Automated driving within one lane is a fascinating experience. Yet, it is even more interesting to go a step ahead: Making automated lane changes without human driver interaction. This thesis presents a concept and implementation demonstrated in "Jack", the Audi A7 piloted driving concept vehicle. Given that automated driving is in the media every other day already, why is it still such a big issue to do tactical behavior planning for automated vehicles? It is one of the core areas where it is surprisingly obvious why humans are currently so much smarter than machines: Tactical driving behavior planning is a social task that requires cooperation, intention recognition, and complex situation assessment. Without complex cognitive capabilities in today's automated vehicles, it is core of this thesis to find simple algorithms that pretend intelligence in behavior planning. In fact, such behavior planning in automated driving is a constant trade-off between utility and risk: The vehicle has to balance value dimensions such as safety, legality, mobility, and additional aspects like creating user and third party satisfaction. This thesis provides a framework to boil down such abstract dimensions into a working implementation. Several of the foundations for this thesis were developed as part of the Stadtpilot project at TU Braunschweig. While there has been plenty of research on concepts being tested in perfect, simulated worlds only, the approaches in this thesis have been implemented and evaluated in real world traffic with uncertain and imperfect sensor data. The implementation has been tested, tweaked, and used in "Jack" for more than 50,000 km of automated driving in everyday traffic.
Towards Tactical Lane Change Behavior Planning for Automated Vehicles
Ein Beitrag zur taktischen Verhaltensplanung für Fahrstreifenwechsel bei automatisierten Fahrzeugen
2018
Miscellaneous
Electronic Resource
English