State-Aware Resource Allocation for V2X Communications.

Area of Science:

• Intelligent Transportation Systems
• Wireless Communication Networks
• Artificial Intelligence

Background:

• Vehicle-to-Everything (V2X) is crucial for intelligent transportation, facing challenges in spectrum utilization and multi-user interference.
• Dynamic vehicular environments and time-varying channels hinder spectral efficiency and increase interference in V2X networks.
• Deep Reinforcement Learning (DRL) shows promise for V2X resource allocation, but agent independence limits cooperative performance.

Purpose of the Study:

• To propose a novel state-aware communication resource allocation algorithm for vehicular networks.
• To enhance agent environmental perception and inter-agent cooperation in distributed DRL frameworks for V2X.
• To improve spectral efficiency and mitigate interference in dynamic V2X scenarios.

Main Methods:

• Developed a state-aware algorithm by expanding the state space for enhanced data representation.
• Incorporated a conditional attention mechanism to improve perception of environmental dynamics.
• Utilized deep reinforcement learning for distributed resource allocation in vehicular networks.

Main Results:

• The proposed algorithm significantly improves agents' environmental perception.
• Enhanced inter-agent cooperation was observed, leading to better resource utilization.
• Superior performance in terms of spectral efficiency and interference mitigation in complex V2X scenarios was verified.

Conclusions:

• The state-aware approach effectively addresses limitations of conventional distributed DRL in V2X networks.
• Expanded state space and conditional attention mechanisms foster improved collaboration and environmental awareness.
• The algorithm offers a promising solution for optimizing resource allocation in dynamic intelligent transportation systems.