top of page


Big data optimization of algorithms for autonomous driving

the task

The aim is to check the quality of the autonomous steering of the vehicle. The so-called “Steering Reversal Rate” was defined as a measured variable, which shows how often and intensively the vehicle has to correct an initiated steering. After the evaluation, this key figure is then correlated with other driving parameters in order to carry out a cause analysis.


The challenge

Autonomous driving is characterized by a high volume and complexity of data, as it typically comprises measurement data, audio and video recordings and detected objects. All data types are recorded in a wide variety of file formats from many different providers, which makes access to and work with the data very difficult.

our solution

For this purpose, the customer was initially provided with an environment for big data analyzes in a cloud environment. The data analysis takes place via a data loading path for proprietary customer data formats, which regularly searches for files and converts them to a standardized big data format. Using an analysis language for time series analysis, it was now possible to examine and optimize the steering behavior on big data via parameter variations.


The customer benefit

The comfortable analysis language enables scalable evaluation in self-service and saves the customer the requirements and pitfalls of parallel computing. This enables the research and rapid testing of hypotheses and thus makes a decisive contribution to the speed of innovation in algorithm development.



Our role

  • Support of the customer by data scientists and data engineers

Our activities

  • Installation of DaSense on an MS-Azure subscription

  • Creation of a workflow for the conversion of MDF and AVRO files into a big data format

  • Implementation of a training in the creation of an analysis notebook and its transfer to an app

  • Support with data analysis

Technologies & methods

  • Applications: DaSense, Oozie

  • Data / databases: mdf4, AVRO

  • Languages / Frameworks: Python, Jupyter, Spark, bokeh

  • Methods: time series analysis

bottom of page