Transforming UML state machines into stochastic Petri nets for energy consumption estimation of embedded systems

This paper presents an ongoing effort towards a methodology for the model-based engineering of energy-efficient automation systems. Energy consumption is an increasingly important decision criterion, which has to be included in the search for good architectural and design alternatives. In the paper, a modelling and performance evaluation technique is proposed, which describes an embedded system with an operational model of the processor hardware and an application model of the software. UML extended with MARTE profile elements is used for this part. Both models are transformed into a stochastic Petri net (SPN), for which we give transformation rules. It is then possible to predict the energy consumption of the hardware / software system by a standard evaluation of the Petri net. An example is provided.