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.
No. The other car brakes but not enough to avoid a collision. If the Waymo had not swerved, there would have been a collision. That was all on the Waymo that there was no collision.
I am just saying that there would have been a collision if the other driver had not stopped. That seems very clear from the video. Obviously if the Waymo driver had not swerved there would have been a collision regardless.

The Waymo had a green light. The other car was occluded by the white car in the adjacent lane, so it is only visible when the Waymo is already at the intersection.
Yeah. The Waymo knew the other car was there well in advance and it saw the white SUV stopped to turn (why would it stop???). Need to put two and two together and ease off. Doesn’t seem very complicated.

I think people seriously underestimate how awesome an alert and competent human driver is. So good at anticipating.
 
I am just saying that there would have been a collision if the other driver had not stopped. That seems very clear from the video.

It is hard to tell in the video, I am not completely convinced the other driver came to a complete stop. They do seem to brake and swerve a bit. Yes but even with the other driver braking and swerving, there would still have been a collision had the Waymo not swerved. So give the Waymo credit for avoiding the collision.

Also, the other driver was completely at fault for making that turn. They almost caused the collision in the first place! Not sure why you seem to be giving the at fault human driver that almost caused the collision, credit for avoiding the collision?
 
Last edited:
They almost caused the collision in the first place! Not sure why you seem to be giving the at fault human driver that almost caused the collision, credit for avoiding the collision?
I'm not. I'm just saying that if they had not stopped there would have been a collision. Has nothing to do with giving credit. It goes without saying that the other driver was at fault. It's usually good when you start making a mistake when driving to try to correct that mistake and yield right of way.

So give the Waymo credit for avoiding the collision.
The Waymo reaction avoided a collision.
They do seem to brake and swerve a bit.
This also contributed to the collision being avoided.
 
The frame rate seemed ok to me.
He's talking about the framerate of the real world video being extremely jerky, as if there are skipped frames. This makes it seem like the car is travelling much faster than it probably actually is. Compare that to the visualization in the top, it is much smoother and it doesn't feel as if the car is travelling that fast.
The Waymo had a green light. The other car was occluded by the white car in the adjacent lane, so it is only visible when the Waymo is already at the intersection. By the time you see the other car making the turn into the Waymo, there was less than 1 second to avoid a collision. The Waymo reacted very quickly. Very impressive reaction time by the Waymo Driver.

No. The other car brakes but not enough to avoid a collision. If the Waymo had not swerved, there would have been a collision. That was all on the Waymo that there was no collision.
I think he is talking about how if the turning car didn't abort, the collision would have happened anyways. Definitely the Waymo swerving also avoided a collision, but the actions of both cars contributed to help.

An alternative scenario is if the Waymo saw the car on the left slowing down, slowed down for the intersection in anticipation, and slammed on the brakes once it saw the car trying to cross (as AEB may do). In that case it wouldn't matter if the oncoming car committed or not, the accident would be avoided completely no matter what.
 
  • Like
Reactions: AlanSubie4Life
Waymo improves overall safety:




The collision avoidance clip is quite impressive.
Collision avoidance video also sucks for me, I'd estimate 10 fps. Of course it might be Twitter than sucks instead of Waymo. Or maybe Twitter's algorithms intentionally make Waymo videos look like crap :)

Crash numbers over 14.8m miles:
Injury crashes - Waymo 12, humans 42
Police-reported crashes - Waymo 32, humans 64

First, I'm surprised human police-reported crashes are only ~50% higher than injury crashes. In my experience the gap is much wider. Maybe it's skewed by a lot of miles in SF where pedestrians, cyclists, skateboarders, etc. get hit often.

Second, this includes all accidents regardless of fault. The average driver is at fault in half the wrecks, so even a perfect driver who never causes a wreck will only see a 2x reduction in two-car crash rate. But a defensive driver should be able to see >2x and Waymo should avoid single car and car-VRU crashes almost entirely. So I'm a little surprised Waymo is only 2x.

Waymo's 3.5x injury crash improvement is more like it. That may be due to defensive driving or maybe from avoiding pedestrians.
 
The Waymo was traveling 34MPH in a 35 and slowed to 10MPH.

Screenshot 2024-06-18 at 6.18.19 PM.png

Screenshot 2024-06-18 at 6.17.47 PM.png
 
  • Helpful
Reactions: diplomat33
Waymo improves overall safety:




The collision avoidance clip is quite impressive.

Kudos. For me that was stellar!

At 8 secs the sensors detect a partial return, only painting the oncoming vehicle's front. Forward motion is detected.

A little before the 11 sec mark the Waymo senses the partially detected object is turning into the intersection and Waymo path almost immediately begins to adjust.

At the end of 11 secs the oncoming vehicle is fully painted and Waymo begins to decelerate.

I'm seeing ~15mph deceleration in ~1 sec for about 0.68g force. That might be about max expected especially when simultaneously using steering to avoid an accident.

This is the type of response needed to prevent accidents. It's human like. So much happened around the 11 second mark. Spectacular sensing and processing!
 
Kudos. For me that was stellar!

At 8 secs the sensors detect a partial return, only painting the oncoming vehicle's front. Forward motion is detected.

A little before the 11 sec mark the Waymo senses the partially detected object is turning into the intersection and Waymo path almost immediately begins to adjust.

At the end of 11 secs the oncoming vehicle is fully painted and Waymo begins to decelerate.

I'm seeing ~15mph deceleration in ~1 sec for about 0.68g force. That might be about max expected especially when simultaneously using steering to avoid an accident.

This is the type of response needed to prevent accidents. It's human like. So much happened around the 11 second mark. Spectacular sensing and processing!
I didn't get those numbers (though I was measuring from the point of perception of the obstacle, not the start of deceleration, which I think is why my deceleration number was nearly half of yours). I just think it was careless driving on Waymo's part which nearly got it in an accident with a driver who had completely screwed up and incorrectly taken right of way from the Waymo, which would not have been at fault.

The deceleration is hard to measure exactly due to the atrocious frame rate and how difficult it is to consistently go frame by frame in Twitter.

Anyway to me obviously it's an accident avoided by Waymo (this is good!), but I think it's not as good as it should be. I would have been clenching hard flying into an intersection at that speed in that situation with traffic stopped in an adjacent lane for unknown reasons (which were actually known; the Waymo Driver knew there was an obscured vehicle), blocking visiblity. Doesn't anyone think this is strange and reckless driving? Put yourself in that position and "feel" what you would do. I know what I would do. Maybe people don't drive very defensively?

That was my point. I also think the reaction was surprisingly slow given the superhuman response time that should be available to the Waymo Driver. I don't know why you'd path adjust before you started bleeding a lot of speed. There must be reasons but it seems non-optimal for a human driver response.

To be clear, I don't think there is any reason to think that any competing system would do better. I just think it points to the extreme difficulty of making a system that is better than an alert and competent human driver.
 
Last edited:
  • Like
Reactions: flutas
I would have been clenching hard flying into an intersection at that speed in that situation with traffic stopped in an adjacent lane for unknown reasons, blocking visiblity. Doesn't anyone think this is strange and reckless driving?

Uh, what?! You make it sound like the Waymo was speeding like a maniac. The Waymo was only going 35 mph on a 35 mph road. That is not a high speed. No, I would not consider it strange or reckless driving to go 35 mph through an intersection on a green light.

I also think the reaction was surprisingly slow given the superhuman response time that should be available to the Waymo Driver.

What?! The other car popped out from the occlusion heading right towards the Waymo, with less than a second to react. The Waymo reacted in a fraction of second to avoid the collision and you think that is too slow and not superhuman? A super attentive human driver might have reacted that fast but most human drivers would not have.
 
Last edited:
  • Disagree
Reactions: flutas
Uh, what?! You make it sound like the Waymo was speeding like a maniac. The Waymo was only going 34 mph on a 35 mph road. That is not a high speed. No, I would not consider it strange or reckless driving to go 34 mph through an intersection on a green light.
Again, consider the whole situation. I feel like you're not even reading what I'm writing. I said nothing about the Waymo speeding like a maniac. I just said it seemed reckless.

You're beetling along with clear visibility. You see a glimpse of car beyond the white SUV (you can see this in the visualization). This vehicle disappears. You no longer have clear visibility.

The white SUV comes to a stop, presumably signaling (the video is of such poor quality it is hard to tell).

The SUV does NOT turn, in spite of there being no obvious traffic. This gives a clue. Also, the vehicle you know exists is nowhere to be seen!

At this point I have released the accelerator and am in regen before entering the intersection. Who knows what is obscured? At least bleed off a few mph!

You see what I'm saying? I would be clenching if some easing off were not taking place, ready to slam on the brakes if any vehicle or pedestrian came into view.

Intersections are dangerous if visibility is not good. Caution is advised.

What?! The other car popped out from the occlusion heading right towards the Waymo, with less than a second to react. The Waymo reacted in a fraction of second to avoid the collision and you think that is too slow and not superhuman? A super attentive human driver might have reacted that fast but most human drivers would not have.
I made it clear that I wasn't talking about most human drivers. Remember the bar is not "most human drivers." You're not going to ask 50% of people to step into a vehicle that is less safe than just driving themselves.
 
Last edited:
Again, consider the whole situation. I feel like you're not even reading what I'm writing. I said nothing about the Waymo speeding like a maniac. I just said it seemed reckless.

You described the Waymo as "flying into an intersection at that speed". Terms like "flying" and "at that speed" sound to me like you were implying the Waymo was speeding like a maniac. Going 35 mph is not "flying" in my book.

You're beetling along with clear visibility. You see a glimpse of car beyond the white SUV (you can see this in the visualization). This vehicle disappears.

The white SUV comes to a stop, presumably signaling (the video is of such poor quality it is hard to tell).

The SUV does NOT turn, in spite of there being no obvious traffic. This gives a clue. Also, the vehicle you know exists is nowhere to be seen!

At this point I have released the accelerator and am in regen before entering the intersection. Who knows what is obscured? At least bleed off a few mph!

You see what I'm saying? I would be clenching if some easing off were not taking place, ready to slam on the brakes if any vehicle or pedestrian came into view.

Yes, I see what you are saying. I just disagree that the Waymo was reckless. It was only going 35 mph. Also, you can see that the lidar tracked the occluded car the whole time. The Waymo was tracking the car. And there was no reason to slow down prior to the car turning, as the car was likely going straight. The white SUV did not turn because there was the car in the oncoming lane, that was likely going straight. The white SUV was likely waiting for the car to pass before making the turn. Prior to the car turning, there was no indication it was going to turn. But as soon as the lidar detects the "occluded" car is turning into the path of the Waymo, the Waymo reacts instantly.

I agree it is important to be cautious but you can't phantom brake at every clear intersection on a green light "just in case" a car might turn into your path, when you have no evidence that it will.

Intersections are dangerous if visibility is not good. Caution is advised.

Agreed. I just think going 35 mph is being cautious.

I made it clear that I wasn't talking about most human drivers. Remember the bar is not "most human drivers."

Yes and the Waymo's reaction time was as good as any experienced, attentive human drivers.
 
  • Disagree
Reactions: flutas
There was no reason to slow down prior to the car turning, as the car was likely going straight.
Why would the car be stopped (or moving very slowly) if it was going straight?
Yes and the Waymo's reaction time was as good as any experienced, attentive human drivers.
But the defensive driving was not and it required the driver at fault to stop in order to avoid a collision!

You described the Waymo as "flying into an intersection at that speed". Terms like "flying" and "at that speed" sound to me like you were implying the Waymo was speeding like a maniac. Going 35 mph is not "flying" in my book.
There was no sign of easing off in an obviously hazardous situation.
 
  • Like
Reactions: flutas
Waymo improves overall safety:




The collision avoidance clip is quite impressive.

It's clear that Waymo is improving safety, as we can compare their past performance to their present performance.

But I take issue with them comparing their accident statistics directly with human drivers for the entire cities of SF and Phoenix. Waymo drives a very different mix of roads as well as a limited set of roads. They would need to measure the human-performance only on their geofenced streets before they can compare the figures like that.

Tesla did this exact same marketing exercise of comparing Autopilot to the overall human rate, and it was never fully comparable.
 
  • Like
Reactions: AlanSubie4Life
Why would the car be stopped (or moving very slowly) if it was going straight?

I am saying the white SUV stopped waiting for the right gap to make the turn.

But the defensive driving was not and it required the driver at fault to stop in order to avoid a collision.

The implication that Waymo is to blame for causing the other driver to stop to avoid a collision when it was the other driver that almost caused the collision in the first place seems backwards to me. If there had been a collision, the other driver would be at fault, not Waymo. They never should have made that reckless turn in the first place! They should not have needed to stop to avoid a collision because they never should have turned into the path of the Waymo!

Look, I believe in defensive driving. But the fault is not on the defensive driver for not being defensive enough, the fault is on the other driver for almost causing the collision.
 
I am saying the white SUV stopped waiting for the right gap to make the turn.

Yeah it was obvious well in advance it was slowing and stopped to turn. Why was it stopped? Where did the other car go? Why is the other car going slowly/stopped?


If there had been a collision, the other driver would be at fault, not Waymo
Yeah, not sure why we're talking about this or why it's relevant. It was the other driver's fault. Who cares whose fault it was? The idea is to avoid a collision no matter what the other driver does.

I was just pointing out that the at-fault driver had to take action to avoid the collision.


Anyway, my point is that this seems like an absolutely terrible example of good driving. That's the last I have to say about it. I asked another person who I know who is a good defensive driver, and he thought it was awful too. I was careful not to prejudice him. Good to know I'm not insane.
Screenshot 2024-06-18 at 5.54.43 PM.png
 
Last edited:
  • Like
Reactions: flutas
Why was it stopped? Where did the other car go? Why is the other car going slowly/stopped?

The white SUV was stopped because the path was not clear to turn. You can see 3 cars in the oncoming lane. And the other car does not disappear. You see it on the lidar view the whole time.

jfs6v1B.png


And here, even though the car is occluded from the cameras, the lidar sees that the car is starting to turn into the path of the Waymo:

SqUGb28.png


and a fraction of a second later, the Waymo takes evasive action as seen by the green path starting to curve.

DlXeVjm.png


I was just pointing out that the at-fault driver had to take action to avoid the collision.

But the at-fault driver did not have to take action to avoid a collision because of the Waymo doing something bad. It had to take action because it made a reckless turn that almost caused a collision. The at-fault driver is to blame for needing to take action.
 
. And the other car does not disappear. You see it on the lidar view the whole time.
Yeah I know, that makes it worse.

Anyway, terrible example of safe driving. Absolutely awful.

The at-fault driver is to blame for needing to take action.
Why on earth are we still talking about this obvious fact? Obviously the at-fault driver was at fault.
 
  • Like
Reactions: flutas
Where I live, if you don’t “phantom brake” for every intersection where visibility is occluded, you won’t last more than a couple of weeks before you smash up your car. What the Waymo did was impressive, but at the same time, that rate of speed just seems much too high for the environment and the situation, posted speed limits or no. This isn’t the suburbs with wide stroads, protected lefts and clear visibility all around.

This is also what peeves me about FSD FWIW, the lack of caution around poor visibility just because one has a green light or no stop. Too many people run stop signs, jump across intersections in heavy traffic without checking to make sure BOTH directions are clear, turn left without checking BOTH lanes are clear, and so on. Slowing down for these situations saves my ass every single week.
 
This is also what peeves me about FSD FWIW, the lack of caution around poor visibility just because one has a green light or no stop. Too many people run stop signs, jump across intersections in heavy traffic without checking to make sure BOTH directions are clear, turn left without checking BOTH lanes are clear, and so on. Slowing down for these situations saves my ass every single week.
Judgment aside on how careful autonomous vehicles should be, one thing should be clear is that from a long distance out Waymo saw that vehicle and even those occluded from the camera's perspective. 360 degrees hypervigilance is one of the key benefits of autonomous vehicles employing surround cameras, radars, lidars and sophisticated algorithms. Companies who are taking autonomous taxi services seriously employ multiple sensor modalities to help mitigate the issue of visibility and occlusion. That is why there is a long range lidar dome on top of the vehicle.

vrw0RF8.png