A Novel Warning/Avoidance Algorithm for Intersection Collision Based on Dynamic Bayesian Networks

Collision on road is a classic and important problem as it brings about great loss on humans' safety and the efficiency of the traffic system. Numerous algorithms have been proposed to address the issue, and most of the existing researches are focused on rear-end, overtaking and lane change scenarios. However, due to the massive computing process, limitation of prediction time and complex layouts at intersections, most conventional algorithms are not suitable for intersection collision avoidance. This paper makes a theoretical analysis on the variations of vehicle states taking advantage of the Dynamic Bayesian Networks (DBNs). Based on the analysis, the risk assessment process is made out to identify a dangerous situation. Different collision avoidance strategies are implemented to avoid accidents if a potential danger is detected and the driver is informed of warning. Simulations are carried out to evaluate the ability of the proposed algorithms on detecting and mitigating a vehicle collision. Results show that the proposed algorithms have great performance in handling the vehicle collisions at intersections.