Automotive Industry Insights

Examples with special relevance for type approval

Scenario-Based Approval of ALKS (Automated Lane Keeping Systems) using X-in-the-Loop

Scenario-based test approaches have several features that make them interesting for assessment, release, and type approval of automated driving functions. The first aspect to mention here is the analogies to methods used for risk evaluation. These analogies allow the use of in-the-loop methods to confirm cases identified during analysis. This might include aspects like severity ratings or input parameters for controllability considerations, as well as identification of edge cases.

Secondly, scenario-based in-the-loop methods might be used to build a positive risk argument along the lines of UNECE R157 (ALKS) Annex 4, Appendix 3, by comparing the system of interest with an (existing) benchmark. Such a benchmark might also be a driver model as suggested in UNECE R157.

Thirdly, to confirm the safety of the intended function (SOTIF from ISO/DIS 21448 or “operational safety” in UNECE sources), it will be necessary to argue the absence of unknown hazardous scenarios (refer to “area 3” from ISO/DIS 21448). This endeavor is sometimes also referred to as the exploration of the unknown. In case this exploration requires a parameter variation, in-the-loop techniques should complement real-world observations.

SW/HW Reprocessing/Data-Replay for Perception Systems
With the currently available methods and experiences, the argument to control risk cannot end after product release. Instead, the final evidence for acceptable risk (e.g., in terms of a positive risk balance) can only be provided in the field. Thus, field observation has become an essential part of the UNECE regulations. Due to limitations in connectivity and in-vehicle data storage capacity, testing capabilities to reproduce adverse interactions in real-world application are required to operate an automated vehicle fleet safely. Reprocessing of recorded (or reconstructed) sensor data might allow a reconstruction, analysis, and future mitigation of such adverse effects.

VIL and Proving Ground
Previously, the outstanding role of in-the-loop and reprocessing techniques has been discussed. However, the key element that makes these methods efficient – namely partial modeling (and thus simplification) of the real world – is also a point of weakness once it comes to systematic deviations. To keep these potential systematic issues under control, a close alignment with other test results that are closer to the real-world application of automated vehicles on public roads is required. This means vehicle-in-the-loop and proving ground tests are key to model validation in other in-the-loop approaches. Beyond this, proving ground tests are integrated in the UNECE multi-pillar approach as a mandatory part, requiring the accomplishment of a minimum set of challenges.