facebook tracking
Master thesis 2019 - Project suggestions

Master thesis 2019 - Project suggestions

Write Your Thesis with Us

Your ideas and research results from your study are too valuable just to disappear in a drawer. Therefore, we offer you the opportunity to write your thesis about current, innovative subjects that are close to practice. 

We are looking for highly motivated students with background in computer science or similar. We prefer Master Thesis projects handled by two students.

Students are welcome with their own ideas but here is suggestions from our team:

    Investigate possible improved storage format for AUTOSAR models, capable of being version controlled.

    AUTOSAR models are usually saved in an XML variant called ARXML. This model format is not friendly to traditional version control programs such as git or subversion, and this results in major complications when a model is continuously changed over ti
    me. It would be a major benefit to the Automotive industry if a loss-less conversion could be developed that allows the saving of ARXML in traditional version control systems, while being readable by traditional automotive industry programs.

    Develop a way to efficiently perform lazy partial loading of massive automotive models.

    AUTOSAR models tends to grow larger and larger as the automotive software gets more and more complicated. This causes longer loading times and higher RAM memory usage from the modeling tools, and it’s starting to reach the point where lazy loading of pieces of models is a must. Complications however arise in that AUTOSAR models are rich in cross-references, both single and bi-directional and that users’ behavior and changes could break or invalidate those links. 
    The problem is to develop a lazy-loading cache that will only load parts of a full model, allow changes, and still not display any out-of-date or misrepresenting data to a user of the system. Ideally the cache should also be able to warn when operations would break the data-links without loading the entire model into memory.

    Develop a high level “language” for modeling automotive systems.

    AUTOSAR models are getting larger and larger as the car systems they are describing are getting more and more complex. It is now reaching the point where it is difficult or even impossible for humans to have any kind of overview of how the entire system works. This problem is made worse by the fact that AUTOSAR demands very detailed models, and thus larger models. 
    Could an abstraction to AUTOSAR be found by analyzing how general car ECUs are modeled and what parts of AUTOSAR are actually used? The goal being to describe the system on a higher level from which the detailed AUTOSAR format can then be automatically generated, or compiled? 

    Develop a standardized test-suite to analyze and compare adaptive AUTOSAR implementations.

    Adaptive AUTOSAR is a very new standard, and for now it is very difficult to know what kind of minimum requirements that it will place on hardware. It also constitutes a large change in how automotive ECUs work and it is not properly known what kind of requirements and performance can be expected of an Adaptive AUTOSAR system.
    It would therefore be good to study similar architectures from other domains and check how they work to measure and improve the performance of their core systems. This could then be developed into an Adaptive AUTOSAR performance test. Which could be used to compare different implementations, and to ensure consistent improvement of ARCCOREs own products. In other words, to establish a benchmark test with metrics for Adaptive AUTOSAR. Such a benchmark would likely involve stressing the system, error injection and AUTOSAR coverage.

    Develop smart ways of interacting between old signal based automotive systems and new service-based ones.

    In the automotive systems world of yesterday, almost all communication was made with “signals”, that is fixed communication links with known content between known parties. With Adaptive AUTOSAR and the industry in general moving more into a “service” based world, where one “provider” of content can be “subscribed” to by any other component in the car.
    For at least the foreseeable future these two design methodologies will have to coexist, both for financial reasons (it’s expensive to remake everything) and for safety reasons (safety critical functionality must be deterministic). Therefore, there is a need to investigate different methodologies to interact between those two designs. Should any signal be directly mapped to a service? Should there be a central lookup of signals for the service world? What are the pros and cons of each approach? What is most efficient? Most safe? Fastest?

    Investigate ways to analyze and look for faults in complex service based automotive systems.

    As the automotive systems gets increasingly more complex it is getting more and more difficult to create fault detecting systems that are meaningful. A traditional approach is to set up “report” points where the different parts of the system have to report what they have done and since the entire system is known this can be used to ensure the right things happen in the right order.
    However, in the newer dynamic system, no one can know from the start exactly what should happen in the system, or exactly when. In such a system, it might be difficult, or maybe even meaningless to use the traditional approaches. Instead there is a need to investigate how other computer architectures handle fault detection, for example how does widely distributed data servers check for faults? Possible approaches might be analyzing data flow using metrics/AI, having fixed behavior subsystems or maybe something completely different.
    People interested in this area will likely need experience of fault detection strategies from at least one service oriented domain.

    In the new upcoming standard Adaptive AUTOSAR it will be possible to download new application software to the car without reprogramming the complete unit and without going to the workshop. This creates new possibilities for development of new car functions but also deploying fixes for existing problems. On the other side this possibility also creates vulnerabilities for the car i.e. there is a risk for viruses, security leaks and other types of malfunctions.
    In this master thesis we would like to evaluate the needs and requirements that should be put on the infrastructure that distributes the software to the car. I.e. what protocols to use, authentication mechanisms, recovery strategies and
    so on. The master thesis should result in a theoretical study of available mechanisms used in other IT systems and a prototype implementation of such a system for the automotive industry.

    Investigate the possibilities of current and coming releases of standard Linux. How does the deterministic behavior vary with CPU load and how should applications be configured to use the included features optimally?

    Historically, the computers in vehicles have been based on small microcontrollers communicating using a CAN bus or similar transport mechanisms. The requirements on the software in future vehicles demands for much more performant hardware and network infrastructure (compare clusters of servers using Ethernet and TCP/UDP). This change put also a high demand on the OS (operating system) used for this type of systems. Currently there are discussions of using standard Linux, real time versions of Linux and specific kernels like PikeOs, QNX and VxWorks.Most development of code for these systems are performed on Windows or on a standard Linux system like Ubuntu.

    This investigation should look at the possibility to use a standard Linux kernel (current, future releases) like Ubuntu (stripped one?)There are many characteristics that are important to investigate. Examples are process isolation, file systems, determinism, context switching, processes and threads, core usage efficiency, safety and real time characteristics including hardware structure.

    The focus for this investigation should be which deterministic behaviour that can be achieved using a standard Linux kernel. How should it be configured; how should the applications load be distributed and how does the characteristics vary with CPU load.The work will include theoretical studies of the area, information search, development of test suites, testing on Standard Linux and special automotive hardware. The result of the thesis work shall be documented in a technical report and presented for people interested in the area.

    C++ is the language that has been selected for development of future applications for complex systems located inside next generation intelligent vehicles. See application areas like ADAS and AD (Advanced Driver-Assistance Systems and Autonomous Driving). Major initiatives are currently on going to standardize the middleware platforms for these types of applications. They are all based on C++ and its extensive STL library. But how suitable is C++ for these types of applications?

    • Which parts in the language is not suitable to use when building real time applications fulfilling QM requirements respective ASIL B, D requirements. Why? What can we use instead? What is done about it from a standards perspective.
    • Which parts are not predictable and why? What can we use instead?
    • Propose replacement components that could be developed and used instead.
    • Analyze the current used coding rules and prioritize them from the perspective of suitable usage in above area.
    • Suggest a review checklist. Which are the most important parts that should be looked at when performing reviews. (In group exercises, when doing pull requests (see Bitbucket, GitHub, GitLab processes)
    • Which are the weak/strong points of the language? (Compare different compilers environments)
    • Based on the information collected during the study, try to formulate a conclusion on how suitable the language is for the type of applications existing in future vehicles. Suggestion on alternatives?
    The work will include deep studies in how C++ is implemented, programming of typical examples, design of replacement/interim components complementing current STL. The result of the thesis work shall be documented in a technical report and presented for people interested in the area.


Are you interested and want more detailed information about the projects? Contact Annie.friden@arccore.com


ARCCORE, founded in 2006 and headquartered in Sweden, is a global supplier of embedded software products for highly complex automotive software solutions focusing on the Autonomous Drive, Infotainment and Embedded Control market. Our products are based on renowned industry standards like AUTOSAR, Linux, Android and more and we are an active contributor to the global development of new standards. We are close to 100 embedded software experts in development centers in Sweden, Germany, India and China.

ARCCORE is a fully owned subsidiary of German headquartered Vector Informatik since 2018. At 24 locations worldwide, over 2,000 Vector employees support manufacturers and suppliers of the automotive industry and related industries with a professional platform of tools, software components and services for developing embedded systems.

Apply for this job

Or, know someone who would be a perfect fit? Let them know!

Workplace & Culture

At Arccore you are given a high degree of freedom, the opportunity to take own responsibility and to make decisions which means that you will grow and learn new things. Arccore has an outstanding company culture, where sharing of knowledge among developers is essential. At Arccore we embrace diversity and are proud to have co-workers that are passionate about their work from all over the world! This makes Arccore a productive workplace where we have fun together.

Already working at ARCCORE?

Let’s recruit together and find your next colleague.


Applicant tracking system by Teamtailor