The SAFEXPLAIN project opened the 8th edition of the Critical Automotive applications: Robustness & Safety (CARS) workshop on 8 April 2024, with a keynote talk, delivered by Ikerlan partner Jon Perez-Cerrolaza on insights on “AI, Safety and Explainability”. SAFEXPLAIN was further represented in the workshop in the session on Safety-critical system development, where Javier Fernandez, from partner IKERLAN, presented the consortium´s work on “AI-FSM: Towards Functional Safety Management for Artificial Intelligence-based Critical Systems”.

The workshop took place in Leuven, Belgium, in conjunction with the 2024 European Dependable Computing Conference (EDCC2024). The workshop provided an important forum for discussing key challenges and advancements in the construction of critical automotive applications and systems with researchers and practitioners.

Keynote presentation

Jon Perez-Cerrolaza gave the keynote talk on “Artificial Intelligence, Safety and Explainability”. The talk examined the potential for AI to drive the creation of autonomous safety-critical systems. It also addressed the challenges for integrating the latest AI technologies with existing safety engineering practices and safety standards to ensure trustworthiness.

The presentation summarized the most relevant contributions of the SAFEXPLAIN project concerning AI, safety and explainability. This included an overview of the Deep Learning (DL) Safety Archtiectural design patterns the project is pursuing and the differences in explainability between traditional safety engineering and AI/Machine Learning Safety engineering and its their effect on trustworthiness. A brief overview of the AI-Functional Safety Management methodology was presented, with more details presented in workshop session 2.

Workshop session on safety-critical system development

Following the keynote speech, the CARS WS delved deeper into the challenges faced by critical automotive applicatinos, namely 1) Risk analysis and assessment, 2) Safety-critical system development, and 3) Resilience and other issues.

SAFEXPLAIN participated in the safety-critical system development session. The presentation provided a deeper look into shared challenges for safety-critical systems. Javier Fernández, from partner IKERLAN, presented the SAFEXPLAIN consortium´s ongoing work on “AI-FSM: Towards Functional Safety Management for Artificial Intelligence-based Critical Systems”.

The presentation helped to contextualize the consortium´s work in this direction and highlighted the proposed lifecycle at the intersection of AI and functional safety. He expressed the gap between DL-based systems development and traditional safety development processes and the main phases affected when including AI&DL. He then presented the AI-FSM created within the framework of the project, based on recommendations from ISO 26262, IEC 61508, IEC TR 5469, EASA Conept Paper, AMLAS, ASPICE for ML.

For more information on the CARS workshop and participants, visit the workshop website here.