World modeling in automated driving

Millions of kilometers have been driven with autonomous test vehicles and AutoPilot-like driver assistance functions are appreciated by many users every day. All big car manufactures invest heavily on automated driving and a world without severe accidents seems realistic within years. At the same the Dutch safety board recently concluded that the introduction of advanced driver assistance systems introduces serious risks on road safety. This conclusions seems to contradict promises we have heard for many years. What can we expect from our automated vehicles and what makes this problem complicated from a technological perspective?

This talk looks for an answer to this question by elaborating on one of the fundamental technologies required for vehicle automation: world modeling. A world model represents an automated vehicle’s surroundings and contains the vehicle’s own position, a road model and positions and intentions of other traffic participants. By explaining both required and available technology and algorithms we explore the current state of the art and the promising future of automated driving.