Deep Reinforcement Learning-Based Resource Allocation for Cellular Vehicular Network Mode 3 with Underlay Approach.

Area of Science:

• Wireless Communication Networks
• Intelligent Transportation Systems
• Machine Learning Applications

Background:

• Vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications are crucial for road safety and traffic efficiency.
• Mode 3 underlay systems face interference challenges when V2V links reuse spectrum allocated to V2I links.
• Optimizing resource allocation is essential to balance V2V low-latency needs with V2I throughput demands.

Purpose of the Study:

• To propose a novel resource allocation framework for V2V communications using deep reinforcement learning.
• To address the challenge of spectrum sharing and interference in V2V and V2I underlay systems.
• To improve V2I link capacity while satisfying V2V low-latency requirements.

Main Methods:

• A deep reinforcement learning framework utilizing a double deep Q network for intelligent resource allocation by the base station (BS).
• Optimization of resource allocation strategies based on partial channel state information (CSI) to reduce signaling overhead.
• Simulation-based evaluation of the proposed framework against existing methods.

Main Results:

• The proposed scheme effectively meets the low-latency requirements for V2V links.
• Significant increases in V2I link capacity were observed compared to other methods.
• The performance using partial CSI was comparable to using complete CSI, demonstrating efficiency.

Conclusions:

• Deep reinforcement learning offers an effective approach for intelligent resource allocation in V2V/V2I systems.
• Partial CSI-based optimization provides a practical and efficient solution for reducing signaling overhead without substantial performance degradation.
• The framework successfully balances competing demands of V2V and V2I communication links.