Intelligent Traffic Flow Assist: Optimized Highway Driving Using Conditional Behavior Prediction

Many conventional systems for driver assistance on highways determine vehicle motion based on static situation configurations only. Adaptive Cruise Control systems for example react to vehicles cutting-in from neighboring lanes only after these vehicles have entered the ego-lane. Some more recent systems include prediction of future behaviors, but often neglect mutual influences between these and the best behavior for the ego vehicle. This paper introduces ?intelligent Traffic Flow Assist?, an assistance system that analyzes the current traffic situation and predicts candidate future behaviors conditioned on different options for the ego behavior. It computes vehicle trajectories optimized for safety, comfort and utility for each of the possible future situations and selects the best option for vehicle control. We describe the system architecture and implementation on a prototype vehicle with close-toproduction sensors and report on tests performed in different situations on proving ground and public highways.