A6
Autonomous Driving 1

Back to overview

10:40
conference time (CEST, Berlin)
Numerical Simulation for Supporting Validation and Certification of Automated Driving Features
26/10/2021 10:40 conference time (CEST, Berlin)
Room: A
E. Landel (ELC, FRA); E. Arnoult (Renault, FRA)
E. Landel (ELC, FRA); E. Arnoult (Renault, FRA)
New features are introduced in vehicle to assist the driver and in the future proposed a full autonomous driving. The vehicle equipped with such systems must be safe by design in very large range of situations. Validation by testing at proving ground or open road are necessary and made as usual but can’t cover an extended range of situations because it becomes too long and too expensive. The use of simulation the only affordable solution to explore the huge number of situations which can be encountered by the vehicle within the system’s ODD. Classically, simulation is used during design phase to support definition and optimization of ADAS in link with other systems. Many methods and tools integrated to the engineering processes have been developed with success. But the characteristics of them are not fully adapted for direct application to certification activities. The main reason is that high fidelity of models is mandatory to ensure credibility of the validation provided by a massive simulation. Methods need to be reinforced. A simulation plan will reproduce many driving situations based on a model of the ego vehicle immerged in a scenario. The collection of scenarios with their parameters related to infrastructure, other vehicles, obstacle and environment conditions must be managed in a way of covering all the ODD attributes, all accident scenarios and all other relevant scenarios within the ODD, in particular for extreme and rare situation which are usually named “corner or edge cases”. The vehicle model is an assembly of subsystems. The vehicle becomes a cybersystem including multiphysics, control and communication. New methodologies are proposed for assembling such complex systems with a high level of reliability. At the end, a high level of fidelity can be obtained by filling the gap between numerical results and testing signals with technics of “hybrid twin” based on AI.
validation Autonomous vehicle
11:00
conference time (CEST, Berlin)
Simulation Credibility for Virtual Validation of Automated Vehicle Systems
26/10/2021 11:00 conference time (CEST, Berlin)
Room: A
M. Benedikt (Virtual Vehicle Research GmbH, AUT)
M. Benedikt (Virtual Vehicle Research GmbH, AUT)
Virtual Validation represents a strictly mandatory key solution concept for testing and approving complex automated vehicle systems. This approach is necessarily based on Modelling and Simulation (M&S) of the vehicle systems at hand for significantly reducing testing efforts, in terms of time and costs. As any simulation represents an approximation of the real-world-system quantification of M&S, quality is essential and is expressed by credibility, i.e. the quality to elicit belief or trust in M&S results. Simulation Governance and Management refers to a strictly top-down approach for ensuring fit-for-purpose simulation quality, starting from a simulation strategy, and is based on quality assessment for process, methodological and tooling levels. It is concerned with (a) selection and adoption of the best available simulation technology, (b) formulation of mathematical models, (c) management of experimental data, (d) data and solution verification procedures, and (e) revision of mathematical models in the light of new information collected from physical experiments and field measurements. For introduction of credible simulations initially the way simulations are used in automotive system development is pointed out. This contribution discusses quality assurance for virtual experiments, where simulation targets (test order) are tailored top- to low-level requirements of simulation artefacts. Therefore, a new Credible Simulation Process (CSP) is highlighted for streamlining system M&S, supplemented by a Credibility Assessment Framework (CAF) for quantification. For arguing quality a so called Credibility Argument is derived. A leading-edge application of this approach was conducted concerning the Automated Lane Keeping System (ALKS, ECE/TRANS/WP.29/2020/81) approval proposal. Based on an Operational Design Domain (ODD, PAS 1883:2020) specification, the formal ALKS approval regulation and the vehicle model specification, requirements (e.g. reliability) are synthesized for ensuring fit-for-purpose simulation quality. Besides general traceability aspects, simulation quality arguments are finally provided via the goal structuring notation for assessor evaluations. The resulting continuum of purpose-driven and quality ensured simulation scenarios are vital for a scalable test coverage according to virtual validation test requirements – (outlook) which may also be used for triggering continuous integration and testing tool-chains subsequently.
11:20
conference time (CEST, Berlin)
Scenario-based Validation of Automated Driver Assistance Systems using Reliability Analysis Methods
26/10/2021 11:20 conference time (CEST, Berlin)
Room: A
Z. Kayatas, M. Rasch, P.T. Ubben (Mercedes-Benz AG, DEU); V. Bayer, S. Kunath, R. Niemeier (Dynardo GmbH, DEU)
Z. Kayatas, M. Rasch, P.T. Ubben (Mercedes-Benz AG, DEU); V. Bayer, S. Kunath, R. Niemeier (Dynardo GmbH, DEU)
One of the most important and current future trends in the automotive industry is the development of Advanced Driver Assistance Systems (ADAS). Due to the ever-increasing complexity of ADAS, the safety validation of such systems is a major challenge. New methods have to be developed, as the previous certification and approval methods are not suitable for this use case. E.g. Monte-Carlo simulation, combined with Software-in-the-Loop (SiL) simulation may help to overcome this limit. In order to achieve this goal, scenario-based testing for the safety validation of highly automated driving systems, where specific traffic scenarios are parametrized, simulated and analyzed by a set of criteria, is the only possible approach to efficiently test and validate thousands of concrete scenarios. By using distribution functions for each input parameter in this approach, a safety statement can be given by approximating the probability of failure for each traffic scenario. This is done by determining the transition between the safe and unsafe region in the parameter space. This process heavily relies on data from real-world traffic scenarios to derive the necessary scenario information for testing. Thus, we present a methodology based on a qualitative modelling of the searched scenarios, by using universal pattern elements of an ontology. Furthermore, depending on the identified traffic scenarios, distribution functions are generated. Additionally, correlations between the extracted parameters are considered by using the Nataf Transformation. Then, a combined criterion is introduced, to describe the criticality of a simulated traffic scenario. This enhances a clear separation between the safe and unsafe region. Finally, an efficient and robust search strategy is proposed by combining dimension reduction technics, surrogate models with Neural Networks and advanced methods of the reliability analysis (Importance Sampling using multiple Design Points). It can be shown that the number of simulations carried out can be significantly reduced in comparison to Monte-Carlo.
ADAS, Scenario-based validation, scenario identification, ontology, parameter extraction, robust search strategy, reliability analysis, Importance Sampling, Neural Networks
11:40
conference time (CEST, Berlin)
Integrative Simulation Architecture for Highly Automated Vehicles
26/10/2021 11:40 conference time (CEST, Berlin)
Room: A
C. Franke, F. Fischer (Prostep AG, DEU)
C. Franke, F. Fischer (Prostep AG, DEU)
Worldwide, initiatives strive for developing highly automated road vehicles of SAE levels 3 and above. Due to the complexity of the open world, requirements for those systems cannot be collected, engineered, and tracked the same way, as it has been established for ‘conventional’ products over the last decades. Correspondingly, highly automated vehicles cannot be verified and homologated ‘as usual’ solely by testing them against classical requirements using ‘conventional’ test catalogs. Instead, new approaches such as scenario-based approaches are needed to verify the vehicle’s correct behavior, especially under critical circumstances. Finding these critical scenarios is one challenge, which can be addressed by applying DOE-similar methods to parameterized families of scenarios. Another is investigating the vehicle’s behavior within such scenarios. [WW15] states that statistically some 1.000.000.000 km need to be driven to verify a highly automated vehicle by physical test for homologation. It is unclear, how many of these test kilometers would have to be repeated after each software iteration. Moreover, physical tests approach their technical limits when it comes to urban traffic scenarios involving pedestrians and other vulnerable road users (VRUs). Both renders classical physical test insufficient for verification and homologation of highly automated vehicles. The solution can only be virtual test, also known as simulation. The amount of virtually driven km scales with computing power, but does not require more engineers, if scenarios are generated automatically. This is the reason why scenario-based testing got the focus in recent years. A system enabling a vehicle to reach SAE levels 3 and above is highly sophisticated and requires highly developed systems engineering processes, capabilities, methods, tools, and even standards. We address these challenges by modern methods, which enable simulation-based decision making, analysis and verification. Methods are derived from model-based systems engineering (MBSE), and will guide us seamlessly all the way down from high level systems architecture to implementable simulation model prototypes. To reach this goal, we apply the RFLP-pattern to the system of the complete simulation, composed of not only the simulation model, but also the simulation software. Diversity and complexity of real traffic participants is tackled by a frame model with configurable components. This all is shown within the context of a credible simulation process, cf. [GH20, Hei21, HV20]. As modeling language, a subset of UML is used. This guarantees that the method is easy to understand and agnostic of the modeling tool. This new approach is explained using a practical application as example.
ADAS, credible simulation process, highly automated vehicle, MBSE, RFLP, SAE level, scenario-based testing, simulation-based decision making, traceability
×

[TITLE]

[LISTING

[ABSTRACT]

[DATE]

[ROOM]

[KEYWORDS]