Iterative alternating water-filling for wireless powered MIMO multiple-access channels

Consider a wireless powered Gaussian multiple access channel (MAC), which consists of a power station (PS), multiple passive users and a sink node. Assume that each node is equipped with multiple antennas, and the PS has reliable power supply while the users have to harvest energy from the PS by wireless power transfer (WPT). The harvest-then-transmit scheme is adopted where all users harvest energy in the downlink (DL) timeslot, and then send their own information to the sink by time-division-multiple-access (TDMA) protocol in the uplink (UL) timeslot. The sum throughput maximization problem is investigated by jointly optimizing time allocation and precoding design. We first formulate the optimal design as a convex semidefinite programming, and an iterative alternating water filling solution is proposed to serve as a low-complexity implementation. The optimality of the DL beamforming scheme is proved. By investigating a special case where each user sends only one data stream to the sink, we propose a suboptimal scheme to further reduce the implementation complexity but incur little performance loss. Simulation results demonstrate the efficiency of the proposed algorithms.