Compiling business process models for sensor networks

Wireless sensor networks are increasingly being used to improve business processes. The behavior of such a process is usually captured in models while its implementation is typically created manually. Besides being expensive, this approach entails deviations between the model and the implementation of a business process, thus rendering it less effective. We aim at closing this gap by automatically generating applications from the model of the process instead. In our approach, software developers provide building blocks that are used by business analysts to model business processes and hereby effectively control their execution on an abstract level. Furthermore, the model editor integrates a compiler and a simulator so that the business analysts can test and debug the modeled processes on the same level of abstraction. Our evaluation results show that the generated code can be executed on resource-constrained sensor nodes consuming only 1% more energy than the hand-written equivalents. However, the benefits of our approach come at the price of 10% more RAM and 44% more flash space consumption on average.