Non-Iterative Reconstruction and Selection Network-Assisted Channel Estimation for mmWave MIMO Communications.

Area of Science:

• Wireless communication
• Signal processing
• Sensing technologies

Background:

• Millimeter-wave (mmWave) MIMO systems are crucial for next-generation wireless networks, enabling high data rates.
• These systems offer potential for integrated sensing, requiring precise object detection and localization.
• Existing channel estimation methods, like learned approximate message passing (LAMP), are computationally intensive, limiting large-scale deployments.

Purpose of the Study:

• To develop a low-complexity channel estimator for mmWave MIMO systems.
• To improve sensing resolution and communication efficiency in mmWave systems.
• To address the limitations of iterative channel estimation methods.

Main Methods:

• Proposed a non-iterative reconstruction network (NIRNet) for efficient beamspace channel reconstruction.
• Introduced a learning-based selection matrix (LSM) generating a signal-aware Gaussian measurement matrix.
• Incorporated a denoising network to enhance accuracy under low signal-to-noise ratio (SNR) conditions.

Main Results:

• NIRNet achieved significantly reduced computational overhead compared to LAMP-based methods.
• The LSM outperformed traditional Bernoulli matrices in channel estimation.
• The proposed algorithm demonstrated superior normalized mean squared error (NMSE) and achievable sum rate (ASR).

Conclusions:

• The NIRNet-based algorithm offers a low-complexity, high-performance solution for mmWave MIMO systems.
• The method enhances both communication efficiency and sensing resolution, vital for real-time applications.
• Reduced training overhead and improved accuracy under low SNR conditions were observed.