Showcasing Project Results in 2 HiPEAC’25 Workshops

Date: February 04, 2025

The 2025 HiPEAC conference successfully brought together more than 750 experts in in computer architecture, programming models, compilers and operating systems for general-purpose, embedded and cyber-physical systems.  This annual event is a premier forum for exploring challenges and innovations in safety-critical dependencies, cybersecurity, energy efficiency and machine learning.

SAFEXPLAIN played in active role throughout the conference with key presentations in 2 workshops and presence in the Barcelona Supercomputing Center booth.

Women@HPC MAR WHPC chapter: Building the diversity continuum in cutting-edge technologies

As part of the workshop, Women@HPC Mar WHPC chapter: Building the diversity continuum in cutting-edge technologies, held on 21 January 2025, IKERLAN’s Irune Yarza presented “Ensuring Functional Safety in AI-based critical systems: A railway case study”. She discussed key concerns regarding functional safety in AI-based critical systems, such as what standards should be followed, how to make AI explainable and if we can make AI safe. She followed this with project results, specifically focusing on the rail case.

Get the presentation here.

MCS: Mixed Critical Systems – Safe Intelligent CPS and the development cycle

21 January 2025

From session: Safe Artificial Intelligent Dependable Autonomous Systems

Robert Lowe from partner RISE delivered a presentation on “SAFEXPLAIN: Addressing uncertainty in software develpment/deployment when integrating AI within safety critical autonomous systems”.  The presentation discused AI for functional safety management, XAI for supporting AI-FSM and minitigating uncertainty in deployable AI systems

Access the presentation here.

From session: Safe Artificial Intelligent Dependable Autonomous Systems II

Exida development partners Carlo Donzella and Stefano Lodico presented “SOTIF-compliant V&V (Verification & Validation) and STPA (Systems Theoretic Process Analysis) for critical systems with AI-based components”. They discussed the safety lifecycle for DL-software, direct mapping to relevant SOTIF clauses, and the application of “traditional” functional safety standards to DL. Moreover they compared SOTIF vs ISO 26262 and validation and verification strategies.

Get the slides here.

Barcelona Supercomputing Center Booth

SAFEXPLAIN partners from the Barcelona Supercomputing Center were also on hand to share information about the project and its results at its booth. The project flyer and video were distributed to interested participants.