The development of modern embedded systems, such as autonomous air and ground vehicles, is often subject to a complex set of requirements. The seamless integration into large-scale networks, a steadily increasing demand for computational performance, and the application of non-deterministic Artificial Intelligence (AI) algorithms are three of the many challenges to be tackled by embedded system developers. At the same time, they need to ensure a sufficient degree of safety and security. To reduce development costs, minimise the risk of catastrophic failures, and prevent unauthorised system access, automated approaches for the systematic consideration of these requirements are of major importance.
The goal of the collaborative research project XANDAR was to research such approaches, combine them into a prototypical design framework, and evaluate their applicability to real-world scenarios. The framework consists of two major components: a model-based toolchain for embedded software systems and a self-healing runtime architecture for modern System-on-Chip (SoC) platforms. The toolchain relies on the X-by-Construction (XbC) paradigm, which is a strategy to auto-generate system implementations that exhibit well-defined runtime properties.
At the end of the project, two practical use cases were employed to evaluate the capabilities of the resulting framework: (1) a sensor fusion pipeline for automated road vehicles provided by the BMW Group and (2) a pilot assistance system provided by the German Aerospace Center. Using the XANDAR framework, the consortium was able to address the requirements of both use cases successfully.
In summary, XANDAR’s ambitious project mission has led to numerous novel approaches, methodologies, and tools for the development of software in autonomous and distributed embedded systems. These results pave the way for follow-up innovations in the form of a commercially viable XbC toolchain and applied research building up the proposed concepts.
