The automotive industry is facing the challenge to gain knowledge from large datasets – originating for example from production or from fleet data. Problems in production or in the fleet most likely become manifest in problems in the data and result in increasing costs. Examples can range from problems in the on-board electrical system to virtual validation of autonomous driving functions in R&D. Hence, one important use case is the automatic detection of anomalous behavior in the data to forecast and identify potential problems as early as possible. We demonstrate new mathematical methods from Topological Data Analysis (TDA) that can help to address these kinds of problems. TDA is a rather new field in mathematics that combines techniques from geometry and topology in order to analyze noisy datasets. Beside academia, it has been applied successfully to various fields including medicine (identification of tumor cells), finance (fraud detection), and materials science (structure analysis). We highlight two main methods from TDA: the ball mapper algorithm and persistent homology and illustrate potential applications in automotive industry. Even though the methods are highly mathematical, we show that they are easy to use and that they can produce valuable knowledge about potential problems - for example in the automotive context - giving an added value to the customer and the OEM.
Session:
AUTONOMOUS DRIVING I
|
| 10:05 - 10:35