Requirements Engineering in a Model-Based Methodology for Embedded Automotive Software

Requirements Engineering in a Model-Based Methodology for Embedded Automotive Software

Jean-Louis Boulanger (CERTIFER, France)
DOI: 10.4018/978-1-60566-731-7.ch002
OnDemand PDF Download:
No Current Special Offers


This chapter resumes the requirements engineering in a model-based methodology for embedded automotive software. The methodology relies on two standard languages: EAST-ADL for automotive architecture description and SysML for systems modeling. The requirements engineering in the methodology describes phases of elicitation, modeling, traceability, verification and validation. It is illustrated by applying on a case study -- the knock controller -- a part of the engine management system.
Chapter Preview


Modern car is now equipped with more and more functionalities dependent on embedded electronics, ranging from powertrain and chassis control to body comfort and infotainment. These functionalities are distributed over a networked Electronic Control Units (ECU). The size and complexity of software for these embedded electronics increase rapidly with its cost raising from 10% of the overall cost in 1970 to 40% in 2010. Actually, 90% of innovations in the automotive industry concerns embedded electronics and 80% among them are software see Bortolazzi, J. (2004).

A big challenge in developing automotive software concerns the quality. Automotive systems are safety-critical systems where failures may cause severe damages or loss, so software errors led directly to car recalls. According to the report Gumbrich, S. (2004), one-third of the recalls in recent year is caused by software errors. More efforts are needed on software's verification and testing.

Another challenge concerns the reduction the time of development. The automotive market is shared by manufacturers, suppliers and tool vendors, and all needs shorten processes which favor the exchangeability among them and the reuse of software in different product lines. They also need to follow requirements along the development, from the specification to design and code, to anticipate and communicate its changes throughout teams.

New paradigm in software development is born in this context to face these challenges. In Europe, automotive actors tried to cooperate on a common base for software development. The result of this corporation is EAST-ADL (2004), a recently defined standard. EAST-ADL is an architecture description language dedicated to automotive software. It provides a mean to describe the functionality of a vehicle, from high level requirement to implementation details. It focuses on structural aspect, leaving behavioral aspect for existing tools. EAST-ADL is based on Unified Modeling Language 2 (UML (2007)) but has automotive-specific constructs and semantics in order to make system models unambiguous, consistent and exchangeable.

Model-based development (MBD) is a preferred approach for automotive software because it improves the specification, design, and implementation phases. Model-based development benefits from the Systems Modeling Language (SysML (2006)), another recently defined by Object Management Group (OMG). SysML (2006) gives a means to early represent into models the requirements and physical parametric of automotive systems. SysML (2006) has also the capacities of facilitate the design and verification.

The research project MeMVaTEx (2008) addresses a model-based methodology that emphasizes the requirements validation and traceability. The methodology invests two languages: EAST-ADL (2004) for automotive architecture description and SysML (2006) for system modeling. The methodology describes step-by-step process with appropriate tools supporting each step. It aims to give a \textit{seamless} solution for industrial use. An automotive case study -- the engine knock controller -- a part of the Engine Management System (EMS) is used to illustrated the methodology.

This paper shows the requirements engineering in the methodology. It describes phases of elicitation, modeling and traceability, verification and validation, and accompanied tools. The methodology concerns other aspects like safety, real-time, variability, or model transformation that will not be addressed here. Our related works A. Albinet, J-L. Boulanger, H. Dubois and al. (2007), C. André, F. Malet, and M.-A. Peraldi-Frati (2007), J-L. Boulanger and Q-D. Van (2007), A. Albinet, S. Begoc, J-L. Boulanger and al. (2008), J-L. Boulanger and Q-D. Van (2008)can be found on the Web site MeMVaTEx (2008).

Complete Chapter List

Search this Book: