Análisis integral de los resultados obtenidos en la Etapa I 47 

Two prominent industry figures Mössinger (2010) of Robert Bosch and Heinecke (2010) of BMW both conclude that to cope with complexity and rate of change, standards for open systems and sharing and reuse of software are absolutely vital, Heinecke referring to it as “the Survival Strategy”.

The major automotive initiative in this respect is AUTOSAR (Automotive Open System Architecture), which aims to create a standard operating environment of basic software as illustrated in Figure 2.6 (www.autosar.org). On top of standardized basic software

to all ECUs and supports the sharing, reuse and flexible deployment of software components with an AUTOSAR interface.

Figure 2.6 AUTOSAR Automotive Open System Architecture (www.autosar.org)

As well as the base layer software and run time environment AUTOSAR also includes an underlying methodology and toolset for the deployment of software functions. This is done initially through linking software components through a Virtual Functional Bus (VFB) which abstracts communication layer away from the application software hence improving composability (the ability to guarantee that a component property is preserved when

integrated) and reusability of software components (Heinecke, Damm, Josko, Metzner, Kopetz,Sangiovanni-Vincentelli, and Di Natale, 2008). This is illustrated in Figure 2.7.

Figure 2.7 AUTOSAR process for deploying standard software components (AUTOSAR 2013)

The AUTOSAR methodology uses a series of design artefacts, such as ECU descriptions, system constraint descriptions and interface databases, and tools which enforce a contractual approach to the interactions between software functions. In a related initiative called Artop, an AUTOSAR Tool Platform or ecosystem is being developed to create an integrated development environment (IDE) for standard AUTOSAR tools and more advanced tools to enhance the design and development process (Knüchel, Rudorfer, Voget, Eberle,

Sezestre and Loyer 2010). The structure for the Artop tool platform is illustrated in Figure 2.8.

Figure 2.8 Artop AUTOSAR Tool Platform (Artop 2013)

AUTOSAR is not the only automotive open-source initiative. The GENIVI initiative is driving the broad adoption of an In-Vehicle Infotainment (IVI) open-source development platform based on a Linux operating system (GENIVI 2013). From a robustness perspective the opportunity from having standardised software elements must be balanced against the risks of taking commercial off-the-shelf software from a wider and potentially less mature supplier base.

Another area highlighted by Mössinger (2010) and Heinecke (2010) where there are significant developments is that area of higher bandwidth and more deterministic networking protocols such as Flexray and Ethernet. Heinecke et al (2008) note that these busses can support new more integrated topologies with less physical interconnections and ECUs and supporting the partitioning of a single physical communication channel to reduce temporal interference and improve error containment.

Flexray was first introduced by BMW, Schedl (2007) gives the motivation for this is be driven by CAN bus bandwidth limitations, unacceptable delays caused by multiple gateways

and unacceptable levels of complexity using multiple CAN buses. This led to additional costs for system integration and warranty and a need for deterministic behaviour. More recent work (Scheickl, Ainhauser, Gliwa, 2012) has used timing extensions from the AUTOSAR environment to apply tools for the specification and verification of timing constraints to optimally exploit Flexray network’s deterministic capabilities.

Hyung-Taek, Volker and Herrscher (2011) note there is a gathering pace for higher bandwidth and lower cost busses initially for diagnostics and software updates and as a lower cost higher bandwidth infotainment bus but future potential to provide additional high speed data communication between larger ECUs, for instance for future driver assistance applications. Figure 2.9 illustrates a roll-out strategy for Automotive Ethernet described by Hank, Suermann and Müller (2012) culminating an automotive Time-Triggered version providing a high bandwidth deterministic network backbone.

Finally the other key approach adopted within the automotive industry, noted by Mössinger (2010) and Heinecke (2010) among others, is model based development methods. To date the main focus of this has been at a functional level to support the auto-coding of software from models but is now starting to be used at higher levels to model systems and architectures which will be further examined in Section 2.9.

2.5

Agile Software Development Methods

If software development is one of the major challenges of coping with complexity for the automotive industry then it is worthwhile considering what lessons can be learned in the wider software development community. One of the key developments in this field is that of the practice of agile software development. Highsmith, one of the founding fathers and leading proponents of agile software techniques, defines agility as “the ability to both create and respond to change in order to profit in a turbulent business environment” and “the ability to balance flexibility and stability” Highsmith (2004).

Larman (2004) describes how software developers have learned that software development is subject to emergent behaviour and have adopted agile and iterative development techniques. An early software development model was the waterfall model whereby the outputs of one stage cascade to the next stage and the outputs are developed in a linear sequential fashion. This model has been augmented to add a mirror image test and validation stream in the V-Model. This model also has been adopted on a wider basis e.g. vehicle product development system is often based on a V-Model. However in practice there are issues in implementing sequential models as projects become more complex and difficult to predict. The key learning from the software development community from the widespread

adoption of agile practices for successfully delivering complex products is to conduct iterative and incremental development which leads to progressive integration and testing.