Advanced driver assistance systems (ADAS), which can either help to prevent accidents or reduce their negative consequences by intervening in a wide variety of ways, are the solution of choice for considerably greater safety in traffic and comfortable driving. Further, they can be used to perform various improvements such as optimizing fuel consumption or emissions. However, Validation and verification of such systems can be seen as a limiting factor on their way to public roads. In order to pass the numerous checks and meet certain standards, ADAS must recognize the surrounding traffic as quickly and accurately as possible in order to act upon it. As a result, rapid and accurate identification methods are required to ensure this. In the first part of the talk, we present a template (or pattern) matching (TM) algorithm for maneuver recognition. By using maneuver patterns we are able to consider scenarios of arbitrary complexity and make the identification procedure more systematic and with minimal man expert involvement. The key element of the proposed TM algorithm is the settlement of a distance measure. We use dynamic time warping (DTW) method for an optimal non- linear alignment between two sequences under certain restrictions. It enables a correct assignment of the same maneuver but having different dynamics (speed, aggressiveness, etc.), and is robust to data with missing values. We also discuss the application of the linear functional strategy developed by Sergei Pereverzyev and his team for this use-case.
Another problem towards ADAS certification is that the standard road testing is not affordable due to the infinite number of possible real-life situations. Therefore there is wide consensus that road tests must be complemented by virtual testing. However, also the latter one cannot be performed for all situations, so a finite catalogue of special test cases, so called scenarios, will be used for virtual testing. This catalogue is expected to offer a good coverage of the general intended use of the driving function under test. To this end, it makes sense to derive these scenarios from real data. In the second part of the talk we discuss a systematic way for building up such a catalogue progressively using sensor data. We use a method based on a variant of an on-line k-means algorithm for time series clustering under their alignment using DTW. The advantage of the proposed method is the intuitive representation of the scenarios enabling their easy interpretations. Using experimental data, it is illustrates how such a catalogue is produced and how it can be used for further scenario identification, for example.
This page was last modified on 01/13/2020 - 10:02 CEST