Segment training based channel estimation and training design in cloud radio access networks

Cloud radio access networks (C-RANs) have drawn considerable interests due to the significant improvements of spectral and energy efficiencies. Since most signal processing functions are moved to the centralized baseband unit (BBU), remote radio heads (RRHs) in C-RANs can be regarded as soft relays to transfer the received signals. The centralization characteristics in C-RANs make traditional channel estimation and training design approaches inefficient, and the requirements of perfect channel state information (CSI) would not be satisfied in turn. To solve this problem, a segment training based individual channel estimation scheme and the corresponding training design are proposed for C-RANs in this paper. Particularly, the channel estimator in terms of the sequential minimum mean-square-error (MMSE) is developed through a prior knowledge of long-term channel correlation statistics and previous channel estimates. The optimal training design for the developed estimator is derived by minimizing the estimation mean-square-error (MSE). Further, the optimal training design for the channel estimation of radio access links is computed by applying the eigenvalue decomposition (EVD). Numerical results show that performance gains of the proposed channel estimation and training design schemes are significant.