Tracking abruptly changing channels in mmWave systems using overlaid data and training

Millimeter-wave (mmWave) multiple-input multiple-output (MIMO) links are sensitive to abrupt changes in the channel due to blockage and node mobility. We propose to estimate the channel by overlaying pilot and data transmissions. The data transmission is performed over the signal subspace of the channel matrix, while the training, for estimating the parameters of newly appearing paths, is performed over the null-space of the channel matrix. A sparse Bayesian learning-based approach is employed for jointly estimating the channel and data at the receiver. Simulations are used to validate the performance of the proposed method in abruptly changing channel scenarios.