Evaluating the Utility of Driving: Toward Automated Decision Making Under Uncertainty

Abstract

The complexity of advanced driver-assistance systems (ADASs) is steadily increasing. While the first applications were based on mere warnings, current systems actively intervene in the driving process. Due to this development, such systems have to automatically choose between different action alternatives. From an algorithmic point of view, this requires automatic decision making on the basis of uncertain data. In this paper, the application of decision networks for this problem is proposed. It is demonstrated how this approach facilitates automatic maneuver decisions in a prototypical lane change assistance system. Furthermore, relevant research questions and unsolved problems related to this topic are identified.