Efficient Constraint Handling During Designing Reliable Automotive Real-time Systems

In modern embedded systems, e.g. avionics and automotive, it is not unusual for there to be between 40 and 100 processors with a great deal of the software having hard real-time requirements and constraints over how, when and where they execute. The requirements and constraints are essential to the overall systems dependability and safety (e.g. to ensure replicas execute on different hardware). This leads to a complex design space exploration (DSE) problem which cannot be practically solved manually especially if the schedule is to be maintained. In this paper it is shown that dealing with the constraints using a conventional state of the art System Configuration Algorithm is less efficient, less effective and does not scale well. This issue can be improved by performing constraint pre-processing as well as constraint encoding. It is shown that our approach can handle typical industrial requirements that come from the automotive industry's AUTOSAR standard in an efficient way.