Not sure how accurate the translation is, it's AI generated for crying out loud, LOL.![]()
Autonomes Fahren - quo vadis? Ich kopiere mich hierzu mal selbst: Fast exakt vor einem Jahr hab ich bei LinkedIn einen Bericht über die damals 37. Fachtagung des VDI e.V. und VDI Wissensforum GmbH zu… | Alexander Bloch | 135 comments
Autonomes Fahren - quo vadis? Ich kopiere mich hierzu mal selbst: Fast exakt vor einem Jahr hab ich bei LinkedIn einen Bericht über die damals 37. Fachtagung des VDI e.V. und VDI Wissensforum GmbH zu autonomen Fahrsystemen geschrieben. Auch im Lichte der neuesten Ereignisse und vor allem nach...www.linkedin.com
Can someone translate please?
View attachment 88113
Autonomous driving - quo vadis?
I'll copy myself: Almost exactly a year ago, I wrote a report on LinkedIn about the then 37th symposium of the VDI e.V. and VDI Wissensforum GmbH on autonomous driving systems. Even in the light of the latest events and especially after half a dozen further tests that I have carried out since then (coming soon on VOX auto mobil), the five main points I mentioned at that time remain up-to-date with small additions.
1. Fully autonomous driving (Level 4) of a production car is still a long way off. None of the numerous experts wanted to estimate a year. Especially if you define it via the "Safety of the intended functionality" (SOTIF) with humans as an error rate reference. Fully autonomous driving systems should not have an accident rate higher than that of humans - ideally much lower. According to Professor Mirko Maehlisch, we are still a long way from this with a clean analysis and, above all, individual test drives are not meaningful.
2. The use of artificial intelligence has significantly reduced the error rate in the test vehicles. But it is not the game changer on its own. According to AI specialist Andreas Kuhn, the situation in real traffic is not to be assessed strictly according to engineering aspects, but here decisions must be made to a large extent according to game theory principles and thus also (error) probabilities.
3. When it comes to sensor technology, the vast majority of experts agree that a monosensory solution is not error-resistant enough for autonomous driving for redundancy reasons alone. Especially if it is to work without any problems even in difficult environmental conditions. I can only support this from my tester experience.
4. The classic car manufacturers in particular - the old economy - must completely rethink their development if they want to keep up with the new economy, for example from California, which is completely different in terms of software. If structures established over decades can only be changed rather sedately in development, it may make sense to start completely anew elsewhere - see joint venture VW and Rivian.
5. There is a discrepancy between the public expectation of autonomous driving, which is strongly influenced by personal brand preferences, wishes and marketing, and what will come in the foreseeable future. Even as a car tester, I am often surprised by effusive testimonials about certain systems that do not agree with my personal tests at all.
As a little treat, a cell phone video of our last emergency brake assistant test: We are sitting in a Model 3, waiting for the next take and two cameramen are standing around it. Nothing else happens. This is what image recognition makes of it.
PS: Subtitles are AI generated