1: \begin{abstract}

2: We consider the uplink of a cloud radio access network (C-RAN), where

3: massive MIMO remote radio heads (RRHs) serve as relays between users

4: and a centralized baseband unit (BBU). Although employing massive

5: MIMO at RRHs can improve the spectral efficiency, it also significantly

6: increases the amount of data transported over the fronthaul links

7: between RRHs and BBU, which becomes a performance bottleneck. Existing

8: fronthaul compression methods for conventional C-RAN are not suitable

9: for the massive MIMO regime because they require fully-digital processing

10: and/or real-time full channel state information (CSI), incurring high

11: implementation cost for massive MIMO RRHs. To overcome this challenge,

12: we propose to perform a two-timescale hybrid analog-and-digital spatial

13: filtering at each RRH to reduce the fronthaul consumption. Specifically,

14: the analog filter is adaptive to the channel statistics to achieve

15: massive MIMO array gain, and the digital filter is adaptive to the

16: instantaneous effective CSI to achieve spatial multiplexing gain.

17: Such a design can alleviate the performance bottleneck of limited

18: fronthaul with reduced hardware cost and power consumption, and is

19: more robust to the CSI delay. We propose an online algorithm for the

20: two-timescale non-convex optimization of analog and digital filters,

21: and establish its convergence to stationary solutions. Finally, simulations

22: verify the advantages of the proposed scheme.

23: \end{abstract}

24: