Multi-Angle Camera View Hybrid Vision Transformer-Based Deep Reinforcement Learning Application in Complex Intersection Scenarios for Autonomous Vehicles

Autonomous vehicles struggle at unsignalized intersections, especially in bad weather like heavy rain at night. Most existing studies focus only on safety, limited scenarios, and good weather. This study introduces a new AI method that helps self-driving cars better understand complex traffic scenes and make safer, more efficient decisions. The system was tested in highly crash-prone intersections under different weather conditions using a driving simulator. Results show it performed better than leading existing methods, successfully navigating most scenarios while maintaining both safety and efficiency.