Rethinking Safety in Autonomous Ecosystems

Abstract

As autonomous cyber-physical systems are responding to the dynamism of our hyper-connected digital world, they are forming so called dynamic autonomous ecosystems, which require a change in methods ensuring their safe behavior. Within this change, reactions to predictable scenarios need to be replaced with adaptability to the unpredictable context, with gradual safety mechanisms, able to decide whether or not to trigger a certain mitigation procedure. In this paper, we outline our vision towards evolution of safety mechanisms to support dynamic and self-adaptive architectures of autonomous ecosystems. We are proposing an approach to address this research problem with the help of trust and reputation combined with gradual adaptation of safety procedures at runtime.