An improved deep reinforcement learning routing technique for collision-free VANET

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Vehicular Ad Hoc Networks (VANETs) face challenges with routing complexity and high control overhead.
• Existing solutions often fail to integrate routing optimization with overhead reduction.

Purpose of the Study:

• To introduce an Improved Deep Reinforcement Learning (IDRL) approach for VANETs.
• To address routing complexities and minimize control overhead simultaneously.
• To optimize routing paths and reduce convergence time in dynamic vehicle densities.

Main Methods:

• Developed an IDRL routing technique leveraging transmission capacity and vehicle data.
• Utilized Vehicle-to-Infrastructure (V2I) communication for packet transport via adjacent vehicles.
• Simulated the IDRL approach to assess resilience, scalability, and efficiency.

Main Results:

• The IDRL approach effectively reduces transmission delay and augmented control overhead.
• Achieved optimized routing paths and reduced convergence time.
• Demonstrated high efficacy in secure message transmission via V2I communication.

Conclusions:

• The proposed IDRL routing approach significantly decreases latency and increases packet delivery ratio.
• IDRL enhances data reliability in VANETs compared to existing routing techniques.
• The method proves effective in managing dynamic vehicle densities and network overheads.