Exploiting Deep Learning in Limited-Fronthaul Cell-Free Massive MIMO Uplink

Abstract

A cell-free massive multiple-input multiple-output (MIMO) uplink is considered, where quantize-and-forward (QF) refers to the case where both the channel estimates and the received signals are quantized at the access points (APs) and forwarded to a central processing unit (CPU) whereas in combinequantize-and-forward (CQF), the APs send the quantized version of the combined signal to the CPU. To solve the non-convex sum rate maximization problem, a heuristic sub-optimal scheme is exploited to convert the power allocation problem into a standard geometric programme (GP). We exploit the knowledge of the channel statistics to design the power elements. Employing largescale-fading (LSF) with a deep convolutional neural network (DCNN) enables us to determine a mapping from the LSF coefficients and the optimal power through solving the sum rate maximization problem using the quantized channel. Four possible power control schemes are studied, which we refer to as i) small-scale fading (SSF)-based QF; ii) LSF-based CQF; iii) LSF use-and-then-forget (UatF)-based QF; and iv) LSF deep learning (DL)-based QF, according to where channel estimation is performed and exploited and how the optimization problem is solved. Numerical results show that for the same fronthaul rate, the throughput significantly increases thanks to the mapping obtained using DCNN.

Uplink Spectral and Energy Efficiency of Cell-Free Massive MIMO With Optimal Uniform Quantization

“…The limited capacity links from the APs to the CPU constitute one of the most important challenges in cell-free massive MIMO [13]- [18]. This fronthaul limitation is a more crucial challenge on the uplink, as in the downlink mode the signals are sent as bit streams to the APs which then apply local modulation and coding whereas as the fronthaul links send the quantized version of the received signals at the APs to the CPU, which introduces additional self-interference to the signals at the CPU.…”

“…Proof: Using Lemma 1 and the analysis in [27], the achievable SINR is obtained by (16). Next, using (15) and (11), and after some mathematical manipulation, we have (18), where W E&Q,lf and F E&Q,lf are defined in (20). It is straightforward to calculate the terms E 1 |Ǧ 2 , E 2 |Ǧ 2 , and…”

“…, ∀ , where we drop the index as we use the same number of bits at all APs and use fh = , (43,8) , (39,9), (35,10), (31,11), (27,12) (23,13), (19,14), (15,15) , (11,16), (7,17), (3,18)} where using fh = , (43,8) , (39,9), (35,10), (31,11), (27,12) (23,13), (19,14), (15,15) , (11,16), (7,17), (3,18) } and fh = 12 Mbits/s for all pairs. 6) The Effect of Total Number of Users:: In this section, we investigate the of total number of users on the system performance.…”

A Review of Deep Learning in 5G Research: Channel Coding, Massive MIMO, Multiple Access, Resource Allocation, and Network Security

“…Vieria et al, specifically utilized CNNs to show that massive MIMO channel measurements can be used to achieve precise positions inference for fingerprintbased inference of user positions by using its sparse channel structure [32]. In addition, [33] and [34] used CNN based models to attempt and solve the non-convex sum rate maximization and sum spectral efficiency optimization problems in massive MIMO. With a deep convolutional neural network consisting of 32 convolution layers, 37 residual layers, average pooling layer, fully connected layer and sigmoid part, [33] is able to determine a mapping from the large scale-fading coefficients and optimal power using the quantized channel.…”

Cell-Free Massive MIMO : Scalability, Signal Processing and Power Control

“…Deep Learning [176][177][178][179][180] Channel Model Rician fading [181][182][183] Spatially correlated Rayleigh fading [184][185][186] Miscellaneous Full-duplex [187][188][189] Channel non-reciprocity [190] Asynchronous reception [191] System information broadcast [192] Over-the-air signaling [193,194] Multi-antenna users [195,196] Frequency division duplex [197] Stochastic geometry [198] Dynamic Resource allocation [199] Wireless power transfer [200]…”