Optimal Training in Continuous Flat-Fading Massive MIMO Systems

Massive multiple-input multiple-output (MIMO) communication promises the full potential of MIMO communication realized through the deployment of hundreds of antennas at each base station. Accurate channel state information (CSI) is crucial for making the most of massive MIMO. In time-varying channels, however, if not periodically updated, CSI obtained at the base station becomes progressively outdated. In this paper, we formulate a training optimization problem for maximizing uplink sum-rates. The optimal training period and update interval are characterized as solutions to fixed-point equations, which depend on the normalized Doppler shift and large-scale channel coefficients. It is found that, for a sufficiently large number of antennas at each base station, the optimal training period is equal to the number of users per cell.