Joint group power allocation and prebeamforming for joint spatial-division multiplexing in multiuser massive MIMO systems

We investigate the joint optimization of the group power allocation and prebeamformer for joint spatial division and multiplexing (JSDM) in massive MIMO downlink systems. In contrast with the approximated block diagonalization (ABD) prebeamformer which is derived by heuristic method in the original JSDM scheme and is restricted to semi-unitary matrix, general prebeamforming matrix together with group power is optimized in the framework of per-user ergodic rate balancing. Thus, both flexibility and optimality are integrated in our work. To ease the difficulty in optimizing the exact ergodic rate, the deterministic approximation is employed. Based on the uplink-downlink duality approach, an iterative algorithm alternatively updating group power and prebeamformers is designed. It is shown that the optimization subproblems of the algorithm can be solved efficiently. Compared with the classic signal-to-noise-and-interference ratio (SINR) balancing algorithm for MISO downlink, the proposed algorithm has similar properties of both convergence and optimality. Numerical experiments are conducted to validate the effectiveness of the proposed algorithm.