1: \begin{abstract}

2: Hybrid precoding, which consists of an RF precoder and a baseband

3: precoder, is a popular precoding architecture for massive MIMO due

4: to its low hardware cost and power consumption. In conventional hybrid

5: precoding, both RF and baseband precoders are adaptive to the real-time

6: channel state information (CSI). As a result, an individual RF precoder

7: is required for each subcarrier in wideband systems, leading to high

8: implementation cost. To overcome this issue, two-timescale hybrid

9: precoding (THP), which adapts the RF precoder to the channel statistics,

10: has been proposed. Since the channel statistics are approximately

11: the same over different subcarriers, only a single RF precoder is

12: required in THP. Despite the advantages of THP, there lacks a unified

13: and efficient algorithm for its optimization due to the non-convex

14: and stochastic nature of the problem. Based on stochastic successive

15: convex approximation (SSCA), we propose an online algorithmic framework

16: called SSCA-THP for general THP optimization problems, in which the

17: hybrid precoder is updated by solving a quadratic surrogate optimization

18: problem whenever a new channel sample is obtained. Then we prove the

19: convergence of SSCA-THP to stationary points. Finally, we apply SSCA-THP

20: to solve three important THP optimization problems and verify its

21: advantages over existing solutions.

22: \end{abstract}