Scalable Nonlinear Multiuser Detection for mmWave Massive MIMO

We consider uplink multi-user reception for a mmWave picocell with a large number of base station antennas and a proportionally large number of users. Receiver complexity can become a bottleneck for scaling such a system, necessitating development of more computationally efficient algorithms. Due to the spatial sparsity of mmWave channels, transforming the received signal into beamspace concentrates the power of each user into a small number of dimensions. Our prior work takes advantage of this fact to develop a local LMMSE estimator that relies on a small window of the beamspace signal for demodulating each user's signal. In this paper, we propose to layer nonlinear interference cancellation on top of the local LMMSE receiver: while most interference is suppressed linearly by the local LMMSE receiver, for each user, a small number of strongly interfering users are handled by interference cancellation after suitably whitening the local LMMSE output. The method provides reliable demodulation at higher load factor (defined as number of users divided by the number of antennas) than enabled by linear interference suppression alone, at order of magnitude lower complexity than standard interference cancellation.