A Software-Defined Directional Q-Learning Grid-Based Routing Platform and Its Two-Hop Trajectory-Based Routing...

Area of Science:

• Computer Science
• Networking
• Artificial Intelligence

Background:

• Vehicle-to-Everything (V2X) networks face routing challenges due to high vehicle mobility and variable density.
• Existing Q-learning based routing algorithms, like QGrid, use grid-based approaches but overlook directional information.
• Limitations in current grid-based methods necessitate improved routing strategies for efficient V2X communication.

Purpose of the Study:

• To propose an enhanced routing platform, Software-Defined Directional QGrid (SD-QGrid), for V2X networks.
• To address the shortcomings of existing grid-based routing algorithms by incorporating directional and trajectory data.
• To improve packet delivery ratio and reduce end-to-end delay in V2X communication.

Main Methods:

• Developed a Software-Defined Directional QGrid (SD-QGrid) platform utilizing an SDN Control Node for centralized V2X control.
• Integrated source-destination directionality, real-time positions, and historical trajectory data into the routing decisions.
• Implemented a two-hop trajectory-based routing (THTR) algorithm for real-time next-hop neighbor selection.

Main Results:

• SD-QGrid demonstrated over 10% improvement in average packet delivery ratio.
• Achieved over 25% reduction in average end-to-end delay compared to baseline algorithms.
• Maintained average overhead at less than 2% increase.

Conclusions:

• The proposed SD-QGrid platform effectively enhances V2X packet routing by leveraging directional and trajectory information.
• SD-QGrid offers a superior solution for V2X routing, balancing performance improvements with minimal overhead.
• The findings validate the efficacy of SD-QGrid in real-world V2X network simulations using Taipei City data.