Two-step Machine Learning Approach for Channel Estimation with Mixed Resolution RF Chains

Massive MIMO is one of the main features of 5G mobile radio systems. However, it often leads to high cost, size and power consumption. To overcome these issues, the use of constrained radio frequency (RF) frontends has been proposed, as well as novel precoders, e.g., a multi-antenna, greedy, iterative and quantized precoding algorithm (MAGIQ). Nevertheless, the best performance of MAGIQ assumes accurate channel knowledge per antenna element, for example, from uplink sounding reference signals. In this context, we propose an efficient uplink channel estimator by applying machine learning (ML) algorithms. In a first step a conditional generative adversarial network (cGAN) predicts the radio channels from a limited set of full resolution RF chains to the rest of the low resolution RF chain antenna elements. A long-short term memory (LSTM) neural network extracts further phase information from the low resolution RF chain antenna elements. Our results indicate that our proposed ML approach is competitive with traditional Unitary Tensor-ESPRIT in scenarios with various closely spaced multipath components (MPCs).

I. Introduction

Consider a massive MIMO system, where every antenna element is connected to a dedicated radio frequency (RF) chain including a full resolution analog-to-digital converter (ADC). With increasing number of antenna elements, the energy consumption becomes a serious implementation issue. In this regard, the usage of constrained RF frontends has been proposed in [1] and later enhanced by suitably adapted downlink multi-user (MU) MIMO precoders, such as the MAGIQ algorithm [2]. MAGIQ relies on accurate channel state information (CSI), which generally means one full receiver RF chain per antenna element. Receiver RF chains are less complex than transmiter chains; however, limiting the uplink receiver complexity is also important to reduce cost and size. Then, a mix of full and low resolution ADCs can provide a reasonable trade-off between performance and power consumption. Here, we analyze suitable machine learning (ML) methods to most accurately infer the CSI for the low to the high resolution RF chains.