In adaptive systems, it is necessary to observe the system and react if the system enters an undesirable state or shows
unwanted behaviour. For this purpose, feedback loops, operationalised as observer/controllers can be employed. This practice describes an automatic
transformation to observer/controller implementations from constraints defined during requirements analysis.
A prerequisite of this practice is that constraints have been captured during requirements analysis. Ideally, these are
expressed as OCL-contraints that define the correct states of the system. If the  Define System Limitations and Constraints task from the practice Goal-driven Requirements Elicitation is performed, constraints should be
available. At the same time, this ensures that a domain model containing the elements the constraints are defined on is
available. Constraints can also be defined in a specialised document separate from the requirements model.
The process can be repeated after the requirements or the domain model have changed, according to a model-driven design
(MDD) approach. Changed parts of the system models and implementation will be re-generated while existing models and
code are preserved.
The basic approach is described in Steghöfer et al. - Synthesis of Observers for Autonomic Evolutionary Systems from Requirements Models.
|
