Providing good mobility solutions gets more and more challenging in times of requirements for carbon reduction, aging society, exploding cities and growing town-country discrepancies. AI has the potential to sustain or even improve future mobility solutions under these constraints. However, AI has the challenge that many things are not completely understood today, especially that there is no agreed method to carefully and exhaustively verify AI-based solutions. AI faces also the challenge that nm-technologies used to build AI-hardware come up with new faults and thus new challenges in fault analysis. We propose in this talk a physical and/or virtual cage around AI enabled autonomous mobility as the standard method to support AI in 26262. Further, a process of releasing the constraints of such cages in a yearly revision of the standard to keep track with the rapidly evolving findings in AI-research.
The browser you are using is not supported that will prevent you from accessing certain features of the website. We want you to have the best possible experience. For this you'll need to use a supported browser and upgrade to the latest version.