in phase one and two without applying convex approximation technique that may result in performance degradation. Simulation results illustrate the advantage of our proposed algorithm in terms of computational complexity over existing alternative methods while maintaining favorable performance.