Mark-Oliver Reiser

Multi-level feature trees: A pragmatic approach to managing highly complex product families

Feature modeling has become a popular technique for domain analysis and variability management. However, it is still a considerable challenge to apply this technique to product families and organizational contexts of high complexity like the product range of a global automotive corporation. Managing everything as a single product family with a global feature tree is virtually impossible owing to the enormous complexity, but if the product range is split up into several smaller, independent product lines with separate feature models, systematic reuse and strategic variability management across these portions is lost. In this article, we present multi-level feature trees, which offer a compromise between a single global and several smaller, independent feature trees. Other development artifacts may also be arranged in this way if the multi-level concept is adapted to them. This is shown exemplarily for requirements artifacts in Telelogic Doors. Finally, we describe scenarios showing how this concept can be put into practice.

Managing Highly Complex Product Families with Multi-Level Feature Trees

Feature trees are a well-established instrument for domain analysis and modeling. But for highly complex product families like a vehicle manufacturers product range comprising well-above a thousand technical features - they become very large, and thus cumbersome and inflexible, especially when managing changes to the trees structure over a long time. Furthermore, a conflict of aims arises when using feature trees in large organizations: while a single global feature tree for the entire company is desirable, local amendments for individual units or projects are indispensable in practice. In this paper, we present a detailed description of this problem and show its great relevance to the automotive domain. We then provide a detailed definition of multi-level feature trees as a possible solution to the above problem. Finally, we describe scenarios how such multi-level feature trees can be put into practice and introduce a prototypical tool implementation of this concept

The EAST-ADL architecture description language for automotive embedded software

Current trends in automotive embedded systems focus on how to manage the increasing software content, with a strong emphasis on standardization of the embedded software structure. The management of engineering information remains a critical challenge in order to support development and other stages of the life-cycle. System modelling based on an Architecture Description Language (ADL) is a way to keep these assets within one information structure. This paper presents the EAST- ADL2 modelling language, developed in the ITEA EAST-EEA project and further enhanced in the ATESST project (www.atesst.org). EAST- ADL2 supports comprehensive model-based development of embedded systems and provides dedicated constructs to facilitate variability and product line management, requirements engineering, representation of functional as well as software/hardware solutions, and timing and safety analysis.

Automatic allocation of safety integrity levels

In this paper, we describe a concept for the automatic allocation of general Safety Integrity Levels (SILs) to subsystems and components of complex hierarchical networked architectures that deliver sets of safety critical functions. The concept is generic and can be adapted to facilitate the safety engineering approach defined in several standards that employ the concept of integrity or assurance levels including ISO 26262, the emerging automotive safety standard. SIL allocation is facilitated by HiP-HOPS, an automated safety analysis tool, and can be performed in the context of development using EAST-ADL2, an automotive architecture description language. The process rationalizes complex risk allocation and leads to optimal/economic allocation of SILs.

Managing Complexity of Automotive Electronics Using the EAST-ADL

The complexity of embedded automotive systems calls for a more rigorous approach to system development compared to current state of practice. A critical issue is the management of the engineering information that defines the embedded system. Development time, cost efficiency, quality and dependability all benefit from appropriate information management. System modeling based on an architecture description language is a way to keep the engineering information within one information structure. The EAST-ADL was developed in the EAST-EEA project (www.easteea.net) and is an architecture description language for automotive embedded systems. It is currently refined in the ATESSTproject (www.atesst.org). This paper gives an overview of the EAST-ADL and accounts for some recent refinements as developed in the ATESST project. Areas covered include the relation to other standardization initiatives such as UML2.0, AADL, AUTOSAR, SysML, Marte profile, requirements management and variability.

Automatic optimisation of system architectures using EAST-ADL

Abstract There are many challenges which face designers of complex system architectures, particularly safety–critical or real-time systems. The introduction of Architecture Description Languages (ADLs) has helped to meet these challenges by consolidating information about a system and providing a platform for modelling and analysis capabilities. However, managing this wealth of information can still be problematic, and evaluation of potential design decisions is still often performed manually. Automatic architectural optimisation can be used to assist this decision process, enabling designers to rapidly explore many different options and evaluate them according to specific criteria. In this paper, we present a multi-objective optimisation approach based on EAST-ADL , an ADL in the automotive domain, with the goal of combining the advantages of ADLs and architectural optimisation. The approach is designed to be extensible and leverages the capabilities of EAST-ADL to provide support for evaluation according to different factors, including dependability, timing/performance, and cost. The technique is applied to an illustrative example system featuring both hardware and software perspectives, demonstrating the potential benefits of this concept to the design of embedded system architectures.

Using product sets to define complex product decisions

Product family engineering consists of several activities commonly separated into the areas of domain engineering and product engineering. The main part of product engineering is the definition of product decisions, which means in the context of feature modeling that for each feature the product engineer has to define in what products it will be included. In the automotive domain – and probably in many other embedded real-time domains as well – the considerations that influence these feature selections are extremely complex and, at the same time, need to be documented as closely as possible for later reference. In this paper, we (1) present a detailed description of this problem and (2) try to show that existing approaches do not sufficiently meet these concerns. We then (3) provide a detailed definition of product sets as a means to solve the problem and (4) show what methodological implications arise from the use of this concept.

Towards improving dependability of automotive systems by using the EAST-ADL architecture description language

The complexity of embedded automotive systems calls for a more rigorous approach to system development compared to current state of practice. A critical issue is the management of the engineering information that defines the embedded system. Development time, cost efficiency, quality and most importantly, dependability, all benefit from appropriate information management. System modeling based on an architecture description language is a way to keep the engineering information in one information structure. The EAST-ADL was developed in the EAST-EEA project (www.east-eea.org) and is an architecture description language for automotive embedded systems. It is currently refined in the ATESST project (www.atesst.org). This chapter describes how dependability is addressed in the EAST-ADL. The engineering process defined in the EASIS project (www.easis-online.org) is used as an example to illustrate the support for engineering processes in EAST-ADL.

Compositional Variability — Concepts and Patterns ∗

Most software-intensive systems rely on a componentbased design and are therefore made up of encapsulated structural units which are hierarchically composed of one another. In this paper, we (1) propose a scheme for rigorously managing variability in the context of such a compositional hierarchy, which consistently extends the paradigm of component-based design to variability management, (2) present several basic patterns of specifying variability when applying this scheme in practice, and (3) show how all this was technically realized in EAST-ADL2, an architecture description language for automotive software development. While the observations and concepts discussed in this paper emerged from an automotive context, they are arguably applicable to many other industrial domains involving software-intensive systems.

Managing variability and reuse of features and requirements for large and complex organizational structures

Feature trees are a well-established instrument for domain analysis and modeling (Czarnecki and Eisenecker, 2000; Kang et al., 1990). But for extremely complex product families like a vehicle manufacturers product range, they become very large, and thus cumbersome and inflexible. This is particularly true when changes to the trees structure have to be managed and documented over a long period of time. Furthermore, a conflict of aims arises when using feature trees in large organizations: while a single global feature tree for the entire company is desirable, local amendments are indispensable because often some applications of the feature models require amendments not visible in other application contexts. A research project at DaimlerChrysler, being conducted in cooperation with the Technical University of Berlin, has examined this problem in detail and elaborated an approach based on a distribution of the feature tree definition over several hierarchical levels. Using this approach, amendments to the feature structure can be introduced locally at first, and can later become valid in a broader context, if desired. Currently the approach is being evaluated and refined in case studies.