Defining the design space Sample Clauses

Defining the design space. To achieve optimisation, an EAST-ADL design model - usually at the analysis and/or design levels, since sufficient data for evaluation must also be present - must be created that contains a number of different variability points and that allows for different design variants to be explored. Existing variability mechanisms proved to be largely sufficient for this task. One of the primary concerns for optimisation is that the designer must still ensure that the variants are substitutable, i.e. that one variant is functionally equivalent to another; this is not always the case in normal variability management, where e.g. a feature may be present in one product line but absent in another. Thus is it still important to distinguish between variations that represent such optional functionality (a.k.a. “product line variability”, that does not define an optimization choice) and variations represent alternative but functionally equivalent realizations of a feature (a.k.a. “design space variability”, that defines the optimization space). This was achieved through the use of the binding time attribute. Furthermore, so variants can be automatically explored by the optimisation algorithm without unnecessary complexity, they need to be encoded in a hierarchical, structured way. In practice one major difficulty in automatic optimisation is the connections between substituted and/or replicated components, e.g. one implementation may not have the same interface (i.e. number and type of inputs & outputs) as another, and this has to be resolved for automatic optimisation to produce valid and sensible results. In EAST-ADL, the variability mehanisms already provide this capability via the Feature Model (see next section), which serves as a kind of hierarchical index of the different variability points and relationships between them. The feature model (or feature tree) is similar to the hierarchical encoding envisaged for the optimisation and broadly speaking can be reused for this purpose. Finally, although the original design model used as input to the optimisation process should contain the necessary variability to define the design space (and product space with take rates), the design solutions identified by the optimisation need to have had this variability resolved such that the solution represents a single possible configuration of the system. This is similar to the concept of binding in variability and allows the optimisation process to be able to subject the resolved mo...