Robust training sequence design for spatially correlated MIMO channels and arbitrary colored disturbance

In this paper, we study the design of optimal robust training sequences for multiple-input multiple-output (MIMO) channel estimation, based on known second order statistics of both the channel and the colored disturbance, but with an uncertainty in the channel covariance matrix. More specifically, the training sequences are designed by taking the least-favorable channel covariance component into account throughout an iterative algorithm. Numerical experiments are carried out to demonstrate the performance gained by employing the proposed design procedure and to compare it with other relevant schemes.