Development of embedded software in the automotive domain is a complex task involving the combination of multi-discipline and safety critical requirements. In such an environment, traceability to and from related software development artifacts is demanded by safety standards. It is also needed to facilitate activities such as impact analysis and software maintenance. Despite a lot of research done on software traceability, challenges still exist for complex domains such as the automotive domain due to development being done with multiple companies (suppliers) and integrated at the Original Equipment Manufactures (OEMs) side. This paper describes a PhD research study that is aimed at developing a traceability process, a set of relevant traceability link semantics and a tool. Expected results are that deployment of the process and tool will improve traceability both in terms of link correctness and completeness and that the semantics will support development activities such as impact analysis.
Software traceability refers to the ability to link related software artifacts to each
other [
An embedded system is a system whose critical function is not computational but
which is controlled with by a computer embedded in it [
Software development activities in the automotive industry lead to the production of a lot of diferent artifacts such as requirements, design models, code and tests. These artifacts are produced and consumed using various tools, in diferent teams within an organization and also sometimes across organizations. In industry, it is important to link related artifacts to each other so as to understand how they are related and what should happen if the artifacts evolve. Software traceability has been proposed as a general solution that can be used to establish the relationship between artifacts through the creation of traceability links between related artifacts. To achieve the full potential of traceability usage, the links created need to be correct and complete. Correctness in this case means that the artifacts contained in the traceability link are actually related, given a specific context. Completeness means that all the artifacts that are related are actually connected by a traceability link. In our context, completeness is therefore not a property of one link but a given set of traceability links. The establishment of correct and complete traceability links is not a trivial task because of three major issues that exist. Firstly, there is no clear definition or formalism of what process should be put in place for creation of these traceability links that will seamlessly integrate with the existing software development processes. Secondly, what semantics the traceability links should contain so that they can be useful after they are created and thirdly, what tools can be used to support this. The following sub-sections describe the above three problem areas in detail. 2.1
The process of creating traceability links is time consuming and the accuracy
of this step is a major factor in determining if the traceability links created will
be useful or not. In existing software development processes such as Agile or
the V-model, there is no explicit description of when in the development life
cycle, traceability links should be created or updated. This in turn makes the
creation and updating of traceability links something that is perceived as an
extra activity that adds overhead to the development process. Existing research
on traceability is mainly focused on the technical side of the problem by trying
to use techniques such as machine learning [
Current practice shows that companies only invest in traceability because either
their customers or standards demands them and not for their benefits [
Having a good traceability process is not enough, proper tools need to be put in
place so as to support the process. As mentioned previously, there exist research
on tools that support the creation and management of traceability links. However,
a persistent problem is when the artifacts to be connected reside in various
tools and are of various formats (e.g. models, code, word documents), and the
traceability links also need to be created and shared by diferent teams and
sometimes diferent companies. In addition, most tools do not allow for complete
traceability throughout the development lifecycle but for only some parts of
it, e.g, requirements to code or code to test cases[
artifacts involved other plugins need to be developed to support this. Currently, this leads to inconsistencies in traceability links as artifacts evolve since the tools are not aware of changes happening outside them. Our research will therefore focus on developing a tool that is configurable, can support traceability between arbitrary artifacts and exchange of traceability information between teams and companies.
Currently many companies spend a lot of efort on traceability in order to be certified by safety standards such as ISO 26262. Coming up with new process and tools to improve traceability will mean that they will save time and therefore costs on their side. It will also mean that they can use traceability for other benefits apart from certification. Moreover, for research, the results of this PhD will give an insight on relevant semantics of traceability links, which processes are suitable, how can they fit in existing processes such as the V-model and also how much benefits traceability actually yields. The research questions to be answered are as follows: 1. RQ1: In the automotive industry, what is the traceability links creation process that can lead to more correct and complete links? 2. RQ2: What are important traceability links semantics that can be used to improve software development activities? 3. RQ3: To what extent can traceability links improve software development activities? 4. RQ3: How can traceability links be maintained and kept consistent when artifacts reside in diferent tools, teams and organizations?
The PhD is partly sponsored by an ITEA 2 funded project, AMALTHEA4Public2. The aim of the project is to develop a common set of tools that can be used throughout the software development life cycle with automotive companies developing embedded systems. The project already has a tool chain called APP4MC3 that is based on Eclipse. The platform however, lacks any support for traceability and the PhD is part of a work package that is dedicated to investigating the existing traceability problems and providing solutions for the project partners. We expect that our solution will consist of a traceability process that can be integrated into normal software development processes such as the V-model and aligned with safety standards, a set of traceability metamodels that carry semantics that are relevant in diferent settings, for instance, in product line environments, multi-core development environments or simple embedded development environments. Moreover, our solution will also consist of a configurable tool that will allow companies to easily swap and extend the traceability metamodels depending on their use cases. The tool will also be able to support the evolution
and maintenance of traceability links. It will also support the exchange of artifacts and their traceability information between various teams and across organizations. The research will be done in small iterations and in close collaboration with the companies to make sure that what we deliver is not only of value to academia, but also to industry. The research methodology is further described in Section 6. With the possibility of industrial evaluation, we can research the best way for creating traceability links in various contexts, how the traceability links should be used and when and how they should be updated. 5
Even though there exist much research on traceability, not much has been done in relation to the embedded automotive industry. As mentioned before, this a complex domain due to the fact that they develop complex safety critical systems and in a distributed environment. For companies in this domain to be ISO 26262 certified, they need traceability. Our original contribution will be a development and integration of the following: first, a conceptual framework comprising of a traceability process for automotive industry and this process will be aligned with the ISO 26262 standard; second a set of meta-models that can be used to support various development environments in this domain, e.g. product lines; third, a prototype that can support the process and semantics and most of all allow for linking between arbitrary elements and exchange of this traceability information between teams and between companies. 6
The PhD will be carried out using the action research methodology [
As depicted in Figure 1, the action research cycles were preceded by a study of the state of the art. This was a literature study that investigated what other researchers have done in the field of traceability and what other tools and solutions have been proposed. The first cycle of the study is done in cooperation with the company rt-labs, a small company developing software for trucks and located in Gothenburg. The cycle began with a state of practice study carried out by sending surveys to all the project partners asking them about the software development activities that they conduct and the artifacts they use and produce. This gave us an idea of the development activities conducted. We also elicited traceability requirements from rt-labs through interviews and focus group discussions with employees at the company. Together with one of the project leaders from the company, we defined a suitable metamodel and implemented a prototype supporting the metamodel. We will then define an initial process for their traceability use cases and evaluate the process, metamodel and tool at the company by deploying it and getting feedback from actual use of the prototype. This will mark the end of the first iteration of the study. The derived process, metamodel, tool and other lessons learned from this cycle will be carried on as input to the next study cycle. The second, third and further cycles (depending on the need) will be carried out in collaboration with other companies. For now we are sure that the second study will be at Bosch in Germany. The cycles will focus on improving the process, enhancing the metamodel and deriving more metamodels when needed and also adding more features to the traceability management prototype. In each cycle we will also evaluate the results through observing the impact in practice and through controlled experiments with the prototype.
Define initial metamodel
Prototype
Implementation Literature study
Requirements elicitation (rt-labs)
State of practice study
1st Iteration
Develop initial
process Evaluate process, metamodel and prototype at rt-labs
Process refinement
Prototype enhancement
Metamodel Outcome enhancement
Nth
Iteration Field study at partner company
Evaluate process, metamodel and prototype at partner company The PhD started in April 2015 with literature review of the research that already exists in the area of traceability in general and traceability in the embedded automotive domain. We also did a literature survey of existing traceability management tools and what they ofer. In May, we sent out a survey to all the project partners to collect information on their workflow. Knowing their workflow helps us in knowing what software development activities they conduct and what artifacts are used and produced in these activities and how these artifacts are related to other artifacts. This information also gave us an insight of which tools are used in creating and consuming these software development artifacts. After the data collection, we conducted two interviews and one focus group with some of the partners on the requirements that they have regarding software traceability.
From the requirements collected, in October we started to implement a prototype for a traceability management tool in collaboration with rt-labs. The aim of implementing this with one project partner is first to get the prototype working on this one company and then expand the evaluation with more partner companies in the project. The next steps for the study will be to conduct a case study at Bosch, because it is one of the companies that are already using the AMALTHEA platform in order to gather more requirements relating to traceability especially when working with the platform. This study will take place in February 2016.