Variability Management on Behavioral Models

This paper deals with managing variability on behavioral models. Such models are generally more complex, less tractable by hand than the static, structural parts of a system description. This calls for specific support to check the consistency of variability expression model-wide: defining some elements as variable may impact several behavioral constructs, even elements that designers may not be aware of – e.g. events attached to triggers. Moreover, variable elements at the structural level usually imply variable behavioral constructs in ways that are not easily foreseeable: a seemingly perfectly valid variability scheme may lead to ill-formed behavioral models after derivation. This paper takes UML state machine diagrams as a case study and presents some technical solutions to maintain consistency between both levels: a propagation mechanism to deal with the impact of variability model-wide and a constructive method to check the well-formedness of state machines obtained