Enhancing the Minimum Awareness Failure Distance in V2X Communications: A Deep Reinforcement Learning Approach

Area of Science:

• Vehicular communication networks
• Intelligent transportation systems
• Machine learning for autonomous driving

Background:

• Cooperative awareness is crucial for vehicular networks, typically defined by vehicles' perception and sharing of kinematic data.
• Existing awareness metrics often overlook the complexities of vehicle detection, tracking, and safety distance maintenance.

Purpose of the Study:

• To propose a novel, multi-faceted definition of awareness in Vehicle-to-Everything (V2X) communications.
• To introduce a deep reinforcement learning framework for joint control of beacon rate and transmit power (DRL-JCBRTP) to enhance V2X awareness.
• To improve driving safety by minimizing awareness failure probability and maximizing awareness distance.

Main Methods:

• Developed a deep reinforcement learning framework (DRL-JCBRTP) utilizing LSTM-based actor and MLP-based critic networks within the Soft Actor-Critic (SAC) algorithm.
• Implemented an innovative reward function to increase the minimum awareness failure distance, leveraging local state information.
• Conducted simulations using SLMLab-Gym-VEINS to evaluate the proposed framework's performance.

Main Results:

• The DRL-JCBRTP scheme significantly outperformed existing beaconing schemes in simulations.
• Demonstrated a reduction in awareness failure probability.
• Showcased an increase in the maximum achievable awareness distance.

Conclusions:

• The proposed DRL-JCBRTP framework offers a superior approach to enhancing cooperative awareness in V2X networks.
• Optimizing beacon rate and transmit power through deep reinforcement learning leads to improved driving safety.
• The novel definition and framework advance the field of intelligent transportation systems.