Scenario-Based Simulation Framework for Validation of Automated Driving Using Next Generation DoE

Authors

Mag. Ing. mech. T. Mihalj, Assoc. Prof. Dipl.-Ing. Dr. techn. A. Eichberger, Institute of Automotive Engineering, Graz University of Technology; S. Gopi MTech, P. Nitsche PhD MSc, Dr. techn. H.-M. Koegeler, AVL List GmbH, Graz:

Year

2023

Print Info

Production/Publication ÖVK

Summary

The future of mobility is being driven by automated driving, connected vehicles, electrification and mobility-as-a-service, which require enormous effort in testing and validation of those systems. Virtual test environments introduce a faster, safer and more economical approach than real field tests. They provide significant benefits in the establishment of product requirements during the early phases of product development. However, despite the virtual approach, an enormous number of scenarios and their variations are still required to achieve sufficient test coverage. This paper proposes a methodology that significantly reduces the simulation effort by identifying relevant test cases. The methodology is based on Active Design of Experiments (Active DoE), an advanced and sequential process that establishes behaviour models of Key Performance Indicators (KPIs), and automatically adapts the test design to accurately predict simulation outcomes within the range of interest. The benefits are an overall reduction in the number of simulations in comparison to the one-shot designs such as full factorial, and reusable, seamless behavior models for optimization and search purposes. This work presents a case study of an automated electric bus, that navigates through an urban environment with the task to identify safety-critical and uncomfortable operating ranges. Results demonstrate an excellent fit of KPI models within the range of interest with a more than 60% reduction in the number of simulations with respect to full factorial design.

ISBN

978-3-9504969-2-5