Experimenting with Evolutionary Algorithms to Reduce Feature Model Configuration Steps

Feature models are frequently used to capture the knowledge about configurable software systems and product lines. However, feature modeling of large-scale systems is challenging as many models are needed for diverse purposes. For instance, feature models can be used to reflect the perspectives of product management, technical solution architecture, or product configuration. Furthermore, models are required at different levels of granularity. Although numerous approaches and tools are available, it remains hard to define the purpose, scope, and granularity of feature models. In this paper we thus present experiences of developing feature models for two large-scale industrial automation software systems. Specifically, we extended an existing feature modeling tool to support models for different purposes and at multiple levels. We report results on the characteristics and modularity of the feature models, including metrics about model dependencies. We further discuss lessons learned during the modeling process.

Creating products according to own needs is omnipresent on the Internet. We choose the properties we want to have in a product from a product configurator, such as a car or printer configurator. The underlying models from the configurators are hidden. The migration of legacy configurators into feature models is meaningful to use the advantages of software configurations for product configurators. We introduce an abstraction level and assumptions for product configurators to readout information needed for a feature model. Furthermore, we develop an algorithm called FeatureDiagramSynthesizer. The algorithm consists of two parts which are finding features and adding constraints. For the algorithm, different strategies are used to create the feature model from a product configurator. Moreover, we implement the extensible algorithm FeatureDiagramSynthesizer and a prototype for the product configurator from Ricoh. Based on our developed algorithm, it is only basically possible to readout a feature models from product configurators which fulfill our assumptions.

Feature models are the de facto standard for modelling variabilities and commonalities in features and relationships in software product lines. They are the base artefacts in many engineering activities, such as product configuration, derivation, or testing. Concrete models in different domains exist; however, many are in private or sparse repositories or belong to discontinued projects. The dispersion of knowledge of feature models hinders the study and reuse of these artefacts in different studies. The Universal Variability Language (UVL) is a community effort textual feature model language that promotes a common way of serializing feature models independently of concrete tools. Open science principles promote transparency, accessibility, and collaboration in scientific research. Although some attempts exist to promote feature model sharing, the existing solutions lack open science principles by design. In addition, existing and public feature models are described using formats not always supported by current tools. This paper presents ▪ , a repository of feature models in UVL format. ▪ provides a front end that facilitates the search, upload, storage, and management of feature model datasets, improving the capabilities of discontinued proposals. Furthermore, the tool communicates with Zenodo – one of the most well-known open science repositories – providing a permanent save of datasets and following open science principles. ▪ includes existing datasets and is readily available to include new data and functionalities in the future. It is maintained by three active universities in variability modelling.

Feature models have been used since the 90's to describe software product lines as a way of reusing common parts in a family of software systems. In 2010, a systematic literature review was published summarizing the advances and settling the basis of the area of Automated Analysis of Feature Models (AAFM). From then on, different studies have applied the AAFM in different domains.AAFM has been applied in different activities along the Software Product Line (SPL) process such as product configuration and derivation, reverse engineering or SPL testing. As the field evolves, there is a need to evaluate the trends in the area and discover where the AAFM is being applied. Systematic Literature Reviews (SLRs) and Systematic Mapping Study (SMS) are the main techniques used to crawl the knowledge in a scientific area and candidates to discover the aforementioned tendencies. While SLRs are suitable to summarize the state of a research area by providing mostly qualitative information, SMSs focus on providing quantitative information and a categorization of the corpus that enables the identification of trends and research opportunities.We present a SMS to identify the evolution and trends in the application of the AAFM since 2010. Concretely, we have performed a search on different databases of AAFM-related papers. We selected 423 primary sources (papers) that followed the defined inclusion and exclusion criteria. The primary sources were classified according to different variability facets that were found during the reading and key-wording phase. It is important to remark that before 2010, AAFM was not well defined and it was referenced using an amalgam of names and concepts. Therefore, we consider that in 2010 the concept of AAFM was coined and then used in different domains and scenarios. This paper studies how AAFM has been used since its definition.First, we found six different variability facets where the AAFM is being applied that define the tendencies: product configuration and derivation; testing and evolution; reverse engineering; multi-model variability-analysis; variability modelling and variability-intensive systems. We also confirmed that there is a lack of industrial evidence in most of the cases. Finally, we present where and when the papers have been published and who are the authors and institutions that are contributing to the field. We observed that the maturity is proven by the increment in the number of journals published along the years as well as the diversity of conferences and workshops where papers are published.We also observed that there are only a few industrial and real evidences of the application of AAFM techniques in most of the cases. We detect in detail where and when the papers have been published and who are the authors and institutions that are contributing to the field. We saw that the maturity is proven by the increment in the number of journals published along the years as well as the diversity of conferences and workshops where papers are presented. Finally, we devise some research opportunities and applications in the future as well as synergies with other research areas. We also suggest some synergies with other areas such as cloud or mobile computing among others that can motivate further research in the future.The reader can find the full text of this paper at https://doi.org/10.1007/s00607-018-0646-1

Software product lines (SPLs) are families of software systems sharing common assets and exhibiting variabilities specific to each product member of the family. Commonalities and variabilities are often represented as features organized in a feature model. Due to combinatorial explosion of the number of products induced by possible features combinations, exhaustive testing of SPLs is intractable. Therefore, sampling and prioritization techniques have been proposed to generate sorted lists of products based on coverage criteria or weights assigned to features. Solely based on the feature model, these techniques do not take into account behavioural usage of such products as a source of prioritization. In this paper, we assess the feasibility of integrating usage models into the testing process to derive statistical testing approaches for SPLs. Usage models are given as Markov chains, enabling prioritization of probable/rare behaviours. We used featured transition systems, compactly modelling variability and behaviour for SPLs, to determine which products are realizing prioritized behaviours. Statistical prioritization can achieve a significant reduction in the state space, and modelling efforts can be rewarded by better automation. In particular, we used MaTeLo, a statistical test cases generation suite developed at ALL4TEC. We assess feasibility criteria on two systems: Claroline, a configurable course management system, and Sferion™, an embedded system providing helicopter landing assistance.

Automating feature model refactoring: A Model transformation approach

Feature models are a de facto standard for representing the commonalities and variability of product lines and configurable software systems. Requirements-level features are commonly implemented in multiple source code artifacts, which results in complex dependencies at the code level. As developers change and evolve features frequently, it is challenging to keep feature models consistent with their implementation. We thus present an approach combining feature-to-code mappings and code dependency analyses to inform engineers about possible inconsistencies. Our focus is on code-level changes requiring updates in feature dependencies and constraints. Our approach uses static code analysis and a variation control system to lift complex code-level dependencies to feature models. We present the suggested dependencies to the engineer in two ways: directly as links between features in a feature model and as a heatmap visualizing the dependency changes of all features in a model. We present results of an evaluation on the Pick-and-Place Unit system, which demonstrates the utility and performance of our approach and the quality of the suggestions.

The traditional method of designing the initial configuration of off-axis reflective optical system is to first obtain the initial configuration of coaxial reflective optical system, and then achieve the unobscured design with an offset aperture stop or a biased input field, or both. Because the aberration distribution of coaxial reflective optical system is not applicable to the off-axis reflective optical system, the obtained unobscured off-axis reflective optical system has large aberration, and the unobscured design process is complicated. In this paper we present a method of designing an initial configuration of off-axis reflective optical system based on vector aberration theory. With this design method, a good unobscured initial configuration of off-axis reflective optical system can be directly obtained by using an offset aperture stop or a biased input field, or both. Based on the vector aberration theory and gaussian brackets, the third-order aberration coefficient is derived for off-axis reflective optical system. Initial configuration performance is important for optical design, especially for the complicated optical system design. The selection of initial configuration highly affects the final system imaging performance, fabrication difficulty and alignment difficulty. An error function is established to evaluate the performance of off-axis reflective optical system, and it consists of aberration coefficients and other constraints. The genetic algorithm is a highly parallel, random and adaptive global optimization algorithm. To obtain a good initial configuration for the off-axis reflective optical system, the genetic algorithm is used to search for the initial configuration with minimum residual aberration. This method can obtain a good initial configuration of off-axis reflective optical system for further optimization. The benefit of this design method is demonstrated by designing an off-axis three-mirror optical system. For the focal plane array, a long-wave infrared off-axis three-mirror optical system is designed. A good initial configuration is obtained with the proposed method, which achieves the unobscured design by using an offset aperture stop and a biased input field. To improve the performance of initial configuration, the obtained initial configuration is optimized with the optical design software. The designed optical system has good imaging quality. As the mirrors are free from the tilts and decenters, the designed optical system is aligned easily.

Feature models are widely used for describing the variability of a software product line. A feature model contains a tree of features and a set of constraints over these features, which define valid feature combinations. In the industrial practice, large feature models containing hundreds of features and constraints are common. Furthermore, in a hierarchical product line a feature model can be related to other feature models through inter-model constraints. Due to the model size and complexity, understanding industrial feature models is a challenging task.In this paper, we describe the feature model understanding challenges reported by feature model developers at Robert Bosch GmbH. To support the developers in model understanding, we extend the idea of a feature implication graph to feature relation graph by abstracting groups of implications to feature relations. A transitively closed relation graph shows all modeled and implicit feature relations and spans all related feature models. The graph is also used to identify modeling problems, such as false optional or dead features, and to show the derivation of any implicit relation or problem from the modeled constraints. In a case study at Bosch, we evaluate the use of feature relation graph for model understanding. We propose further use cases of the graph, supporting model maintenance, evolution and configuration.

Since the mid-2000s, Propositional Logic (PL) has been the de facto language to express constraints in Feature Models (FMs) of Software Product Line (SPLs). PL was adequate because product configurations were formed by binary decisions including or not including features in a product. Inspired by both prior research and practical systems (eg., SPLs that use KConfig), future FMs must go beyond PL and admit numerical (and maybe even text) variables and their constraints.The Object Constraint Language (OCL) is a general-purpose declarative constraint language for Model Driven Engineering (MDE), which admits virtually any kind of variable and constraint in metamodels. We should expect future FMs to be examples of MDE metamodels. This raises a basic question: Should OCL be used to express constraints of future variability modeling language(s)?In this talk, I outline the pros and cons for doing so.

Augmented Reality (AR) has recently found high attention in mobile shopping apps such as in domains like furniture or decoration. Here, the developers of the apps focus on the positioning of atomic 3D objects in the physical environment. With this focus, they neglect the configuration of multi-faceted 3D object composition according to the user needs and environmental constraints. To tackle these challenges, we present a model-based approach to support AR-assisted product configuration based on the concept of Dynamic Software Product Lines. Our approach splits products (e.g. table) into parts (e.g. tabletop, table legs, funnier) with their 3D objects and additional information (e.g. name, price). The possible products, which can be configured out of these parts, are stored in a feature model. At runtime, this feature model can be used to configure 3D object compositions out of the product parts and adapt to user needs and environmental constraints. The benefits of this approach are demonstrated by a case study of configuring modular kitchens with the help of a prototypical mobile-based implementation.

Automated diagnosis of feature model configurations

Configurators can be evaluated in various ways such as efficiency and completeness of solution search, optimality of the proposed solutions, usability of configurator user interfaces, and configuration consistency. Due to the increasing size and complexity of feature models, the integration of recommendation algorithms with feature model configurators becomes relevant. In this paper, we show how the output of a recommender system can be evaluated within the scope of feature model configuration scenarios. Overall, we argue that the discussed ways of measuring recommendation quality help developers to gain a broader view on evaluation techniques in constraint-based recommendation domains.

Testing every member of a product line individually is often impracticable due to large number of possible product configurations. Thus, feature models are frequently used to generate samples, i.e., subsets of product configurations under test. Besides the extensively studied combinatorial interaction testing (CIT) approach for coverage-driven sample generation, only few approaches exist so far adopting mutation testing to emulate faults in feature models to be detected by a sample. In this paper, we present a mutation-based sampling framework for fault-based product-line testing. We define a comprehensive catalog of atomic mutation operators on the graphical representation of feature models. This way, we are able (1) to also define complex mutation operators emulating more subtle faults, and (2) to classify operators semantically, e.g., to avoid redundant and equivalent mutants. We further introduce similarity-based mutant selection and higher order mutation strategies to reduce testing efforts. Our implementation is based on the graph transformation engine Henshin and is evaluated concerning effectiveness/efficiency trade-offs.

Needs for mass customization and economies of scale have pushed engineers to develop Software Product Lines (SPLs). SPLs are families of products sharing commonalities and exhibiting differences, built by reusing software assets abstractly represented by features. Feature models describe the constraints that link the features and allow the configuration of tailored software products. Common SPLs involve hundreds, even thousands of features, leading to billions of possible software products. As a result, testing a product line is challenging due to the enormous size of the possible products. Existing techniques focus on testing based on the product line's feature model by selecting a limited set of products to test. Being created manually or reverse-engineered, feature models are prone to errors impacting the generated test suites. In this paper, we examine ability of test suites to detect such errors. In particular, we propose two mutation operators to derive erroneous feature models (mutants) from an original feature model and assess the capability of the generated original test suite to kill the mutants. Experimentation on real feature models demonstrate that dissimilar tests suites have a higher mutant detection ability than similar ones, thus validating the relevance of similarity-driven product line testing.