SceneGen - Intelligent Scenario generation for test and validation of highly automated driving functions

Project: Research project

Project Details

Description

Virtual testing and validation is a key technology to bring automated driving to scale with cost and time efficient development processes. In the predecessor projects VDC-DAS and IAFA, a methodology for the development of test cases has already been developed for the ACC, LKA, AEB and the lane change assistant, which specifically targets the detection of corner cases. This consisted of two separate approaches:

-Generation of fixed scenarios that represent reproducible test cases for system design and checking of requirements specifications.
-Generation of random scenarios by coupling microscopic traffic flow simulation to the virtual development environment.

These developed methods were limited to the automated lane change and included the SAE Level 1 functions ACC and LKA as well as a generic straight, three-lane highway. As part of the development of the ALP.Lab development environment, a defined section of a motorway between Graz and Lassnitzhöhe serves as a test environment for the test and validation of a Level 3+ highway pilot. For this purpose, this section is set-up in the software package IPG CarMaker on the basis of highly accurate digital maps. The traffic is generated via co-simulation with the microscopic traffic flow simulation software PTV Vissim. In order to be comparable to reality, it is necessary to calibrate the Vissim model with measured individual vehicle data. Subsequently, automated post-processing based on innovatiove evaluation metrics will create an interactive tool for the systematic scenario-based testing of automated driving functions. As a test object, the project develops the function of a Highway Pilot, which combines the functions ACC, LKA, lane change, entries and exits, as well as construction sites.
StatusFinished
Effective start/end date1/01/1831/12/18

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.