Automated Driving in Real Traffic: from Current Technical Approaches towards Architectural Perspectives 2014-01-0159

In 2012, Continental received an autonomous vehicle testing license from the US state of Nevada and has subsequently operated an automated driving vehicle on more than 15.000 miles of testing on public roads. The present paper describes the development of this vehicle by explaining the necessary system modules - structured along the signal processing chain from sensing and representing the vehicles environment up to a holistic actuator concept. Therefore, the functional specification for automated lateral and longitudinal vehicle guidance is addressed, also giving an answer on how to cope with challenging scenarios such as stop-and-go traffic and narrow road lanes. Focus is set on a hybrid environment representation, integrating model-based tracking for moving traffic participants and an occupancy grid for maneuvering space.

To put the described technical realization into the context of architectural research, we present a scalable and modular functional architecture to provide a clear perspective on the generic design of future automated vehicles.

As a result of this paper, the need for a sensor setup and data processing that provides maneuvering space information was clearly demonstrated. Furthermore, it is shown that the proposed concept of a hierarchically layered architecture can efficiently organize the driving task from mission planning down to trajectory planning, reducing the implementation and testing effort due to its modularity.