This work presents a predictive safety state evaluation in vehicles with machine learning methods. Based on logged and labeled data of driving sequences predictive models are trained. By continuously evaluating those models during a drive, a criticality value for potential upcoming accidents can be estimated up to 10 s in the future. One important aspect is the integration of slow changing criticality measures. Thus, information about the driver state, driving maneuver, road condition or environment can be included via feature design in the prediction process.
«
This work presents a predictive safety state evaluation in vehicles with machine learning methods. Based on logged and labeled data of driving sequences predictive models are trained. By continuously evaluating those models during a drive, a criticality value for potential upcoming accidents can be estimated up to 10 s in the future. One important aspect is the integration of slow changing criticality measures. Thus, information about the driver state, driving maneuver, road condition or environ...
»