Massive expansion and implementation of Advanced Driver Assistant Systems and advent of Highly Automated Driving functions brings huge challenges in terms of design and development, but also function validation and certification process which is a limiting factor for their market introduction. To ensure safety of such systems, whose complexity is rapidly growing, it is essential to evaluate functionality of automated driving systems within the mandatory certification before it’s deployed on the road. And after their deployment, they must be a subject to periodical technical inspection during life cycle as well. The number of regulations and standards considering safety of AD functions gradually increases, but current safety standards and regulations still have to be adopted and enhanced. For highly automated driving functions and AVs that do not require permanent monitoring by the driver, a theoretically infinite number of possible traffic situations, that a self-driving car could possibly encounter, needs be tested. One promising method to overcome this matter is the scenario-based approach focused on critical, dangerous and extreme situations. Such approach ensures a repeatability and robustness of an approval process if it is supported by a significant sample of harmonized scenarios. Since confronting conventional physical driving tests with this test effort is not feasible anymore, virtualization of testing methods by means of computer simulation needs to be emphasized. To meet above described challenges, TÜV SÜD is developing a methodology for scenario-based evaluation of AD functionality as a supplement for either development or future certification of automated driving systems. The methodology combines virtual-based approach and physical testing and guarantees repeatability of test conditions. Virtual-based testing is provided by an in-house simulation toolchain with an open architecture. The toolchain consists of functional blocks as: database of standardized scenario, virtual environment model, high fidelity physics-based sensor simulation, model of vehicle dynamics, control functions and algorithms, automated and standardized post-processing and reporting. Physical testing provides real-world data measurement used among other purposes for validation of the simulation toolchain and its relevant functional blocks respectively. Physical testing is performed on our own test track using typical equipment as: driving robots, inertial measurement unit, guided soft target, soft VRU targets, master control station and others. In presentation, an overview of the current state of methodology is given and the workflow is demonstrated for a specific operational design domain (ODD). Architecture of simulation toolchain is described and explanation how functional blocks are embedded into overall architecture and how they interact with each other is given. Trustworthiness for virtual test execution will be discussed by means of a comparison and correlation between real-world and virtual-simulation measurement results for a specific operational design domain.
Dr.-Ing. Jiří Svoboda, TÜV SÜD Czech s.r.o, CZECH REPUBLIC Ing. Vladislav Kocián, TÜV SÜD Czech s.r.o, CZECH REPUBLIC