Threat assessment design under driver parameter uncertainty

We consider a model-based threat assessment method, which enables the activation of assisting safety interventions in case an accident threat to the driver is detected. The method relies on vehicle and driver mathematical models and reachability analysis tools. In particular, we focus on the problem of false threats detection that can occur due to uncertainties in the driver model, i.e., the driver is incorrectly deemed incapable of accomplishing a driving task. This paper proposes a novel approach, to compensate for uncertainties in the driver model, for the considered threat assessment method. In particular, we show how the considered threat assessment method can be designed such that, if a threat is detected, the driver is guaranteed to be unable to perform the assigned driving task. In such case, an automated assisting intervention can be motivated.