Automatic deployment space exploration using refinement transformations

To manage the complex engineering information for real-time systems, the system under development may be modelled in a high-level architecture description language. This high-level information provides a basis for deployment space exploration as it can be used to generate a low-level implementation. During this deployment mapping many platform-dependent choices have to be made whose consequences cannot be easily predicted. In this paper we present an approach to the automatic exploration of the deployment space based on platform-based design. All possible solutions of a deployment step are generated using a refinement transformation. Non-conforming deployment alternatives are pruned as early as possible using simulation or analytical methods. We validate the feasibility of our approach by deploying part of an automotive power window optimized for its real-time behaviour using an AUTOSAR-like representation. First results are promising and show that the optimal solution can indeed be found efficiently with our approach.