Klaas Ebel's career in the automotive industry has seen him work for various OEMs at different levels across multiple teams. His favourite aspect of the job is making the world of transportation a little safer every day.
After studying snow physics, his first role in automotive was in numerical crash-test simulation. There, he learned always to determine the root cause before trying to fix a problem. This skill brought him to lead teams through every phase of the vehicle-development process, from architecture to homologation and certification.
He contributes to committees concerned with evaluation software for vehicle safety and nurtures an attitude of progressing a little further every day.
Presentation
Consistent ADAS Evaluation Across the Entire Development Process
Where human beings work together, accurate communication becomes essential. The accuracy of that communication needs to be reviewed at intervals to ensure its quality, particularly over a vehicle development cycle. This is especially important in ADAS development, a field that is no longer new, as vehicles become increasingly defined by software and computing power with every model iteration.
With the help of ML and AI tools, it has become very convenient to generate slides from data repositories regarding key KPIs. This raises the question of how humans, AI agents and KPIs stay synchronised. As these KPIs may change during a project, so too may boundary conditions such as rules and regulations. How do modern AI-based systems react to this, and how can we make sure everyone involved is made aware at the right time?
The data channels need to be as clearly defined as the tools used to extract the KPIs. What are the options for mastering this challenge? Even more AI? Back to paper? Or perhaps something basic in between? One option could be to use the same evaluation tool at various stages of the process and have it hard-wired to the boundary conditions.
Is this perfectly outdated, or maybe the next step?
