Most traffic accidents that result in injuries or fatalities occur in intersections. In Japan, where cars drive on the left, most of such accidents involve cars that are turning right. This situation serves as the basis of the development of our Advanced Driver Assistance System (ADAS) for intersection right turns. This research focuses on the scenario in which an object darts out from the blind spot created by heavy oncoming traffic as a vehicle is making an intersection right turn. When this happens, even if the driver brakes as hard as possible or an active safety function such as the Autonomous Emergency Braking System (AEBS) applies the brakes, the natural limits of physical friction may make it impossible to avoid a collision. To improve traffic safety given the limited potential of physical friction, this research seeks to develop a risk-predictive right-turn assistance system. The system predicts potential oncoming objects and reduces the vehicle velocity in advance. Blind corners can be detected by on-board sensors without requiring information from surrounding infrastructure. This paper presents a right-turn assistance system that avoids conflict with the AEBS in emergencies by decelerating the ego vehicle to a safe velocity.
Design and Implementation of Driver Assistance System using Machine Learning
