Halbzeit beim viel beachteten Verification-Validation-Methods (VVM)-Projekt: Mitte März präsentierten die Projektpartner erste Ergebnisse und Methoden. Die Relevanz des Projekts sei hoch, betonten Vertreter aus Politik, Industrie und Forschung, denn sicheres autonomes Fahren werde nur dann möglich, wenn verlässliche Validierungs- und Verifikationsmethoden eingesetzt werden. dSPACE bringt in das Projekt neben umfangreicher Expertise in der SIL- und HIL-Simulation zur Analyse und Bewertung der Projektanforderungen prototypische Umsetzungen zur Demonstration der Methoden ein.

Results of three main areas were presented in the half-time presentation: The requirements for the test methods, orchestration and validation of the test infrastructures, and the data flow as well as the tools used in the project. dSPACE successfully tested the methods developed in the project for practicality with two demonstrations on criticality analysis and sensor model validation, among other things.

Criticality Analysis

Due to the virtually unlimited number of different road traffic situations, moving real tests to simulation alone is not enough to be able to handle the sheer infinite number of test kilometers. A central task for limiting the parameter range to the relevant cases is the criticality analysis. By systematically identifying causalities that lead to critical situations on the road (called criticality phenomena), the test scope is reduced to the relevant test cases and thus to a manageable level. The highly scalable scenario-based testing plays a particularly important role. While accident databases and expert knowledge formed the basis for the project, simulation was used very early on to effectively complement real data and specifically generate evidence for the refinement of the criticality phenomena. A causality graph shows the factors that influence the criticality phenomenon and their relationships between each other, allowing for their analysis before they are deliberately varied and observed in the simulation. A demonstration that analyzed the particularly easy-to-observe criticality phenomenon “reduced friction value” was implemented in dSPACE SIMPHERA by means of a use case that was defined in one way and used throughout the project. The KPI-based evaluation makes it easy to filter for critical simulation runs and analyze them a second time – this time in more detail.

Eine Demonstration zur Untersuchung des anschaulichen Kritikalitätsphänomens „verminderter Reibwert“ wurde anhand eines einheitlich definierten Use Cases mit SIMPHERA implementiert.

Sensormodellvalidierung

To systematically move tests from reality to simulation, their limits of application must be known to build trust in the models used. For this, the simulation and all used models first have to be validated for the use case and for different samples from the possible parameter set. The method that was designed in the first project half was demonstrated using the model specification of the lidar sensor simulation, based on the general simulation requirements. For this, the chain of effects in the sensor system was methodically prepared and tested for relevance. From this, the project team derived requirements for the lidar system simulation and implemented them in the dSPACE tool chain. For the subsequent demonstration, a lidar sensor model provided by the SET Level project was integrated into the dSPACE sensor environment simulation with its powerful ray tracing engine via open interfaces. A particular focus was on validated material parameters of the virtual environment, which simulate the optical reflection behavior of objects with a high degree of realism. This is what makes model validation based on echo pulse width at the point cloud level possible. The second half of the project will now focus on refining the demonstrated results and implementing more concept demonstrators. While the criticality analysis and sensor model validation will remain as important as before, test orchestration, i.e., the distribution of test cases to the most suitable test equipment, and the required continuity of the test infrastructure between SIL and HIL will be another focal point of dSPACE’s efforts in the project.

VVM

Das Projekt VVM hat zum Ziel, Testverfahren zu entwickeln und Systematiken sowie Methoden bereitzustellen, um den Sicherheitsnachweis für automatisierte Fahrzeuge zu führen. Das Projekt arbeitet am Use Case der urbanen Kreuzung und fokussiert sich auf Fahrfunktionen bis zur kompletten Automatisierung von Fahrzeugen (SAE Level 4 und 5). Im Projekt soll der Sicherheitsnachweis als integraler Bestandteil im Entwicklungsprozess integriert werden. Bereits in der Entwicklung von Komponenten und Subsystemen wird die Testbarkeit, die Verifikation und Validierung als Designziel betrachtet („Design für Testbarkeit“). Die entstehenden Systeme lassen sich dann hierarchisch testen, so dass bei Aktualisierungen von Komponenten nicht mehr das gesamte System neu getestet werden muss. dSPACE ist einer von 23 Projektpartnern. Das Projekt läuft bis Mitte 2023.

Newsletter abonnieren

Abonnieren Sie unsere Newsletter, oder verwalten oder löschen Sie Ihre Abonnements