The Dresden Method": A toolbox for the holistic evaluation of active safety and automated driving functions

被引：0

作者：

Mai M. ^{[1
]}

Bäumler M. ^{[1
]}

Lehmann M. ^{[1
]}

Siebke C. ^{[1
]}

Blenz K. ^{[1
]}

Prokop G. ^{[1
]}

Bönninger J. ^{[2
]}

Höpping K. ^{[2
]}

机构：

[1] FSD Fahrzeugsystemdaten GmbH, Dresden

来源：

VDI Berichte | 2022年 / 2022卷 / 2387期

关键词：

Clustering algorithms;

D O I：

10.51202/9783181023877-419

中图分类号：

学科分类号：

摘要：

In the past ten years, the Chair of Automobile Engineering (LKT) of TU Dresden and the FSD – Central Agency for PTI have focussed their common research and development efforts on a universal methodology toolbox for the holistic effectiveness and risk evaluation of active safety functions and automated driving functions: “The Dresden Method”. Real world data from several sources (naturalistic driving data, camera based traffic observations, accident data, etc.) are being merged into a common data basis by fusion algorithms, which also allow the supplement of partly missing information by means of artificial intelligence methods. Critical traffic situations are being labelled in the normal traffic data to better identify the relevant traffic behaviour. Self-developed criticality measures are being used for this labelling, which are based on the controllability of a situation for the human behaviour of road users. For a function under test, a representative test scenario catalogue, which covers the field of application of the function in a testable manner, is being extracted from that common data basis by means of self-developed clustering algorithms. This scenario catalogue is used by several evaluation m thods subsequently. The safety evaluation on a macroscopic level is done by means of stochastic traffic simulations using physio-psychological models of road user behaviour, which provide a realistic virtual traffic and accident behaviour for a statistical analysis of the impact on road safety (with a reduced level of detail). The safety evaluation on a microscopic level is done by means of driving simulator studies and real world test drives (with a higher level of detail). This methodology toolbox provides a tool set for function development and validation, assessment, authorisation, periodical technical inspection, and field monitoring. It is the foundation of development for the initiative “Safety of connected and automated road traffic”, a common undertaking of TU Dresden, FSD, and further partners for the generation and supply of evaluation scenarios, methods, and tools for future road safety. © 2022, VDI Verlag GMBH. All rights reserved.

引用

页码：419 / 434

页数：15

共 32 条

[1] Mai M., Virtuelle Methoden der TU Dresden zur Verkehrssicherheitsbewertung hochautomatisierter Fahrfunktionen, Forum Zum Verkehrssicherheitspotential Moderner Fahrzeugtechnik, (2018)
[2] Nalic D., Mihalj T., Baumler M., Lehmann M., Eichberger A.S., Scenario Based Testing of Automated Driving Systems: A Literature Survey, FISITA World Congress, Online, (2020)
[3] Prokop G.M., Methodik zur Bewertung sicherheitserhöhender Fahrfunktionen, IEDAS Interdisziplinärer Expertendialog Aktive Sicherheit und automatisiertes Fahren, Essen, (2017)
[4] Trautmann T., Muller B., Staffetius T., Bonninger J.J., Fahrerindividuelle Erkennung von fahrdynamischen Grenzwerten, TÜV-Tagung Sicherheit Durch Fahrerassistenz, München, (2010)
[5] Hopping K., Bonninger J., Prokop G.M., Driver’s license for automated vehicles – towards representative test scenarios, Safetyweek 2021: Auto[Nom]Mobil, (2021)
[6] Baumler M., Dziuba-Kaiser L., Yin Z., Lehmann M.G., Use information you have never observed together: Data fusion as a major step towards realistic traffic scenarios, IEEE International Conference on Intelligent Transportation Systems, (2020)
[7] Dziuba-Kaiser L., Fusion von Unfallszenarien für die Repräsentativitätsüberprüfung eines Testszenarienkataloges zur Absicherung automatisierter Fahrfunktionen, Diplomarbeit, Technische Universität Dresden, (2020)
[8] Siedel G., Evaluation von Machine-Learning-Modellen und Konzeptionierung eines Modell-Ensembles für die Vorhersage von Unfalldaten, Diplomarbeit, Technische Universität Dresden, (2021)
[9] Baumler M.G., Datenfusion als Mittel zur Anreicherung von Unfalldatenbanken, Forum für Verkehrssicherheit, München, (2020)
[10] Lehmann M., Baumler M.G., Use of a criticality metric for assessment of critical traffic situations as part of SePIA, Internationales Stuttgarter Symposium, Stuttgart, (2019)

← 1 2 3 4 →