1: \begin{abstract}

2: We propose a hierarchical interference mitigation scheme for massive

3: MIMO cellular networks. The MIMO precoder at each base station (BS)

4: is partitioned into an \textit{inner precoder} and an \textit{outer

5: precoder}. The inner precoder controls the intra-cell interference

6: and is adaptive to local channel state information (CSI) at each BS

7: (CSIT). The outer precoder controls the inter-cell interference and

8: is adaptive to channel statistics. Such hierarchical precoding structure

9: reduces the number of pilot symbols required for CSI estimation in

10: massive MIMO downlink and is robust to the backhaul latency. We study

11: joint optimization of the outer precoders, the user selection, and

12: the power allocation to maximize a general concave utility which has

13: no closed-form expression. We first apply random matrix theory to

14: obtain an approximated problem with closed-form objective. We show

15: that the solution of the approximated problem is asymptotically optimal

16: with respect to the original problem as the number of antennas per

17: BS grows large. Then using the hidden convexity of the problem, we

18: propose an iterative algorithm to find the optimal solution for the

19: approximated problem. We also obtain a low complexity algorithm with

20: provable convergence. Simulations show that the proposed design has

21: significant gain over various state-of-the-art baselines.\end{abstract}