Specification and execution of system optimization processes with UML activity diagrams

Designing complex systems requires domain knowledge as well as tool-supported modeling and analysis techniques. General-purpose as well as domain-specific tools can be used for this task. The latter has the advantage of not requiring low-level model knowledge from systems designers that are domain experts, but is only possible with specialized tools that have to be programmed for a certain purpose or field by software engineers. The gap between general-purpose tools and domain-specific applications can be bridged by a (meta-)model-based description of structure and behavior of domain objects and the subsequent generation of a software tool. This approach has been successfully demonstrated in earlier work and the result is termed simulation-based application (SBA). One of the main applications of such tools is to find the best solution for design decisions, which can be done by manual evaluations of design ideas or automatically if parameters and design space are well understood. Such an automatic indirect optimization method should be adapted to the specific domain, which would require programming effort for the SBA. A logical extension to the usual system description is to apply the model-based paradigm to such a method description as well. This paper proposes an approach to model optimization processes (i.e., heuristics) graphically with UML activity diagrams that describe the data and control flow of such an algorithm. The resulting models are transformed into an executable algorithm automatically and control the work flow of an optimization with the software tool that has been designed for this task. An example of a heuristic optimization process for a wireless sensor network setup is presented.