Technology Category
Machine Learning: Hardware-aware MLCognitive Modeling & Cognitive Systems: Neural Spike CodingSearch and Optimization: Non-convex Optimization
Learning to Quantize and Precode in Massive MIMO Systems for Energy Reduction: a Graph Neural Network Approach
🤖 AI Summary
To address the hardware complexity and power consumption bottlenecks caused by high-resolution DACs in massive MIMO systems, this paper proposes an end-to-end graph neural network (GNN)-based nonlinear precoding framework tailored for low-bit DACs (e.g., 1-bit). The method jointly models quantization and precoding by designing a differentiable GNN architecture that directly outputs quantized precoding vectors. A self-supervised learning objective maximizes user achievable rates, while a straight-through Gumbel-Softmax gradient estimator overcomes the non-differentiability of DAC quantization. Experimental results demonstrate that, in single-user scenarios, the 1-bit DAC scheme achieves higher sum rate than conventional 3-bit maximum-ratio transmission (MRT). Baseband and RF DAC power consumptions are reduced by 4–7× and 3×, respectively; consequently, total system power consumption exhibits a net reduction for bandwidths below 3.5 MHz.
Technology Category
Application Category
📝 Abstract
Massive MIMO systems are moving toward increased numbers of radio frequency chains, higher carrier frequencies and larger bandwidths. As such, digital-to-analog converters (DACs) are becoming a bottleneck in terms of hardware complexity and power consumption. In this work, non-linear precoding for coarsely quantized downlink massive MIMO is studied. Given the NP-hard nature of this problem, a graph neural network (GNN) is proposed that directly outputs the precoded quantized vector based on the channel matrix and the intended transmit symbols. The model is trained in a self-supervised manner, by directly maximizing the achievable rate. To overcome the non-differentiability of the objective function, introduced due to the non-differentiable DAC functions, a straight-through Gumbel-softmax estimation of the gradient is proposed. The proposed method achieves a significant increase in achievable sum rate under coarse quantization. For instance, in the single-user case, the proposed method can achieve the same sum rate as maximum ratio transmission (MRT) by using one-bit DAC's as compared to 3 bits for MRT. This reduces the DAC's power consumption by a factor 4-7 and 3 for baseband and RF DACs respectively. This, however, comes at the cost of increased digital signal processing power consumption. When accounting for this, the reduction in overall power consumption holds for a system bandwidth up to 3.5 MHz for baseband DACs, while the RF DACs can maintain a power reduction of 2.9 for higher bandwidths. Notably, indirect effects, which further reduce the power consumption, such as a reduced fronthaul consumption and reduction in other components, are not considered in this analysis.
Problem
Research questions and friction points this paper is trying to address.
Reducing energy consumption in massive MIMO systems with coarse quantization.
Proposing a GNN-based solution for non-linear precoding in quantized downlink MIMO.
Overcoming non-differentiability in DAC functions using Gumbel-softmax gradient estimation.
Innovation
Methods, ideas, or system contributions that make the work stand out.
GNN for non-linear precoding in MIMO
Self-supervised training for achievable rate
Gumbel-softmax for gradient estimation
🔎 Similar Papers
No similar papers found.
