How we’ve built the World’s Most Experienced Urban Driver

https://blog.waymo.com/2021/08/MostExperiencedUrbanDriver.html

by alphabetting • 5 years ago

40 58 5 years ago

blog.waymo.com

andyxor• 5 years ago

and probably the most expensive, I see them going in circles all day long in my neighborhood, fuel alone must have cost a fortune

cinntaile• 5 years ago

It's the human driver that costs money. Fuel is a very tiny part of that, especially in the US (compared to the EU) where fuel is practically free.

jimbob45• 5 years ago

I’m from the US and I don’t know why you’re saying that. Is EU fuel comparatively more expensive?

wmf• 5 years ago

What Waymo doesn't talk about is their human remote drivers and chase vehicles. I wonder what their ratio of humans to "autonomous" cars is.

mtwshngtn• 5 years ago

Was it a Jaguar you saw? because those vehicles are electric.

jeffbee4• 5 years ago

Even an electric car costs 11¢ per mile according to AAA. So that's $11,000/week to drive 100,000 miles per week. This may be "fortune"-level spending by some standards.

criticaltinker• 5 years ago

We’re building systems that operate without any reliance on human drivers. Either the system can complete the entire trip or it cannot. There is never any confusion about who is in control and responsible for the safety of driving. So we focus on the hardest parts of the challenge, which is critical to reliably scaling vehicles with no human drivers.

Wow and really impressive demos to support their claims. Waymo's tech seems likely to surpass the capabilities of Tesla and others if it hasn't already.

Has there been any progress on an open test or benchmark that would reliably highlight the differences in performance and degree of autonomy provided by these competitors?

gundmc• 5 years ago

Waymo's tech seems likely to surpass the capabilities of Tesla and others if it hasn't already.

Waymo has been the consensus tech leader in this space for 10 years.

jeffbee4• 5 years ago

Yes, Tesla would have to revolutionize their whole approach to be able to do any of the things that Google/Waymo already demonstrated prior to 2011. For Waymo to "surpass" Tesla we'd have to be living in Oppositeland where Tesla actually possess self-driving car technology.

CorrectHorseBat3• 5 years ago

Looking at a fragment another recent video I don't feel like it's really that great yet: https://www.youtube.com/watch?v=yjztvddhZmI&t=644s

Yes it's amazing what they can do, but if after all this time they can't even react correctly in such a simple situation I really wonder if full self driving is feasible.

Crash0v3rid3• 5 years ago

What exactly was wrong in the example clip you provided?

To me, it looks like the car rightfully stopped because it assumed the person was going to cross the parking lot.

CorrectHorseBat3• 5 years ago

The person was clearly not going to cross right there at the corner so such an abrupt stop was totally unnecessary. If the car assumed he was going to stop a bit further at the yellow lines it should have either braked earlier or just keep on driving.

The way it reacted it IMO just looks like clueless new driver.

deadmutex• 5 years ago

can't even react correctly in such a simple situation I really wonder if full self driving is feasible

I don't know if you can extrapolate from "can't react correctly in such a simple situation" for a few reasons. Fristly, I will point out that car stopped, but the complaint was that it was little bit more jerky than preferred. Also, automated drivers will excel and fail at different scenarios. Imagine the other way around: Imagine if Waymo's tech was ready for all roads, and we were evaluating letting humans drive in manual. One could say things like:

"I saw a human apply the break suddenly on the highway because they were checking their mirrors for a lane change when the car in front of the slowed down. The human had a delayed reaction because they were checking their mirrors. An automated driver would be able to analyze the lane change, and the traffic in front of them at the same time. Also the reaction time of the automated system is superior than humans... so I wonder if human driving is feasible."

I think when comparing FSD, one needs to compare it to human driving at some percentile, and not perfection :).

CorrectHorseBat3• 5 years ago

An automated driver would be able to analyze the lane change, and the traffic in front of them at the same time. Also the reaction time of the automated system is superior than humans...

And yet here we see that the car either has a very slow reaction time or doesn't understand how pedestrians move around. That's what concerning me. Looking at the video again maybe, maybe the car thought the cart would roll further on the road when making the corner? That would justify the sudden reaction.

It's also the only interaction with a pedestrian in the video and I see no interactions with bikes. I'd say that if they were able to deal with those they would show it.

gerash• 5 years ago

I agree that completely solving self driving is not feasible but a constrained version is.

That said, there is also the factor of the car brake system. My hybrid car has a fairly rough transition from regenerative brake to the friction one which sometimes results in such sudden stops

Causality1• 5 years ago

That may not be such a bad thing. When self-driving cars are better than a human driver in every single way, car companies have no reason to sell the cars instead of adding them to a massive robo-taxi fleet. A nearly-uncrashable car that handles everything except the first and last sixty seconds of every trip is close to the ideal outcome for consumers, in my opinion.

tibbydudeza2• 5 years ago

If it can handle the traffic in Delhi then I will be impressed.

https://www.youtube.com/watch?v=mDLCd70iTeE

marai2• 5 years ago

The first 30 seconds are slow but after that crossing the intersection gives me a new appreciation for the human brain :-)

a9h74j• 5 years ago

Beautiful. Take that, Mr. Hollywood montage director!

tgsovlerkhgsel2• 5 years ago

It doesn't seem that much harder. The difference between normal car driving and that seems to be a much lower speed, and a much higher tolerance to get close to pedestrians.

Route finding will be a bit more difficult, but other than that, it could probably be approximated as "if (closest_obstacle_in_front.distance > 30 cm) { drive(SPEED_SNAIL_PACE); } else { honk(); }"

tibbydudeza2• 5 years ago

Or this roundabout in the UK

https://www.youtube.com/watch?v=6OGvj7GZSIo

rhacker3• 5 years ago

It took a while, here's the video I watched that made me not want ANY Waymos near where I live:

https://www.youtube.com/watch?v=zdKCQKBvH-A

I think the thing that made me so furiously mad is that there's all these situations in the video that people might laugh at, but there are situations where I've been stuck in traffic in an emergency and if I found out it's because Google was trying to make it's next pentillionth dollar I would blow my top.

Can you imagine one of these during the Campfire evacuation?

UncleMeat• 5 years ago

https://www.youtube.com/watch?v=GiYO1TObNz8

One can think of all sorts of tail cases where some better technology is worse, but we should be careful not to let the tail cases cloud our overall analysis of a technology.

"I might not be able to escape if my whole town was on fire" is also an argument against virtually all public transportation - but clearly should not be used as a reason to disregard public transportation.

alphabetting2• 5 years ago

Totally agreed. That youtube channel op linked to was my main source of following waymo until he stopped. There's close to 100 full driverless rides and nearly all are normal. The one that has a hickup got maybe 1000x more coverage than any other ride but that's how it goes. I liked how the guy, despite clearly being a waymo fan, makes sure to show an unvarnished view at the state of the tech in contrast to some enthusiasts who mislead or overhype the tech they cover.

jeffbee4• 5 years ago

Imagine believing that escaping a fire by car would ever work. Of the 25 people who died in the 1991 Oakland fires, many were killed when the fire overtook them in their cars in a massive traffic jam. Mass evacuation is when the colossal stupidity of car-based transportation is most clearly evident.

rhacker3• 5 years ago

Well, 26000 people could have walked out of Paradise, but it would have been a lot worse.

nradov2• 5 years ago

The workers repairing potholes in my neighborhood rely on vague and confusing hand gestures to direct traffic. They don't even have those little stop / slow signs. I wonder how Waymo would handle such situations?

thepasswordis2• 5 years ago

It’s the workers that need to adapt.

Self driving cars are here and they’re not going away.

toiletaccount• 5 years ago

lol, you havent worked with road workers

if a self driving car cant handle something like a guy in an orange vest directing traffic then its a very questionable venture even for scroogle or crapple. unexpected things happen on the road and corner cases happen all over, every day.

908B64B1974• 5 years ago

Have someone somewhere remote-control the cars when they get on a construction site.

nradov2• 5 years ago

A lot of roads don't have reliable, high speed cellular data service. There are major dead zones just a few miles from my house.

contingencies• 5 years ago

We've built the world's most expensive Uber driver... in misreading, my subconscious espousing wisdom.

imapeopleperson• 5 years ago

The hubris of the title distracts from the achievement. Still curious how it fairs pumping gas and driving in the snow; two simple things many urban drivers can do.

How we’ve built the World’s Most Experienced Urban Driver

Related Stories