Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Waymo

This site may earn commission on affiliate links.
Another interesting thing from today.

Stop signs printed on T-shirts are an attack vector for Waymo vehicles (not that I'm surprised, this is one of those "how the hell do you handle this" scenarios). Also gives me a fun experiment to try on FSD (not a printed stop sign shirt though).

This is such a controversial statement by MT

and Tesla’s Full Self Driving (FSD)—that really isn’t self-driving at all—
 
It looks like this Waymo got charged with some high-proof electrons:


Surely, any human would get pulled over for driving under the influence if they drove like that.

It appears that @diplomat33 may be correct, and that they abended their HD maps, but don't have the replacement fully dialed-in yet. They certainly aren't driving on rails anymore.
I have to wonder if the odd combination of a tree and a truck driving down the road blew Wamo’s little NN mind.
 
  • Like
Reactions: spacecoin
Edgecase: I didn’t realize how important this is as a passenger.

Ahhhh … safety amd smoothness is the most critical. I have had a chauffeur for 20 years who had perfected the art of driving making you feel like you are in a Rolls Royce. Once you have that, you will not care for how it routes and how mnay times it changes lanes, and why it stops on a yellow light. Nothing will matter.
 
I have to wonder if the odd combination of a tree and a truck driving down the road blew Wamo’s little NN mind.
That was my first reaction. It may have been identifying the tree as an obstacle rather than a vehicle, and the NN kept figuring that it had to go around it rather than follow it. But the "obstacle" kept changing position and it kept going through the same little decision-making loop. This could explain it moving into the bike lane (and over the double yellow) to try to see if it is safe to get around it.

This may be an example of one of the difficult 9's in the March of 9's - there will always be another scenario the training set has never seen before. A human can easily see that it's a tree inside a trailer attached to moving vehicle, but maybe the Waymo couldn't work that out - NN's are not capable of reasoning.
 
  • Like
Reactions: Ben W and spacecoin
Not only Waymo under investigation...

"News of the investigation comes hot on the heels of a separate review of Zoox’s autonomous driving technology.
In that case, two autonomous Toyota Highlanders were involved in collisions after they “unexpectedly braked
suddenly” and were rear-ended by motorcyclists.

 
Yes, I think this is another example that supports my thesis. If Waymo were driving "on rails" on the HD map, we would expect it to just stick dead center in the lane. The fact that it ping pongs in the lane is because the ML planner is given more freedom to maneuver. In this case, it looks like the ML planner wanted to pass the trailer. Presumably, the trailer was moving a bit slow and also the tree in the back was blocking the Waymo's sensors from seeing ahead. In any case, I think the ML planner wanted to get around the trailer. That ping pong behavior definitely looks to me like the planner was like "can I pass now? No. Can I pass now? No. Ok, can I pass now? No etc...". It saw that the right lane was a bike only lane so it was illegal to pass, but it kept looking for another opportunity to pass.

It was definitely awkward driving. Hopefully, it is straightforward to fix by just training the driving policy not to try to pass, especially if it knows it can't because the right lane is bike only. And yes, hopefully, we see Waymo retrain the ML planner to fix these issues. If I am right that Waymo is trying to be less dependent on HD maps and more reliant on the ML planner, then hopefully, once they fully "dial in" the planner, we should see a big leap in driving that is more generalized.
I think the perception might classify the object as a tree and not a trailer which would explain why it wants to go around it all the time. But the stupid tree is always one step ahead....
 
  • Like
Reactions: Daniel in SD
I think the perception might classify the object as a tree and not a trailer which would explain why it wants to go around it all the time. But the stupid tree is always one step ahead....

Yes that is the other theory. If true, then the answer would be to train the perception stack to understand this case. We've seen research examples from Waymo that deal with this type of case of a composite object (one object inside or connected to another object). So they are aware of the problem but this could be an edge case.
 
Not only Waymo under investigation...

"News of the investigation comes hot on the heels of a separate review of Zoox’s autonomous driving technology.
In that case, two autonomous Toyota Highlanders were involved in collisions after they “unexpectedly braked
suddenly” and were rear-ended by motorcyclists.

Interesting that we are seeing these reports.

Waymo and Zoox have very tiny fleets - and yet we hear about Tesla phantom braking all the time but rarely actual accidents.
 
Waymo and Zoox have very tiny fleets - and yet we hear about Tesla phantom braking all the time but rarely actual accidents.

That's because Tesla FSD is supervised so there is a safety driver to prevent a collision. I've had phantom braking that would have resulted in a collision had I not immediately intervened.
 
That's because Tesla FSD is supervised so there is a safety driver to prevent a collision. I've had phantom braking that would have resulted in a collision had I not immediately intervened.
But we also see so many reports of - "if there was someone at the back there would have been an accident".

So, I wonder, is FSD less likely to brake if there is a vehicle coming at the back. Since FSD would calculate the risk weighted probability of an accident, if there is no one at the back - chance of a rear ending becomes near zero. So, even a small chance of a forward collision - because the perception for whatever reason thinks there is something in ahead with say 10% chance - is handled by FSD braking. But in the same situation if someone is at the back, it might calculate a rear-ending chance as being 50% and thus won't break (or may be not as hard).

This is somewhat like when people complain about crossing the line. "If a vehicle was there at that it there would have been an accident" - but if there was a vehicle, FSD wouldn't have likely crossed the line !
 
Yes that is the other theory. If true, then the answer would be to train the perception stack to understand this case. We've seen research examples from Waymo that deal with this type of case of a composite object (one object inside or connected to another object). So they are aware of the problem but this could be an edge case.
Might not even be a tree classification issue, just a general problem of a "drive around" target that moves. The incident where the Waymo drove on the wrong way may be similar, where it identified a moving object as something it needed to drive around, except that object moved with it, which kept the Waymo in the wrong lane.
 
It looks like this Waymo got charged with some high-proof electrons:


Surely, any human would get pulled over for driving under the influence if they drove like that.

It appears that @diplomat33 may be correct, and that they abended their HD maps, but don't have the replacement fully dialed-in yet. They certainly aren't driving on rails anymore.
Thats hilarious.

I definitely want to see what FSD would do. Its one of those "edge" cases where Tesla theoretically has an advantage.
 
With the dust settling on the Chinese EV tariffs announcement, something that I'm curious about.

Will Waymo's new vehicles be hit with the tariffs?

Since they are designed by Zeeker and made by Geely in China as well as being all electric, one would assume they are impacted as well and may push Waymo to use a different manufacturer or vehicle entirely.

For those unaware: the tax rate on imported Chinese EVs will rise to 102.5% this year, up from 27.5%
 
With the dust settling on the Chinese EV tariffs announcement, something that I'm curious about.

Will Waymo's new vehicles be hit with the tariffs?

Since they are designed by Zeeker and made by Geely in China as well as being all electric, one would assume they are impacted as well and may push Waymo to use a different manufacturer or vehicle entirely.

For those unaware: the tax rate on imported Chinese EVs will rise to 102.5% this year, up from 27.5%
Does sound like it. I know someone with some extra American made AVs

1715813799933.png