Waymo Self-Driving Unit Slows Autonomous Trucking (bloomberg.com) 57
Waymo, the self-driving unit owned by Alphabet, is slowing the development of autonomous trucking that's being done by its Via subsidiary. From a report: "With our decision to focus on ride-hailing, we'll push back the timeline on our commercial and operational efforts on trucking, as well as most of our technical development on that business unit," the company said in a statement. "We'll continue our collaboration with our strategic partner, Daimler Truck North America, to advance technical development of an autonomous truck platform."
The move comes as Alphabet is prioritizing financial discipline. The company said on Tuesday that it promoted Chief Financial Officer Ruth Porat to president and chief investment officer, saying that it will stick to the more thrifty culture she has instilled. Self-driving technology has taken a step back in the past several years. Autonomous ventures like Waymo have spent billions of dollars in capital only to bring in little, if any revenue. Waymo has made more progress monetizing its robotaxi business than it has in trucking.
The move comes as Alphabet is prioritizing financial discipline. The company said on Tuesday that it promoted Chief Financial Officer Ruth Porat to president and chief investment officer, saying that it will stick to the more thrifty culture she has instilled. Self-driving technology has taken a step back in the past several years. Autonomous ventures like Waymo have spent billions of dollars in capital only to bring in little, if any revenue. Waymo has made more progress monetizing its robotaxi business than it has in trucking.
when you can't get cars right why do trucks that a (Score:2)
when you can't get cars right why do trucks that can be harder to drive / deal with stuff on the road.
Re: (Score:3)
In a lot of instances trucks can (theoretically) be much easier for autonomous driving. For example, Port-to-Wearhouse trips are generally exclusively on major arterial highways and limited access freeways, and don't pass through school zones or residential areas with bicyclers and cross walkers. There are fewer "edge cases" to work through. BUT, they also aren't all that profitable for $100 Billion and years and years of investment verses just pay a driver $25/hour to pull a rig.
Trucks are depreciated (Score:3)
BUT, they also aren't all that profitable for $100 Billion and years and years of investment verses just pay a driver $25/hour to pull a rig.
Human drivers have to take mandatory breaks and can only drive about 10 hours a day (roughly, omitting important context info). Autonomous trucks can run 24/7.
Trucks are a depreciated capital asset, so getting more hours of use from them would offset the extra cost. If you can use them roughly twice as much as a human piloted truck, you only need half as many trucks in your yard with associated savings in capital investment, real estate for parking space, garage sizes, and so on.
Re: (Score:2, Informative)
use them roughly twice as much as a human piloted truck. will put more wear and tear on them / will have higher fuel costs / etc.
Re: (Score:2)
Presumably a self driving truck would drive in the most efficient way possible. Lack of human drivers might mean they can go a bit slower to save fuel, since they don't have to arrive at the destination before the driver is too tired to continue.
Re: (Score:2)
The trucks will depreciate faster and have incur more operating costs. But if you get more freight-miles out of them in the process, you can come out ahead. Twice the miles per day, twice the fuel - sure, but if you can take a 4-day human-driven cross country trip and make it 1.5 days machine-driven, that's a win.
A big rig sitting idle at a truck stop, because the driver
Re: (Score:2)
You might not actually need as many as half as many trucks in your yard, because they would all be out doing jobs.
Obvoiusly a certain proportion of them will require maintenance at any given time, and there will be fluctuations in demand for jobs.
Re: when you can't get cars right why do trucks th (Score:4, Interesting)
I had this very discussion with a friend who drove trucks in the military (both off-road and on road) back when autonomous driving was getting massive investment because technologists thought they could get to level 5 in another couple of years after they got high on getting to level 2 in a couple of years. His take was that driving a sedan and a heavy duty truck in similar environment, truck is significantly harder and would require a lot more from the driving system.
But driving a heavy duty truck on a highway is significantly easier than driving a sedan in city traffic. So he thought that they'd probably replace the intercity work with automation, and then have drivers do the "last mile" work. Similar to how some automated tsubway systems work today, where moving locomotives around the depot is done by human, and then automation handles the actual pre-programmed route once train is prepped and set to go.
He also got into off-road with off-road military trucks, specifically that this isn't getting automated because of just how much of unique adaptation one needed to do that vs how little you need to do it. You're not going to get enough data for ML to go Big Data on it.
So driving a truck to some terminal point in the city in a nation with huge distances between cities like US automatically, and then having human drivers do the "last mile" of the delivery would bring massive savings. Less so in much of Central Europe, but probably comparable savings here in Nordics, where distances are also massive. And it would be much easier than getting full level 5 on a sedan in city traffic, but much harder than getting level 5 on a sedan on highway.
Re: (Score:2)
AI has much better situational awareness that humans. It's problem isn't in general awareness. It's in cognition. Specifically it lacks self-awareness, which is a completely different thing to situational awareness. I.e. it is much better at observing and comprehending the environment, because you can add far more inputs into it. But it doesn't understand context of any of it, and so it cannot take actions based on context. It has to brute force "learn" from relationships between each situation presented in
Re: (Score:2)
Unironically correct. It is aware of everything in said situation. It is utterly incapable of grasping what the situation is, and therefore sorting things in situation into "relevant" and "irrelevant" bins. Perfect situational awareness coupled with complete absence of context of the situation.
That is the reason why we can't get to level 5, even when we mount incredible sensor arrays that are vastly superior to human sensor package that ensure that vehicle has near total situational awareness of everything
Re: (Score:2)
Re: (Score:1)
Again, AI can have excellent situational awareness, because all that needs is accurate and numerous inputs. AI can have as many sensors as you can bolt to it and have drivers for. They can be for things humans can't even sense, like IR/UV, they can have lidar, radar, you name it, it can be done. It's situational awareness is off the charts, and eclipses a platform as limited as biological humans by nearly infinite margins if you're willing to invest in the sensor package. It's why we have long stopped putti
Re: (Score:2)
You really are an AI-like thing, aren't you?
Zero comprehension of context, coupled with Big Data spam.
Re: (Score:2)
But no self awareness to grasp why x+y does not equal x when y isn't zero in your quote.
Re: (Score:2)
For pre-determined routes they can do a high resolution scan of the entire thing, making it much easier for the autonomous vehicle to know precisely where it is and where to look for things like traffic lights. For some routes they could even have a separate lane, especially where there are tricky parts.
For long distance, they should put rail tracks along side the highways. With automated vehicles that can drive on/off, they would be a lot more efficient than roads. They could electrify them too. They could
Damn I'm too slow (Score:2)
You beat me to the "You've successfully re-invented the train" meme.
Re: (Score:2)
The issue with trains seems to be acquiring the land to build the line on. The state builds and maintains roads, and businesses can use them for free (vehicle tax excepted).
Maybe it's time the government developed a standard for drive-on rail sleds for trucks. Build lines along long highways and require haulage to use them.
Real-world example (Score:2)
The issue with trains seems to be acquiring the land to build the line on.
(That seems to be a US-specific problem. In several European countries, the state has no problem managing a good rail network. Including acquiring the necessary land).
The state builds and maintains roads, and businesses can use them for free (vehicle tax excepted).
Maybe it's time the government developed a standard for drive-on rail sleds for trucks. Build lines along long highways and require haulage to use them.
Real-world example of something similar being done: Switzerland.
Differences with your suggestion are:
- you don't load the whole truck on rail, only the cargo is transferred from truck to train (thank you, standardization in shipping containers).
- rails aren't 100% always along the highway (Switzerland maintains an excellent rail network coveri
Freight Trains? (Score:2)
I agree that separating the shipping journey into three phases - first mile, hub-to-hub & last mile - makes sense, especially as the first and last miles can tend towards electric, but why does the hub-to-hub part have to be as-yet unavailable robo-trucks?
What's wrong with freight trains? We already take this phased approach to international shipping... with ships. Why not put the containers on trains instead of trucks when you unload at the port?
Is it that trains are too old-hat, not VC/tech-bro sexy e
Re: when you can't get cars right why do trucks th (Score:2)
Most trucking runs involve a last mile component that is way more difficult in a semi than in a car. The difference between main roads and residential roads is much less than the difference between driving a 15 foot vehicle and a 65 foot one. Navigating involves things like turning into oncoming lanes with visual communication with other drivers. Most importantly for self driving, it involves technically breaking the law in many cases. Iâ(TM)m talking things things like Walmart stores, with no legal tr
Re:when you can't get cars right why do trucks tha (Score:4, Insightful)
The questions makes an assumption that isn't justified. On what basis do you say that Waymo "can't get cars right"? Because they aren't perfect yet? As Winston Churchill said, "Perfection is the enemy of progress." Waymo has made great progress. They are farther along the road towards fully autonomous vehicles, than any other company. Perfection will never come, but that doesn't mean we / they shouldn't continue developing the capabilities and types of self-driving vehicles.
"Instilled" (Score:2)
That link has nothing to do with the story (Score:3)
That link has nothing to do with the story.
Editors need to edit, man.
Re: (Score:3)
Congratulations!
You’re literally the only person who attempted to RTFA!!
There are few industries that play the long game (Score:1)
Fracking comes to mind as one of the few long term private investments that truly paid off. It was know that vast quantities of oil and gas were trapped in shale formations for decades, and it took decades and decades of research and development for that investment to pay off.
Driverless vehicles is like that. Everyone knows how big an economic win it would be for the first company to crack it. But how much investment can you pour in to make that happen? It's an interesting investment problem: unlimited inve
Re: There are few industries that play the long ga (Score:2)
Re: (Score:1)
That's exactly the point of Alphabet's Other Bets. Research paid for by Search profits. The question is whether Ruth sees the long-term value, or just the short-term cost.
Given the preponderance of evidence, I think it is almost certainly safe to say that she sees everything in terms of short-term gains and losses, with no real consideration to anything more than a quarter away.
Nobody with a long-term outlook would lay off the entire Area 120 R&D team, eliminating basically all research that isn't expected to make a short-term profit.
Similarly, nobody with a long-term outlook would blanketly lay off 12,000 people knowing that they would hire half again more people than t
And economic win for who though? (Score:2)
It's entirely possible self-driving cars is going to be more like fracking in that sense. There's 3.5 million truck drivers and another 300,000 taxi cab drivers. That's a noticeable chunk of
Seems shortsighted (Score:2)
I’m gonna guess that the peeps at google think trucking is too boring and old school industry to bother with. If it doesn’t involve 8 second videos, sells ads and looks like an apple p
Re:Seems shortsighted (Score:5, Informative)
Long-haul, drive-24-hours-in-a-nearly-straight-line autonomous trucking is a way, way, WAY easier nut to crack than 800-sharp-turns-and-change-lanes-every-5-seconds autonomous consumer city driving. Also, trucking is a near-trillion-dollar business. Whoever gets that one first will mint money for quite a while.
I’m gonna guess that the peeps at google think trucking is too boring and old school industry to bother with. If it doesn’t involve 8 second videos, sells ads and looks like an apple product, who needs it, right?
I think one big issue is liability. Urban traffic accidents are usually non-fatal, and one thing self-driving cars. But given the mass and velocity of a big rig on a highway an accident involving another vehicle is probably fatal.
But the bigger issue might actually be technical. Highway driving seems easier, but the one thing self-driving cars do really well in an urban setting is breaking. Esp, with LIDAR if something gets in their way they can stop with inhuman reaction speed and reliably avoid the collision.
But in good road conditions a semi takes 5.5 seconds to stop [schneiderjobs.com].
That's a REALLY long time. That means the driver of the semi needs to recognize the stopping condition 5.5 seconds or 160m ahead of time, that's no longer reaction, that's a judgment call of road conditions, other driver behaviour, animals, etc, etc. And remember all those hypotheticals about the self driving car having to decide who to kill? We don't hear about it anymore since it turns out that passenger vehicles can stop quickly enough for the answer to be "no one".
But if an 18-wheeler self-driving truck realizes that there's a pile-up 100m ahead and it doesn't have time to stop... well now you've got some decision making to do.
Re: (Score:2)
The standard for automotive LIDAR is 300m, so seeing 160m ahead is no problem. Especially on a truck where the LIDAR can be mounted high up to see over things.
If you watch some of Waymo's older videos, you can see that a huge amount of effort went into predicting what other road users would do. Not just other vehicles, pedestrians and cyclists too.
Re: (Score:2)
The standard for automotive LIDAR is 300m, so seeing 160m ahead is no problem.
Sure, but straight ahead of you isn't the only source of danger. There's a lot of other stuff that may be extreme danger, or may be completely innocuous.
If you watch some of Waymo's older videos, you can see that a huge amount of effort went into predicting what other road users would do. Not just other vehicles, pedestrians and cyclists too.
That means they recognized the problem, that doesn't mean they solved it.
Re: (Score:2)
LIDAR doesn't just look ahead. On Waymo vehicles it looks all around.
Re: (Score:2)
LIDAR doesn't just look ahead. On Waymo vehicles it looks all around.
It's not about whether the LIDAR can see everything.
It's about the AI recognizing that a currently safe situation will become dangerous in 5.5 seconds.
Re: (Score:2)
Long-haul, drive-24-hours-in-a-nearly-straight-line autonomous trucking is a way, way, WAY easier nut to crack than 800-sharp-turns-and-change-lanes-every-5-seconds autonomous consumer city driving. Also, trucking is a near-trillion-dollar business. Whoever gets that one first will mint money for quite a while.
I’m gonna guess that the peeps at google think trucking is too boring and old school industry to bother with. If it doesn’t involve 8 second videos, sells ads and looks like an apple product, who needs it, right?
I think one big issue is liability. Urban traffic accidents are usually non-fatal, and one thing self-driving cars. But given the mass and velocity of a big rig on a highway an accident involving another vehicle is probably fatal.
But the bigger issue might actually be technical. Highway driving seems easier, but the one thing self-driving cars do really well in an urban setting is breaking. Esp, with LIDAR if something gets in their way they can stop with inhuman reaction speed and reliably avoid the collision.
But in good road conditions a semi takes 5.5 seconds to stop [schneiderjobs.com].
That's a REALLY long time. That means the driver of the semi needs to recognize the stopping condition 5.5 seconds or 160m ahead of time, that's no longer reaction, that's a judgment call of road conditions, other driver behaviour, animals, etc, etc. And remember all those hypotheticals about the self driving car having to decide who to kill? We don't hear about it anymore since it turns out that passenger vehicles can stop quickly enough for the answer to be "no one".
But if an 18-wheeler self-driving truck realizes that there's a pile-up 100m ahead and it doesn't have time to stop... well now you've got some decision making to do.
Its not unusual for a tired or distracted driver to take even longer.
Driving isn't going to be autonomous for a long time, haulage will be the last to convert for no other reasons than liability and schedules (fail safe option for an autonomous vehicle when it doesn't know what to do will be to stop) so a load sitting on a highway doing nothing is not only a huge risk of causing a major collision and major traffic problems, but it's also a load that needs to be somewhere and burning money out of the haul
Re: (Score:1)
But, here is the problem with driverless trucking: The driver isn't actually that big of a portion of the total cost. A company has to front billions and billions and billions of dollars in research investment up front in order to oust a guy making $25 an hour. AND you have to convince companies to buy new trucks or heavily modify their current fleet. AND many trucks are actually owned by the drivers themselves.
Ultimately, you have to make a lot of trips to pay off the initial investment.
Re: (Score:2)
Ultimately that is true of many industries, including agriculture. It's cheaper to pay a guy a decent salary than to invest in self-driving tractors and such. Although it's getting harder to find people willing to do that sort of work. It's cheaper for me to hire someone to walk my fields every few days monitoring moisture requirements and disease or insect pressure than it is to buy remote sensing to automate this task for me. Goes on and on.
Technology and automation is really neat, but if you step bac
Re: (Score:2)
At $25/hour, assuming 12 hours/day for the trucks(could be much higher), that's ~$110k/year of potential savings. Don't forget that due to benefits and such, a dude making $25/hour probably actually costs $35/hour.
And $100k is around the cost of a new semi-truck a year.
I'd argue that yes, it actually IS a big cost.
~$100k for the truck ($25k/year in capital costs, maybe?)
~$70k for fuel
~$100k on wages
Let's go with $30k for maintenance.
I'd argue that at roughly 1/3rd of operating costs, maybe more, it's still
Re: (Score:2)
Not quite. Try more than double that cost for a new truck.
Re: (Score:2)
100k is the middle of the range when I searched for "price of new semi"
Re: (Score:2)
Last quote I got for a bare bones day cab truck was $200k CAD, delivered in about a year. A fully-loaded highway truck with sleeper is much, much more than that. Significantly more.
Re: (Score:2)
Yeah, but that's CAD - you guys nearly always pay more.
I'll admit that given recent inflation $150k USD would be closer
Doesn't change the figures significantly though. Labor is still a significant cost.
Re: (Score:2)
I can't drive 45! (Score:2)
I can't drive 45!
Re: (Score:2)
Long-haul, drive-24-hours-in-a-nearly-straight-line autonomous trucking is a way, way, WAY easier nut to crack...
In fact, this problem was cracked more than 100 years ago - it's called the freight train.
reason for the slowdown (Score:2)
The trucking industry is busy enough with just electrification. Compared to the difficulty of that, autonomous driving looks like pie in the sky. They'll come back to it in 15 years.
The Easiest-problem First Rule (Score:2)
Any engineer worth their salt knows that you don't go trying to solve the hardest problem first, if you want your company to survive. You start with he easier steps first.
Both a truck and a car have nominally the same sub-sytems for detecting surrounding obstacles, just with