A Structure-variable Bayesian Network Model for Vehicle Threat Assessment

Abstract In driving assistant system and automated driving system, accurate and real-time threat assessment is necessary to improve the safety. Besides moving targets, the influence factors of safety also include environment and driver. To evaluate the threat level of host vehicle, a modified Bayesian network (BN) model is proposed in this paper. In addition to objects’ state and environment condition, the driver’s subjective factor is also considered to structure a more adequate model and assess a more accurate threat level, including physical factor and psychological factor. Because of the rate of change of various factors is different greatly, the structure-variable Bayesian network (VBN) model is proposed to increase the computational efficiency. Finally, simulation is designed to verify the availability of two kinds of network models, and it verifies that the VBN is better in computational efficiency than static BN.