Citation

Kuhn W, Muller M, Höppner T. Transp. Res. Proc. 2017; 27: 222-229.

Copyright

(Copyright © 2017, Elsevier Publications)

DOI

10.1016/j.trpro.2017.12.011

PMID

unavailable

Abstract

Vehicles need to recognize and record the roadway and the associated driving area elements with the help of sensors (stereo cameras, radar, lidar, laser scanners etc.) in highly automated driving processes; this information then needs to be converted into a digital 3D model in real time. The vehicle can then locate and orient itself and move in a so-called obstacle-free and restricted 3D area. Localizing the vehicle precisely in the surroundings is often difficult for several reasons: The volume of data needing to be processed in real time, the accuracy of the object recognition process and the multiple disturbances like the weather, daytime and nighttime or the traffic situation etc. To gradually solve the object recognition problems in real time when relying on the available sensors and disturbances, the vehicle should have a detailed prior knowledge of the traffic infrastructure on the planned route before the journey starts; this can take place through highly developed maps (HD maps with separate layers) within its navigation system. The localization of the vehicle can take place faster and more accurately as it compares the prior knowledge and the knowledge obtained from its surroundings. When selecting a route, the prior knowledge about the existing roads is then directly retrievable and can considerably accelerate the localization process in the surroundings; and it can particularly make things safer if any problems occur. The digital data on roads and surroundings can also be used to calculate recommended speeds and the necessary distances between vehicles and for highly automated driving. There are separately layers inside of the HD maps for the digital geometric data and the kinematic parameters.

Language: en

Keywords

automated driving; critical velocity; prior knowledge; road data