Power Minimization in Wireless Network Virtualization with Massive MIMO

Abstract

This paper presents a new method of wireless network virtualization to share the downlink of a massive-MIMO base station among multiple Service Providers (SPs). The SPs are allowed to simultaneously utilize all antennas and channel resource of the Infrastructure Provider (InP). This can improve resource utilization but also requires the InP to control the interference between SPs that are oblivious of each other. We develop novel precoding schemes to minimize the InP's transmission power while satisfying certain prescribed maximum deviation between each SP's intended signal to its users and what the InP delivers. This problem is studied for both perfect and imperfect channel state information (CSI). Under perfect CSI, the proposed precoding is optimal and substantially outperforms a time-sharing alternative. Under imperfect CSI, we use a numerical lower bound to show that the proposed precoding is nearly optimal.