Obviously this only matters if the passenger can feel it (they probably can't). This is certainly an example where an L2 system needs to be better than an L4 vehicle as that would be super annoying if you were holding the steering wheel.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Waymo
- Thread starter Daniel in SD
- Start date
AlanSubie4Life
Efficiency Obsessed Member
I imagine this specific issue with Waymo has been fixed or reduced now.
But I am glad someone has finally responded and sees my argument (probably others have made it too, before me) that (useful) L2 has to be be better than L4 in many ways. Counterintuitive, but true!
powertoold
Active Member
It'd be great if we could see unedited videos of Waymo driving with a safety driver in challenging SF areas, like we see all the time with FSD beta.
All the available Waymo 1x videos are in simple square intersections and suburban areas.
All the available Waymo 1x videos are in simple square intersections and suburban areas.
Why with a safety driver?
If Waymo ever opens up driverless rides to the public in San Francisco you should go there and make an unedited video for us.
powertoold
Active Member
Because Waymo avoids these more challenging areas.
If you disagree, feel free to post some Waymo videos of what you think is a challenging SF area.
Ah, I see. I think your best bet to get an unbiased video is to follow a Waymo around.
diplomat33
Average guy who loves autonomous vehicles
Waymo does not avoid more challenging areas. These videos are not 1x but they show over 40mn of Waymo doing driverless in SF in more complicated scenarios. These are not simple square intersections and suburban areas:
willow_hiller
Well-Known Member
We know they geofence around construction zones, at the very least.
If you recall from this JJRicks video:
The safety driver says: "I don't know why this construction zone wasn't taken off the map."
diplomat33
Average guy who loves autonomous vehicles
That was a year ago with the 4th Gen Pacificas. I don't believe that is true anymore. AFAIK, the 5th Gen I-Pace don't avoid construction zones. Waymo has even mentioned that the 5th Gen can handle construction zones. They even posted clips of the 5th Gen handling different construction zones.
I'm excited for Waymo's success. As long as they can keep the costs down per ride, they will dominate the market. The younger generation doesn't want to drive as much as evidence by the rising age of initial drivers license applications.
Let's do some math to see how it will work:
Current average cost of a new car - $47,000. Let's assume some down payment, so the cost is $30,000.
Loan over 60 months - $500/mo car payment
Let's assume the standard 12,000 miles per year milage
Gas average cost per gallon - $3.80
Average fuel economy - 35 miles per gallon
Total gallons used per month - 29 - $110/mo
Average cost of car insurance - $104/mo
Total cost per month to own and drive a car - $714
This does not include maintenance and wear & tear
The average cost for Uber per mile varies widely between $1 and $2/mile. Let's take the middle - $1.50. Uber would cost about $1500 per month to do all the person's driving.
Waymo would need to come in under $1/mile to make economical sense for people to abandon their cars in favor of robotaxi. At $0.70/mile, it's a wash!
Let's do some math to see how it will work:
Current average cost of a new car - $47,000. Let's assume some down payment, so the cost is $30,000.
Loan over 60 months - $500/mo car payment
Let's assume the standard 12,000 miles per year milage
Gas average cost per gallon - $3.80
Average fuel economy - 35 miles per gallon
Total gallons used per month - 29 - $110/mo
Average cost of car insurance - $104/mo
Total cost per month to own and drive a car - $714
This does not include maintenance and wear & tear
The average cost for Uber per mile varies widely between $1 and $2/mile. Let's take the middle - $1.50. Uber would cost about $1500 per month to do all the person's driving.
Waymo would need to come in under $1/mile to make economical sense for people to abandon their cars in favor of robotaxi. At $0.70/mile, it's a wash!
willow_hiller
Well-Known Member
Best I can find is a short clip from a Waymo employee where the route happens to be near some construction:
But I can find no recent videos showing what a Waymo driver would do if it unexpectedly found the lane it needed was blocked by construction.
Doggydogworld
Active Member
When possible they update the map to show that a lane is unavailable. Why wouldn't they?
If the map is not updated, as in this case, the car attempts to follow the signs. If it's not almost 100% certain, it phones home. When Joel's car phoned home the remote monitor incorrectly told the car to drive in the rightmost lane. Things quickly fell apart from there.
diplomat33
Average guy who loves autonomous vehicles
Well, presumably, it would automatically reroute. That's what an AV would need to be able to do. And Waymo is building an AV to handle anything without a safety driver.
willow_hiller
Well-Known Member
I'm not saying they shouldn't, I'm saying it's something worth remembering when comparing between Waymo and Tesla disengagements.
FSD Beta drivers are constantly putting their vehicles in difficult situations, occasionally on purpose, so the disengagement rate is higher than it would be if Tesla chose to geofence FSD Beta away from construction or other difficult road conditions.
People still use this old video? Imagine if people were pulling up videos from FSD beta from almost 2 years ago…
willow_hiller
Well-Known Member
We don't need to pull up old FSD Beta videos because there are dozens of videos on YouTube for any given firmware version in any given situation. I bet you can find 5 videos on how 10.69.1.1 handles cones in 60 seconds.
Meanwhile, if you can find a video of the 5th gen Waymo driver routing around construction, I'd be interested in seeing it.
You don't need to search for difficult situations.
Tesla fails and requires disengagement at areas Waymo is driverless in, its just that simple. Why make it hard?
And there are more videos. Tesla fans are going to be shocked at what happens in the next 24 months if all they are doing is grasping at an almost 2 years old video.
willow_hiller
Well-Known Member
Thanks for wasting my time. Neither of those videos showed the Waymo vehicle anywhere near construction zones.
And I'm not even using that JJRicks video to talk about the failure exhibited in that video. I'm using it as proof that Waymo had, and presumably still has, a policy of geofencing their vehicles away from construction zones. A Waymo employee said "I don't know why this construction zone wasn't taken off the map."
Until you can show me video proof of a 5th Gen Waymo vehicle substantially navigating through, or needing to on-the-fly reroute around construction, the only conclusion I can draw is that it's still Waymo policy to meticulously map and geofence their vehicles away from construction.
powertoold
Active Member
I've basically watched all the Waymo videos in SF. They're driving 98% of the time in the SF suburbs / "rural" areas (which look different than what you'd consider a suburb in a normal city).
To keep it simple, lets assume the above cost structure reflects Uber drivers vehicle with the exception that Uber driving 3x the amount above. With that and some wear and tear, car costs would be ~ $1000 / month for around 3000 miles.
Let's say, out of 3000 miles driven 2500 miles would be billed at $1.5 / mile average. This yields $3750. Assuming Uber driver earns ~ 60% of it: $2250. After car costs, the driver would net $1250.
With Waymo, lets say the driven miles is 12k/m and billed 10k/m. Wrongly assuming the ICE car, car costs would go to ~ $2000 / month. Also, sensor suite is far from being free. Assuming $0.75 / mile pricing, gross revenue would be $7500 / car / mile. Deduct direct car costs of $2000 and some car services like daily wash and inspection, Waymo could still see up to $5000 gross margins per car per month with pricing that is 50% below Uber.
Many assumptions above are very inaccurate, but this napkin calculation should still demonstrate that robotaxi business could be insanely profitable even and push human drivers from any market overnight. This is due to two fundamental reasons:
- No compensation for drivers
- Utilization rate can be very high for cars driven by robots
In practice every robotaxi company will want to drive electric to keep maintenance and energy costs low. So some additional adjustments should be done.
Similar threads
- Article
