In this paper, we present a technique for incremental and commutative composition of state-machine models of features, using the Feature House framework. The inputs to Feature House are feature state-machines (or state-machine fragments) modelled in a feature-oriented requirement modelling language called FORML and the outputs are two state-machine models: (1) a model of the whole product line with optional features guarded by presence conditions, this model is suitable for family-based analysis of the product line, and (2) an intermediate model of composition that facilitates incremental composition of future features. We discuss the challenges and benefits of our approach and our implementation in the Feature House.

A software product line (SPL) is often constructed as a set of features, such that individual products can be assembled from a set of common features and a selection of optional features. Although features are conceptualized, developed, and evolved as separate concerns, it is often the case that, in practice, they interfere with each other -- called a feature interaction. In this paper, we precisely define what it means for one feature to have a behaviour interaction with another feature, where the behaviour of one feature is affected by the presence of another feature. Specifically, we use a form of bisimilarity to define when the behaviour of a feature in isolation differs from its behaviour in the presence of an interacting feature. We also consider the case where features are modelled in a language that allows the specification of intended interactions, and we adapt our use of bisimilarity to provide a formal definition for unintended behaviour interactions.

In this paper, we present a feature-oriented requirements modelling language (FORML) for modelling the behavioural requirements of a software product line. FORML aims to support feature modularity and precise requirements modelling, and to ease the task of adding new features to a set of existing requirements. In particular, FORML decomposes a product line's requirements into feature modules, and provides language support for specifying tightly-coupled features as model fragments that extend and override existing feature modules. We discuss how decisions in the design of FORML affect the evolvability of requirements models, and explicate the specification of intended interactions among related features. We applied FORML to the specification of two feature sets, automotive and telephony, and we discuss how well the case studies exercised the language and how the requirements models evolved over the course of the case studies.

Our goal is to develop a method for creating models of functional software requirements in which features are explicit, with a focus on the automotive software domain. In the current state of our proposed method, feature requirements are modelled using state machines that describe changes to a shared domain model.