With its implementing regulation (EU) 2022/1426 the European Union provides a basis for Level 4 (L4) Automated Driving (AD) homologation, but it lacks the necessary specificity for OEMs, Tier X suppliers and other industry stakeholders to effectively implement system homologation. In addition, the industry moves towards virtual validation and the respective homologation to enable a deep and broad scaling of simulations for diverse and dynamic driving environments.
Consequently, regulatory discussions were conducted to clarify and specify requirements. A basic understanding of the scope of the regulations in the European framework, the integration of the regulations into the overall virtual validation workflow, as well as the derivation of checklists for every section and highlighting of critical aspects was possible.
Subsequently, HAIViSH is developing a virtual validation methodology that integrates advanced system simulations and AI technologies to address the level of detail and potential gaps for the evaluation of AD/ADAS systems based on functional equivalence.
Establishing a comprehensive and consistent understanding of the regulation is essential for creating standards and methodologies that end users can rely on in the long term. Accordingly, based on the integration in the overall workflow of virtual validation as well as the highlighted checklists and critical points, overarching and well-founded discussions can follow. The objective is to establish confidence in the vendors applications and, over time, in autonomous vehicles as well.
Understanding the regulation and integrating this into the overall virtual validation workflow, highlighting the necessary actions and requirements in respective pipelines, enables vendors to implement pipelines and assign regulation responsibilities to subgroups. Therefore, a possible agile workflow to be aligned with the regulation and virtual validation was presented.
Building on this, the derived requirements were categorized and summarized. Key outcomes include checklists for each section, as well as the identification and highlighting of critical aspects. Consequently, vendors are able to assess their current processes and focus on critical aspects.
A persistent challenge lies in comprehending the repercussions of regulatory formulations on technical solutions. Consequently, we put increased effort to discuss questions and open points in interactive workshops. This addresses the second challenge of establishing contact with industrial partners. A common understanding depends on the incorporation of industrial partners, which was conservative at first but improved along the workshops. Current efforts aim to showcase first implementations as result of WP4 to increase understanding and visualize first outcomes.
To facilitate practice-based discussions, a radar sensor model will be developed and validated using measurement data. In case of deviation between model and reality, a structured method for handling imperfectly matching models will be established.
The potential of AI methods in virtual validation will be analyzed by categorizing application areas and highlight explainability. We will provide recommendations regarding these methodologies.
Finally, we will prepare next steps for integration of AI approaches in virtual validation to further address 2022/1426 in development.
Due to its integration within the prostep framework, implementing and launching HAIViSH presented certain challenges. However, we were able to reach an operational level quickly. Building on this foundation, we have established, refined, and optimized the project's strategic direction and roadmap for 2025, with the objective to attract and integrate industrial partners in continuing discussions and workshops in 2026. In particular with regards to known AI applications, we are confident that HAIViSH will showcase their integration potential and capture interdisciplinary feedback and discourse with industrial partners.
Regulation “act” by its proverbial definition means “do something”. – Jürgen Pannek