Deep Reinforcement Learning-Based Coordinated Beamforming for mmWave Massive MIMO Vehicular Networks.

Area of Science:

• Wireless communication systems engineering
• Signal processing for telecommunications
• Machine learning applications in networking

Background:

• Millimeter wave (mmWave) beamforming is crucial for beyond fifth-generation (B5G) technology, utilizing Multi-input Multi-output (MIMO) systems.
• High-speed mmWave applications face significant challenges including signal blockage, latency, and high training overhead for beamforming vector discovery in massive antenna arrays.

Purpose of the Study:

• To propose a novel deep reinforcement learning (DRL) based coordinated beamforming scheme for mmWave systems.
• To address challenges of blockage, latency, and training overhead in highly mobile mmWave applications.
• To enhance the efficiency and performance of mmWave massive MIMO systems.

Main Methods:

• A coordinated beamforming scheme is developed where multiple base stations (BSs) jointly serve a single mobile station (MS).
• A deep reinforcement learning (DRL) model is employed to predict optimal beamforming vectors from a codebook of candidates.
• The scheme focuses on mitigating training overhead and latency in dynamic mobile environments.

Main Results:

• The proposed DRL-based coordinated beamforming scheme significantly increases achievable sum rate capacity.
• The system demonstrates dependable coverage and low latency for highly mobile mmWave applications.
• A remarkable reduction in training and latency overhead was observed compared to conventional methods.

Conclusions:

• The novel DRL-based coordinated beamforming scheme effectively enhances mmWave massive MIMO system performance.
• The solution provides a robust framework for dependable, low-latency, and efficient mmWave communication for mobile users.
• This approach offers a promising direction for future B5G wireless networks.