Beam management optimization for V2V communications based on deep reinforcement learning.

Area of Science:

• Wireless Communications
• Artificial Intelligence
• Intelligent Transportation Systems

Background:

• Intelligent connected vehicles are crucial for smart cities and intelligent transportation.
• Vehicular networks transmit diverse data (safety, sensing, multimedia) requiring high spectral efficiency, low latency, and reliability.
• 5G NR FR2 (24-71 GHz) is recommended for V2X, but faces challenges like high path loss and fluctuating channels.

Purpose of the Study:

• To propose a deep reinforcement learning (DRL)-based intelligent beam management method for vehicle-to-vehicle (V2V) communication.
• To address the trade-offs between spectral efficiency, delay, and reliability in 5G NR FR2 V2X communications.
• To optimize beam alignment and tracking in complex, dynamic vehicular environments.

Main Methods:

• A deep reinforcement learning (DRL) algorithm was developed for intelligent beam management.
• The DRL method focuses on optimal control of beam alignment and tracking.
• Simulations were conducted to compare the proposed method against 5G standard and Extended Kalman Filter (EKF) approaches.

Main Results:

• The DRL-assisted method significantly reduced communication delay compared to the 5G standard.
• The proposed method demonstrated superior reliability and spectral efficiency over the EKF-based approach.
• DRL effectively managed beam alignment and tracking in challenging 5G NR FR2 V2X scenarios.

Conclusions:

• Deep reinforcement learning offers an effective solution for intelligent beam management in 5G NR FR2 V2X communications.
• The proposed DRL method successfully balances spectral efficiency, delay, and reliability.
• This approach enhances the performance of vehicular networks in complex and dynamic environments.