A model-driven software construction approach for cyber-physical systems

Systems and software engineers realize innovative functionality in cyber-physical systems by connecting previously independent systems. The development of these systems becomes more challenging because the systems’ complexity, the amount of the safety-critical software, the heterogeneity of the underlying platforms, and the connectivity of the systems increase steadily. However, existing model-driven approaches do not cover the different engineers’ concerns efficiently during the different development tasks. Many of these approaches have a focus on the formal design and the verification of safety and real-time properties. Though, the question how to retain the systems’ safety during following construction tasks remains open. This thesis focuses on the construction tasks: physical system integration validation, resource allocation, and the implementation. It presents a highly automated end-to-end modeldriven approach for the development of interacting cyber-physical systems that are realized on heterogeneous, distributed platforms. The approach covers the validation of the system integration via model-in-the-loop simulation, constraint-driven software to hardware allocation, and generative software construction. Consequently, the approach retains the system’s safety with respect to systematic errors through different development tasks. It enables engineers to detect and avoid design and implementation flaws. Thereby, it improves the reliability of systems during their operation. It enables an effective and efficient seamless development. This thesis evaluates the approach by providing an integrated development environment and conducting several case studies from the automotive and automation domain. The evaluation shows the effective and in comparison to state of the art approaches more efficient application of: (1) the model-in-the-loop simulation in the context of virtual prototypes, (2) the allocation constraint specification and the automated allocation planning, and (3) the generation of executables for heterogeneous, distributed embedded platforms.