A New Active Safety Distance Model Of Autonomous Vehicle Based On Sensor Occluded Scenes

Safety Distance Model (SDM) is a key component of collision avoidance systems. However, traditional SDMs focus only on the obstacles that can be detected by vehicular sensors, ignoring obscured areas where obstacles cannot be detected timely by sensors due to tall trees and buildings on both sides of a road. This paper presents an innovative safety distance model (SDM-PTA) based on real-time identification of potential traffic accident (PTA) areas by a convolution neural network. The research focuses on analyzing relationships between relative speed and distance of an autonomous vehicle and PTA areas. The SDM-PTA accounting for rear-end collision is obtained, which could provide decision-making information for autonomous braking system of intelligent vehicles. Research results show that the SDM-PTA can increase collision avoidance performance of intelligent vehicles, especially at corners and crossroads which are PTA areas.