Learned-SBL-GAMP based hybrid precoders/combiners in millimeter wave massive MIMO systems.

Area of Science:

• Wireless communication systems
• Signal processing
• Machine learning

Background:

• Millimeter-wave (mmWave) massive Multiple-Input Multiple-Output (MIMO) systems require efficient hybrid precoder/combiner designs to boost antenna gain and minimize hardware complexity.
• Traditional Sparse Bayesian Learning via Expectation Maximization (SBL-EM) faces high computational challenges with increasing signal dimensions, limiting its practical application.
• Existing methods struggle with the computational demands of designing optimal hybrid precoders for large-scale mmWave MIMO systems.

Purpose of the Study:

• To propose a novel algorithm, Learned-Sparse Bayesian Learning with Generalized Approximate Message Passing (L-SBL-GAMP), for designing optimal hybrid precoders/combiners in mmWave Massive MIMO systems.
• To address the limitations of high computational complexity and suitability for large datasets associated with traditional SBL-EM algorithms.
• To enhance the spectral efficiency and performance of mmWave Massive MIMO systems through an advanced machine learning approach.

Main Methods:

• The proposed L-SBL-GAMP algorithm extends the SBL-GAMP framework by integrating a Deep Neural Network (DNN).
• The DNN component is trained to optimize the design of hybrid precoders/combiners based on the characteristics of the training data.
• The algorithm leverages generalized approximate message passing principles for efficient signal estimation and precoder design.

Main Results:

• The L-SBL-GAMP algorithm demonstrates reduced computational complexity, fewer iterations, and lower training overhead compared to the SBL-EM algorithm.
• Simulation results show that L-SBL-GAMP achieves higher achievable rates and improved accuracy.
• The proposed method outperforms existing algorithms like OMP, SOMP, SBL-EM, and SBL-GAMP in mmWave Massive MIMO architectures.

Conclusions:

• The L-SBL-GAMP algorithm offers a computationally efficient and effective solution for designing hybrid precoders/combiners in mmWave Massive MIMO systems.
• The integration of DNNs within the SBL-GAMP framework significantly enhances system performance, including spectral efficiency and achievable rates.
• L-SBL-GAMP presents a promising advancement for practical deployment in next-generation wireless communication systems.