Tesla: Failure of the FSD's degradation detection system [pdf]

"In the crashes that ODI has reviewed, the system did not detect common roadway conditions that impaired camera visibility and/or provide alerts when camera performance had deteriorated until immediately before the crash occurred."

Does it not detect them at all, or fail to deal with detected sensor degradation adequately? Does "Full Self Driving (assisted)" slow down under conditions of poor visibility?

Does Tesla even look for the road surface? One big advantage of those up-top LIDAR units is that you have a good scan of the pavement ahead. If you're not sensing flat pavement ahead, don't go there. That's basic. Vision-only systems, going back to Mobileye, have been overly dependent on looking for known kinds of obstacles. Original Mobileye could only detect car rear ends.

"Cameras only" is a cost cutting for profit only feature that is subject to Wile E. Coyote attacks.

It is a shameful engineering design to leave out LIDAR and it has cost human lives.

Let's hope Musk does not leave out something important for the moon landing. His proposal for it is absolutely ridiculous, it looks like a children's book fantasy and many smaller top-heavy craft have already toppled on the moon!

Not too much to specifically take away yet, but appears that the degradation detection system did not function well. That is pretty egregious for FSD given a human won't be able to tell if FSD is confident or needs the human to intervene. I'd expect this to be a VERY important test case with high reliability of passing, but who knows.

Overall, yikes.

Tesla is a premium product - if someone is going to use FSD they know its a luxury feature and should pay for the most comprehensive safety features available which in my mind would of (of course) require lidar

Ironic to see this on the front page just next to the report about Waymos being 13 times safer than humans.

>In each of these crashes, FSD also lost track of or never detected a lead vehicle in its path.

Oh good, Tesla vehicles apparently struggle with the task of "Hey, there's a car there" in degraded conditions.

Probably don't need to worry about that while driving though.

>Tesla also described internal data and labeling limitations that prevented a uniform identification and analysis of crash events with the subject system engaged. ODI believes this limitation could have led to under-reporting of subject crashes over portions of the defined time-period.

I thought Tesla was a "Software" company!

This report is insanely vague though. It's very preliminary, opened yesterday.

This moved from the top spot on the front page so fast

[dead]