State-the-art active safety systems assume simplified presumptions about the environmental events, i.e. they reduce the trajectory prediction for other traffic participants to their current path and if applicable lane markings. With regard to short prediction times, these simplifications only imply small errors, but to increase the active field of ADAS functions as well as for autonomous driving, longer prediction times are required. Therefore, a more sophisticated modelling of all objects participating in a traffic scenario with respect to the infrastructural constraints and traffic rules is required. Such a model should also incorporate the interaction between individual objects.
The goal of this research project is to develop such an interaction model and implement it in a stochastic prediction algorithm which is based on the maneuver forecast of relevant surrounding objects. Based on this prediction, authoritative activation decisions should be made for deescalating, automated interventions to the vehicle dynamics to a preferably early instant of time. This methodology should be applicable without the use of Car2X information, since even in the future non-connected road users will be present and have to be considered in the prediction of traffic scenarios.