Cyber-physical Systems are distributed, embedded systems that interact with their physical environment. Typically, these systems consist of several Electronic Control Units using multiple processing cores for the execution. Many systems are applied in safety-critical contexts and have to fulfill hard real-time requirements. The model-driven engineering paradigm enables system developers to consider all requirements in a systematical manner. In the software design phase, they prove the fulfillment of the requirements using model checking. When deploying the software to the executing platform, one important task is to ensure that the runtime scheduling does not violate the verified requirements by neglecting the model checking assumptions. Current model-driven approaches do not consider the problem of deriving feasible execution schedules for embedded multi-core platforms respecting hard real-time requirements. This paper extends the previous work on providing an approach for a semi-automatic synthesis of behavioral models into a deterministic real-time scheduling. We add an approach for the partitioning and mapping development tasks. This extended approach enables the utilization of parallel resources within a single ECU considering the verification assumptions by extending the open tool platform App4mc. We evaluate our approach using an example of a distributed automotive system with hard real-time requirements specified with the MechatronicUML method.

Modern Cyber-physical Systems are executed in physical environments and distributed over several Electronic Control Units using multiple cores for execution. These systems perform safety-critical tasks and, therefore, have to fulfill hard real-time requirements. To face these requirements systematically, system engineers de- velop these systems model-driven and prove the fulfillment of these requirements via model checking. It is important to ensure that the runtime scheduling does not violate the verified requirements by neglecting the model checking assumptions. Currently, there is a gap in the process for model-driven approaches to derive a feasible runtime scheduling that respects these assumptions. In this paper, we present an approach for a semi- automatic synthesis of behavioral models into a deterministic scheduling that respects real-time requirements at runtime. We evaluate our approach using an example of a distributed automotive system with hard real-time requirements specified with the MechatronicUML method.

Die Komplexität von mechatronischen Systemen wird stetig größer. MechatronicUML (MUML) ist eine Methode zur Entwicklung für Software von mechatronischen Systemen. Im Rahmen einer Bachelorarbeit wurde ein bestehender Quelltextgenerator für MUML-Modelle so erweitert, dass Quelltext für ein Echtzeitbetriebssystem generiert werden kann, welches auf einem LEGO Mindstorms-Roboter installiert ist.