Risks Level Assessments for Automotive Application

The article presents a modelization and assessment of automotive risk accidents taking into account the interactions between environment, driver and vehicle. The evaluated risk is composed of two parts: one concerns the impending risk (i.e. risk of a clearly identified danger and which is present in a short time horizon) and the other one, the latent risk (i.e. risky behavior of the driver which can lead to an accident). The developed tool uses information present in the CAN bus, additional sensors and car communication for shared sensing. With the collected information and estimated variables (e.g. grip and reaction time), it infers a probability of risk with a Bayesian Network. The tool can also be used for evaluating autonomous car driving and driver decisions.