Solution Study: Efficient mining of relevant scenarios – From data collection to processing

by Thomas Mauthner - Department Manager Real-World ADAS/AD Solutions

Representative and reliable data is essential for ADAS/AD development, particularly in perception and machine learning tasks. Moreover, future homologation and certification of ADAS/AD functions will demand even more real-world data. While collecting and digitizing real-world driving scenarios is indispensable, it remains a time- and cost-intensive process. AVL solutions support the optimization of data collection from the outset, covering efficient campaign planning, hardware setup, and data logistics. The process begins with precise route planning to effectively cover the required Operational Design Domain. Vehicles are then equipped with the necessary hardware and software to capture raw data. A robust and reliable data collection setup, combined with seamless campaign execution, is crucial for enabling real-time insights and informed decision-making during data collection. However, state-of-the-art sensors generate vast amounts of data, necessitating efficient processing to extract meaningful scenarios. Additionally, the digitalization of both dynamic and static scenario elements plays a key role in maximizing the value of collected data.

This presentation will provide insights into:

- The application of route planning

- The types of tags that can be generated in real-time

- How online tagging can significantly reduce data volume

- The automated digitalization of raw data into standardized formats