SAFEXPLAIN’s success can be understood as a combination of outstanding scientific results and the vision to put them together to solve fundamental industrial challenges to make AI-based systems trustworthy. The project’s results and network of interested parties highlight the project’s significant progress in both scientific and industrial fronts.
On the scientific side, a recent SAFEXPLAIN publication authorted by BSC researchers has received an outstanding paper award at one of the top real-time conferences worldwide. The contribution highlights BSC’s work in the project providing solutions to guarantee the timeliness of safety-critical real-time tasks, such as those in autonomous automotive, space and railway applications.
Outstanding Paper Award for BSC researcher and SAFEXPLAIN contributor Blau Manau Corderroure at the 37 Euromicro Conference on Real-Time Systems
Similarly, a number of other publications stemming from the project have shown that traditional safety measures, such as diverse redundancy, can be smartly tailored in AI-based systems to improve prediction accuracy, fault tolerance and resilience against adversarial attacks, all at once, and with analogous cost to those solutions providing only fault tolerance.
On the application side, SAFEXPLAIN’s approach has proven to be industrially-ready and be of high interest for commercial application. Follow-up activities have already been set up with automotive partners, international organizations, and standardization workgroups.
The key behind the project’s successful strategy is its end-to-end approach (Figure 1). It starts from system conception until operation to enable the trustworthy use of AI in safety-critical systems and to develop practical implementations and their instantiation in automotive, space and railway use cases.
Figure 1: SAFEXPLAIN end-to-end approach
Overall, SAFEXPLAIN faces the very last months of its execution with a sense of success thanks to its unique end-to-end approach based on outstanding scientific solutions and its effective application to industrial use cases across multiple domains. Follow-up exploitation activities, as well as research actions are being carried out these days to guarantee that SAFEXPLAIN results will have a real short-time impact through direct industrial application, and mid/long-term impact through the research and evolutions of the project’s results in the scope of other European and national funded projects.
