The sleepwalker framework: Verification and validation of autonomous vehicles by mixed reality LiDAR stimulation

Abstract

Verification and validation of autonomous mobile systems, such as autonomous vehicles, is indispensable, since conflicts and serious incidents are rarely acceptable when human beings are involved. Although integrative simulation frameworks are commonly applied to test these systems, such simulations are usually too idealistic, while real world tests are both, expensive and not reproducible. To overcome this problem, we present the framework Sleepwalker for verifying and validating autonomous vehicles: Similar to a human sleepwalker, our framework stimulates the automated driving function at a sensor close level with virtual laserscans mixed with sensor data from the real environment. Thus, the autonomous driving function explicitely builds up a mixed reality environment model as a basis for the subsequent components and therefore enables an overall performance assessment. The instantiation of the framework is adaptable so it to can be balanced between the required result's plausibility and scenario criticality. We demonstrate the distinguished benefits of our framework by different instantiations stimulating an autonomous vehicle and conclude with further research questions.