Hybrid intelligent systems for safer traffic: intersection collision avoidance through vehicle to everything and machine learning

Abstract The absence of clear signals at unsignalized intersections can increase drivers’ uncertainty and risk, making the decision-making process more complex. Drivers must make safe decisions based on distance to other vehicles, speed, acceleration, ability, etc., to avoid collisions. Unfortunately, factors like visibility, weather conditions, and the driver’s abilities can influence these decisions. While previous research has addressed these issues, there remains a substantial necessity for more efficient solutions that improve decision accuracy and decrease response times. This paper presents the intelligent graph model-based cooperative intersection collision avoidance system (IG-CICAS), an innovative solution for drivers’ complex decision-making challenges at unsignalized intersections. IG-CICAS integrates Vehicle-TO-Everything communication with real-time graph databases for real-time data exchange and efficient intersection management. The system encompasses driving behavior classification, cloud-based machine learning for collision risk prediction, dynamic adaptation to intersection conditions, and resilience to infrastructure failures. Simulations conducted using the vehicles in network simulation framework across urban intersection scenarios demonstrate that IG-CICAS consistently improves safety and offers significant insights into driving behaviors. This innovative approach addresses the need for smarter, safer traffic management at intersections without clear traffic signals.