AI – Challenges in application with bus data in the automotive sector

Autoren

Alexander Faul, Maria Floruß, Vector Informatik GmbH;
Felix Pistorius, Karlsruher Institut für Technologie

Jahr

2019

Zusammenfassung

Artificial Intelligence (AI) is a scientific field which emerged in the 1950s shortly after the introduction of the first electronic and programable computers. Since then, the field has hit many astonishing milestones, like defeating the then world chess champion Garry Kasparow in 1997 (Deep Blue, IBM) or surpassing human abilities in areas like visual object recognition in images (ISBI [1], ICPR [2]). Today it is used widely in many different areas of our daily lives, from autonomous cars and drones to medical expert systems to recommender systems in online shops. One important factor for these accomplishments where the increased computation performance of CPUs and especially the usage of GPUs introduced in the 1990s. The development of special hardware and better algorithms lead to further accomplishments like defeating a human in the
strategic board game Go (AlphaGo, March 2016) and the possibility to use AI to analyse growing data collections, more commonly referred to as “big data”. While the big internet companies like Google and Facebook are well known to amass user data, more and more “classic” industry companies also start collecting their business and product data which oftentimes is too large and complex to deal with manually.
To analyse these ever-growing masses of data, several so-called Data Mining (DM) processes were introduced early on. Most commonly known are the Knowledge Discovery in Databases (KDD) [3] process, which will be described in this paper, as well as the Cross-Industry Standard Process for Data Mining (CRISPDM) [4]. These processes not only comprise AI algorithms, but describe the whole pipeline: from collecting the data and selecting the relevant parts, to preprocessing, filtering,
and transforming it, to the actual analysis using AI algorithms, and finally an interpretation of the results.
As a consequence of this trend of growing masses of data, complex processes, and algorithms,
engineers, who up to now have analysed this data with conventional methods, need more and more knowledge in the field of AI, or must be supported by a data scientist.
This paper gives a broad overview of the field of AI, and analyses the challenges and requirements the mentioned DM processes entail, with emphasis on applications in the automobile industry and bus data. It also proposes an improvement to the current DM processes as a solution to communication challenges.

Mitglieder-Login