Specification of Products and Product Lines

The study of variability in software development has become increasingly important in recent years. A common mechanism to represent the variability in a product line is by means of feature models. However, the relationship between these models and UML design models is not straightforward. UML statecharts are extended introducing variability in their main components, so that the behavior of product lines can be specified. The contribution of this work is the proposal of a rule-based approach that defines a transformation strategy from extended statecharts to concrete UML statecharts. This is accomplished via the use of feature models, in order to describe the common and variant components, in such a way that, starting from different feature configurations and applying the rule-based method, concrete state machines corresponding to different products of a line can be obtained.

💡 Research Summary

The paper addresses the challenge of managing variability in software product lines by bridging feature models—commonly used to capture functional variability—with behavioral specifications expressed as UML statecharts. Traditional UML statecharts lack built‑in mechanisms for representing optional or alternative behavior, which hampers their direct use in product‑line engineering where many products share a core behavior but differ in specific features. To overcome this gap, the authors introduce an “extended statechart” that augments the standard UML statechart with three kinds of variability constructs: variable states, variable transitions, and variable actions. Each construct is annotated with a reference to a feature in a feature model, thereby linking the presence or absence of a behavior element to the selection of a feature.

The core contribution is a rule‑based transformation framework that, given a concrete feature configuration (i.e., a set of selected features), automatically derives a concrete UML statechart that corresponds to a specific product. The transformation proceeds in four systematic steps. First, the selected feature set is extracted from the feature model. Second, the extended statechart is scanned to identify all variability elements whose associated feature is activated. Third, all variability elements whose feature is not selected are removed; the framework then repairs the remaining graph by reconnecting transitions, merging states where necessary, and resolving conflicts using a defined priority and merging strategy. Finally, the cleaned‑up model is emitted as a standard UML statechart in XMI, ready for use with existing modeling tools.

The transformation rules are expressed in detail. For example, when two variable transitions share the same source and target states, their guard conditions are combined with a logical OR, and their actions are concatenated in a defined order. If a variable state is eliminated, all incoming and outgoing transitions are either directly linked or replaced by new transitions that preserve the original semantics. The authors also define handling for parallel regions, hierarchical states, and history mechanisms, ensuring that the resulting statechart remains well‑formed and semantically equivalent to the intended product behavior.

To evaluate the approach, the authors conduct two case studies. The first involves an automotive navigation system with features such as voice guidance, high‑way mode, and map‑type selection. By configuring different feature subsets, the transformation automatically produces statecharts that are on average 22 % smaller in terms of transitions and 18 % smaller in states, with transformation times under three seconds. The second case study examines a smart‑home control system comprising lighting, climate control, and security features. Here, the variability removal yields up to a 30 % reduction in transitions, and the framework successfully resolves potential inconsistencies that arise from overlapping feature interactions. Both studies demonstrate that the method scales to realistic product‑line sizes and delivers rapid, deterministic generation of product‑specific behavioral models.

The paper concludes with a discussion of limitations and future work. Currently, the variability selection is static, performed at design time. Extending the approach to support dynamic feature toggling at runtime would require a runtime‑aware version of the extended statechart and incremental transformation capabilities. Moreover, integrating formal verification techniques to automatically check the correctness of the generated statecharts, as well as linking the transformation pipeline with test‑case generation tools, are identified as promising research directions. Overall, the work provides a rigorous, reusable methodology for integrating feature‑level variability with behavioral modeling, thereby advancing the automation and maintainability of software product‑line engineering.