Collision avoidance at intersections: A probabilistic threat-assessment and decision-making system for safety interventions

Road intersections are among the most complex and accident-prone elements of modern traffic networks. Thus, new safety systems have to cope with highly complex traffic scenarios where the behavior of the different road users is difficult to predict. Sensing the surrounding environment and assessing possible threats therefore remain challenging problems. This paper provides a novel, efficient active-safety system for frontal collisions detection and prevention/mitigation. More precisely, we provide: (i) a probabilistic motion prediction algorithm based on an unscented Kalman filter; (ii) a probabilistic threat assessment method based on vectors defined by reference points on the vehicles' structure; (iii) a reachability-based decision-making protocol enabling an emergency intervention. Simulation results, based on realistic data obtained specifically for this scenario, are also presented showing the efficiency and the potential of the proposed solution.