Predictable timing behavior of gracefully degrading automotive systems

Abstract Fail-operational behavior of safety-critical software for autonomous driving is essential as there no driver available a backup solution. In failure scenario, tasks can be restarted on other hardware resources. Here, graceful degradation used cost-efficient solution where resources are redistributed from non-critical to at run-time. We allow actively use that reserved critical tasks, which would otherwise unused and only required in scenario. However, such it paramount importance achieve predictable timing applications safe operation. has ensured even after the restart guarantee execution times given. this paper, we propose approach using composable scheduling. our present, first time, performance analysis able analyze constraints fail-operational distributed degradation. Our method verify during Electronic Control Unit failure, always adheres end-to-end constraints. Furthermore, present dynamic decentralized mapping procedure performs constraint solving run-time analytical combined with backtracking algorithm. evaluate by comparing success rates state-of-the-art approaches active redundancy an based resource availability. experimental setup fit about double number same architecture compared approach. Combined, enable, gracefully degrading automotive systems solutions applications.