Assuming the truck must have a data link to its base, it could be equipped with a UAV that could be launched and operated remotely to place triangles or flares as needed, more rapidly and safely than a driver could. A possible alternative might be an array of mortar tubes that could project folded warning devices in a predetermined pattern, with the devices unfolding on impact. Just two instant ideas; surely a good engineering study would quickly come up with better.
I assume they've thought about options like this and decided not to do them. I asked to talk to someone at Aurora about this and they declined to share more than their boilerplate statement, so I'm not sure what they've been thinking.
> Robotaxi companies like Waymo get a lot of attention, but economically speaking, trucking is far more important. The trucking industry had revenues of almost $1 trillion in 2023, an order of magnitude larger than the $80 billion consumers spent on taxis and ride hailing apps that year.
This makes about as much sense as saying that the development of reliable fusion power will have next-to-no economic impact because the US currently spends <$3 billion a year on fusion.
The majority of personal driving happens in personal vehicles because it's cheaper and more convenient. If Waymo sufficiently reduces the cost of ride-hailing, people will switch over to it instead. The actual economic import of driverless taxis is relative to the entire amount spent on non-recreational driving, (plus alternatives like biking, walking, and public transit), which is much more than $1 trillion.
I agree that in the long run self-driving technology will come to passenger vehicles and that will have a much bigger impact than on taxis. Whether you want to consider that to be an extension of the taxi market or its own separate market seems like a question where reasonable people can disagree. And don't forget that the induced demand point could apply to trucking as well.
That's fair, although that TAM is a LONG way off. I think Waymo needs to prove they can compete with Uber first at scale. And then they can go after personally owned vehicles, which would have much lower utilization and thus a longer payback period.
People *buying* a driverless car has a longer payback period, sure. But the majority of personal cars are unused for the majority of the time, which is a complete waste. If "Uber but only 10% the price and wait times" existed, many people would just sell their cars and use only ride-hailing.
I was writing articles predicting the end of personal car ownership a decade ago ( https://www.vox.com/2014/5/28/5758560/driverless-cars-will-mean-the-end-of-car-ownership ) but I feel less confident in this conclusion today. People like to have their own cars so that they can have car seats in the back, golf clubs in the trunk, etc. Probably those preferences will shift as cohorts age and driverless technology gets cheaper, but I think this is going to be quite a slow process.
And I don't think 10x cheaper is going to happen any time soon. Labor is half the cost of the taxi, and not all of that will go away—you still need people to clean, repair, and charge the vehicles. So they will absolutely be cheaper than today's Ubers but I'm guessing it'll be more like 2x cheaper a decade from now rather than 10x.
That idea did occur to me. I would guess it would be complicated and expensive to redesign the truck to add a place for the drone to live inside the truck and automatically get released. Also they might not want to set a precedent that would lead to them having to keep those drones around for the next decade.
As an engineer it seems implausible to me that there would be any great difficulty in accommodating and launching a UAV. However, I see it as pretty likely that the multitube mortar would prove superior. Similar systems are widely used to launch various protective payloads from military vehicles, and it would probably be possible to buy one off the shelf. The foldup device surely poses no great engineering challenges.
My main counter-thought too was precedent, but it really feels like it wouldn't be *that* difficult, and definitely worth exploring if the triangles are adjudicated 100% required.
My view is that this isn't about the technical details behind the triangles, it's a matter of principle that it's a somewhat silly rule designed for humans that having to work around sets a bad precedent for; I think they're just playing hardball here.
If the trucks are driving the highway and so the rescue drivers can travel 10 miles — *in either direction* — in 10 minutes, wouldn’t you just need someone every 20 miles? So that’s 10 people over 200 miles.
It's a fair question. A rescuer can't cross lanes of traffic, so if the stopped truck is heading the opposite direction, they'll need to go past the truck, get off the freeway, turn around, and come back from the other direction, which will limit how far they can travel in 10 minutes.
Also, the rescue vehicles presumably can't start out being on the freeway. They need to wait in a parking lot somewhere off the freeway, so that will add a minute or two of delay.
So the gap between drivers would be somewhere between 10 and 20 miles. I suspect that in practice it would be closer to 10 but it depends on how much they care about getting there within the 10-minute window 100 percent of the time.
You are missing an important consideration and that is that 41k people die in auto accidents each and every year and 100k+ are injured, some crippled for life in the USA alone. Medical costs related to auto accidents are HUGE!
Once dependable autonomous driving comes to cars, it will not be long before human drivers are prohibited from driving. Possibly this will be kicked off by some lawsuit concluding that human drivers are an unacceptable risk to life and limb and should no longer be allowed to drive cars, likely initially on the highway.
As humans are banned from driving and fully autonomous cars take over on the highways, it will be only a short time until they are networked linked and therefore speed limits will be abolished on highways. Cars will travel at whatever speed is deemed safe by the software based on every car being in millisecond contact with every other car. People will get where they need to go faster and be productive while en route.
After the banning of humans from highway driving, it won't be long before humans are similarly banned from city driving. The potential savings in life, limb and accident avoidance will be too great to ignore.
Next step will be the removal of steering wheels and control pedals from cars savings some costs AND the standardization of driving on one side of the road worldwide.
If you aren't allowed ot drive any longer, WHY do you need a car sitting at your house that is unused 90% or more of the day? This will lead to people giving up their autonomous driven cars and turning to fleet services, which to be successful, will need to be able to dispatch an on-call car to someone within say, a 10 minute window.
Of course, with no cars owned by most people, we can then do away with garages for cars and parking spaces on streets.
Auto insurance and auto mechanics as careers? Buh-bye! All the maintenance will be done by the fleet owners. Insurance will no longer be necessary as accidents will be few and far between.
IMO, all of this could take place in the next 10-20 years.
They should just set up a system near the exhaust on the back that can shoot little safety triangles attached to a line out. Then a feature that reels it back in after. Or work with some of those little robotics companies to wheel one out and back in. The latter would probably be more logistically challenged, but both could be a solution.
Automated trucking is in Elon's plan. So, yes, other people paid significant amounts to play, but not the unions. Better to consider the Teamsters more as Trump's emotional support group. It's a nontransactional relationship, unfortunately for them.
From the article it sounds like maybe they are planning to have convoys with a chase car that could put out the triangles as needed. As long as your convoy is not longer than whatever distance is covered at 10 minutes of top speed, your chase car will always be in range.
The little quadcopter onboard whose entire function is to be dispatched to drop two little orange triangles behind the truck seems extremely dumb, but totally do-able.
As far as regulations making self-driving vehicles less cost-effective, this actually seems *less bad* than self driving taxis keeping one seat open for an imaginary driver (or realistically an emergency responder).
How hard would it be to fasten a warning triangle to an iRobot vacuum or similar small robot and have them deploy when a truck makes an emergency stop? Seems easier automation than the actual truck itself. If a truck turns on its side as the union rep claims is a situation that a flashing light would be inadequate, how likely is the driver to climb out a deploy triangles anyway.
And how did TuSimple handle this situation in the trucks they were driving autonomously between Tucson and Phoenix on Interstate 10?
Just a note that you only need 9 extra people over 200 miles to reach a point with in 10 minutes, since they can travel both directions, and the starting and ending hubs can also cover the 10-mile sections near them - and that's assuming 60mph travel (typically exceeded in TX)
I think this gets complicated pretty quickly. Someone at a particular exit will be able to cover 10 miles of the northbound lane to the north and 10 miles of the southbound lane to the south. And they'll be able to cover some of the "opposite direction" lane by going past the point you want to reach, exiting, turning around, and coming back in the other direction. How much opposite-direction coverage you can get will depend on how close together the exits are and how long it takes to do the turnaround. But I agree that in principle one person can probably cover more than 10 miles of freeway.
On the other hand, there might be stretches of freeway where the exits are spaced inconveniently, such that you have to have some overlap to get full coverage. Also, it presumably wouldn't be safe for people to be waiting in their car on the freeway. They'll need to be pre-positioned in a parking lot off the freeway, which might add a minute or two to the time they need to get to the freeway entrance. And you've gotta leave some amount of time for the person to get a notification that they're needed and respond. And of course some areas of the freeway are going to be congested, so a company might want to leave itself a bit of a buffer.
Ultimately I think it would make more sense for Aurora to send trucks out in convoys once or twice a day, in which case they could have a single car following along behind the group ready to place the warning triangles. But that seems to contradict Sterling's "no escort" comment so I'm not sure what they're planning.
I think a key issue is that trucks might be stopped or pulled over for traffic reasons, or they may create traffic backup by stopping. In those cases--when the triangles are most needed--no one could reach the site in ten minutes by regular car.
Ques: What about the technical challenge of driving on the highway? How good is Aurora's truck 🚛? Why did highway truck driving not take off before since highways are relatively easy to drive on (more interested in knowing the technical challenges rather than legislative or other challenges)?
Highway driving is easier on average, but the higher speeds mean that the worst-case scenarios are much worse. You can't just slam on the brakes if you get into trouble, and a crash is much more likely to be deadly. This is why Waymo is only starting to test driverless operation on freeways six years after they started doing driverless service on surface streets. Aurora also says that they have a superior lidar sensor that allows them to see further ahead, which is important whe you're traveling down the freeway at 70 mph.
Interesting topic. On the commercialization side, I think truck driving is a no brainer for AVs and the 'triangle issue' may explain why it's taken so long. You can have human drivers on both ends to help with loading/unloading, drop and hook, etc. And then AVs are perfect for long barren stretches of highways inbetween cities. My Tesla 'FSD' excels on freeways, so I don't see any reason why better/more advanced tech (ie lidar) couldn't do the same. Waymo's CEO also said in an interview at the All In Summit that their vehicles can do freeways no problem, but they learn more on city streets (I think there's also major negative PR implications if they get into a deadly accident on freeways whether at fault or not).
Traditionally, red states have been more receptive to business interests but the automation debate may be different. I'm all for advancing automation in a safe/thoughtful way (ie I think we should consider a soft landing for Uber/truck drivers displaced by AVs to avoid catastrophe like the NYC cabbie suicides) but it seems like today's news with Trump banning automation at ports, and Florida's Ron DeSantis banning lab grown meat (let the free market decide, as long as it's safe to eat. People say it tastes terrible though..), etc. there is a clear shift in the old guard's thinking. Will be interesting to see how that jives with the new conservative tech leaders like Musk and others who are definitely pro-automation and see the benefit that will outweigh any short term pains for small groups of workers.
Did any cabbie actually commit suicide? If so it’s probably a medallion owner that spent a fortune for a government monopoly token that dropped in value, not just the loss of a driving job. I imagine independent semi driver-owners will be impacted, but I doubt a self driving truck takeover would be rapid enough to cause great losses in the lifespan of a typical semi.
That’s from 2018 and says, as I expected they were owners of medallions that lost value, so you can blame Uber and Lyft, or you can blame the NYC government trying to over regulate small business with price controls. No one is blaming the drivers.
I'm not blaming Uber and Lyft but disruption has consequences is all. Wouldn't have been hard to set up a soft landing for some cabbies that owned medallions and maybe avoided some suicides. Could care less about the sleazy guys like Michael Cohen that owned a bunch.
Same principle applies with disrupting truck driving jobs (blue collar and full time).
Assuming the truck must have a data link to its base, it could be equipped with a UAV that could be launched and operated remotely to place triangles or flares as needed, more rapidly and safely than a driver could. A possible alternative might be an array of mortar tubes that could project folded warning devices in a predetermined pattern, with the devices unfolding on impact. Just two instant ideas; surely a good engineering study would quickly come up with better.
I assume they've thought about options like this and decided not to do them. I asked to talk to someone at Aurora about this and they declined to share more than their boilerplate statement, so I'm not sure what they've been thinking.
I love this kind of content, thanks!
> Robotaxi companies like Waymo get a lot of attention, but economically speaking, trucking is far more important. The trucking industry had revenues of almost $1 trillion in 2023, an order of magnitude larger than the $80 billion consumers spent on taxis and ride hailing apps that year.
This makes about as much sense as saying that the development of reliable fusion power will have next-to-no economic impact because the US currently spends <$3 billion a year on fusion.
The majority of personal driving happens in personal vehicles because it's cheaper and more convenient. If Waymo sufficiently reduces the cost of ride-hailing, people will switch over to it instead. The actual economic import of driverless taxis is relative to the entire amount spent on non-recreational driving, (plus alternatives like biking, walking, and public transit), which is much more than $1 trillion.
https://en.wikipedia.org/wiki/Induced_demand
I agree that in the long run self-driving technology will come to passenger vehicles and that will have a much bigger impact than on taxis. Whether you want to consider that to be an extension of the taxi market or its own separate market seems like a question where reasonable people can disagree. And don't forget that the induced demand point could apply to trucking as well.
That's fair, although that TAM is a LONG way off. I think Waymo needs to prove they can compete with Uber first at scale. And then they can go after personally owned vehicles, which would have much lower utilization and thus a longer payback period.
People *buying* a driverless car has a longer payback period, sure. But the majority of personal cars are unused for the majority of the time, which is a complete waste. If "Uber but only 10% the price and wait times" existed, many people would just sell their cars and use only ride-hailing.
I was writing articles predicting the end of personal car ownership a decade ago ( https://www.vox.com/2014/5/28/5758560/driverless-cars-will-mean-the-end-of-car-ownership ) but I feel less confident in this conclusion today. People like to have their own cars so that they can have car seats in the back, golf clubs in the trunk, etc. Probably those preferences will shift as cohorts age and driverless technology gets cheaper, but I think this is going to be quite a slow process.
And I don't think 10x cheaper is going to happen any time soon. Labor is half the cost of the taxi, and not all of that will go away—you still need people to clean, repair, and charge the vehicles. So they will absolutely be cheaper than today's Ubers but I'm guessing it'll be more like 2x cheaper a decade from now rather than 10x.
Maybe I'm stupid, but why not have drones attached to these trucks that can be piloted from HQ that set down the triangles?
That idea did occur to me. I would guess it would be complicated and expensive to redesign the truck to add a place for the drone to live inside the truck and automatically get released. Also they might not want to set a precedent that would lead to them having to keep those drones around for the next decade.
And thank you for responding. I listened to your episode of Derek Thompson's podcast a couple weeks ago and you kinda blew my mind.
As an engineer it seems implausible to me that there would be any great difficulty in accommodating and launching a UAV. However, I see it as pretty likely that the multitube mortar would prove superior. Similar systems are widely used to launch various protective payloads from military vehicles, and it would probably be possible to buy one off the shelf. The foldup device surely poses no great engineering challenges.
My main counter-thought too was precedent, but it really feels like it wouldn't be *that* difficult, and definitely worth exploring if the triangles are adjudicated 100% required.
My view is that this isn't about the technical details behind the triangles, it's a matter of principle that it's a somewhat silly rule designed for humans that having to work around sets a bad precedent for; I think they're just playing hardball here.
If the trucks are driving the highway and so the rescue drivers can travel 10 miles — *in either direction* — in 10 minutes, wouldn’t you just need someone every 20 miles? So that’s 10 people over 200 miles.
It's a fair question. A rescuer can't cross lanes of traffic, so if the stopped truck is heading the opposite direction, they'll need to go past the truck, get off the freeway, turn around, and come back from the other direction, which will limit how far they can travel in 10 minutes.
Also, the rescue vehicles presumably can't start out being on the freeway. They need to wait in a parking lot somewhere off the freeway, so that will add a minute or two of delay.
So the gap between drivers would be somewhere between 10 and 20 miles. I suspect that in practice it would be closer to 10 but it depends on how much they care about getting there within the 10-minute window 100 percent of the time.
Thank you for your reply. Love the blog!
You are missing an important consideration and that is that 41k people die in auto accidents each and every year and 100k+ are injured, some crippled for life in the USA alone. Medical costs related to auto accidents are HUGE!
Once dependable autonomous driving comes to cars, it will not be long before human drivers are prohibited from driving. Possibly this will be kicked off by some lawsuit concluding that human drivers are an unacceptable risk to life and limb and should no longer be allowed to drive cars, likely initially on the highway.
As humans are banned from driving and fully autonomous cars take over on the highways, it will be only a short time until they are networked linked and therefore speed limits will be abolished on highways. Cars will travel at whatever speed is deemed safe by the software based on every car being in millisecond contact with every other car. People will get where they need to go faster and be productive while en route.
After the banning of humans from highway driving, it won't be long before humans are similarly banned from city driving. The potential savings in life, limb and accident avoidance will be too great to ignore.
Next step will be the removal of steering wheels and control pedals from cars savings some costs AND the standardization of driving on one side of the road worldwide.
If you aren't allowed ot drive any longer, WHY do you need a car sitting at your house that is unused 90% or more of the day? This will lead to people giving up their autonomous driven cars and turning to fleet services, which to be successful, will need to be able to dispatch an on-call car to someone within say, a 10 minute window.
Of course, with no cars owned by most people, we can then do away with garages for cars and parking spaces on streets.
Auto insurance and auto mechanics as careers? Buh-bye! All the maintenance will be done by the fleet owners. Insurance will no longer be necessary as accidents will be few and far between.
IMO, all of this could take place in the next 10-20 years.
They should just set up a system near the exhaust on the back that can shoot little safety triangles attached to a line out. Then a feature that reels it back in after. Or work with some of those little robotics companies to wheel one out and back in. The latter would probably be more logistically challenged, but both could be a solution.
Automated trucking is in Elon's plan. So, yes, other people paid significant amounts to play, but not the unions. Better to consider the Teamsters more as Trump's emotional support group. It's a nontransactional relationship, unfortunately for them.
From the article it sounds like maybe they are planning to have convoys with a chase car that could put out the triangles as needed. As long as your convoy is not longer than whatever distance is covered at 10 minutes of top speed, your chase car will always be in range.
The little quadcopter onboard whose entire function is to be dispatched to drop two little orange triangles behind the truck seems extremely dumb, but totally do-able.
As far as regulations making self-driving vehicles less cost-effective, this actually seems *less bad* than self driving taxis keeping one seat open for an imaginary driver (or realistically an emergency responder).
This is all a bit silly. Trump curried favor with Palestinians in Michigan last Fall.
How hard would it be to fasten a warning triangle to an iRobot vacuum or similar small robot and have them deploy when a truck makes an emergency stop? Seems easier automation than the actual truck itself. If a truck turns on its side as the union rep claims is a situation that a flashing light would be inadequate, how likely is the driver to climb out a deploy triangles anyway.
And how did TuSimple handle this situation in the trucks they were driving autonomously between Tucson and Phoenix on Interstate 10?
Just a note that you only need 9 extra people over 200 miles to reach a point with in 10 minutes, since they can travel both directions, and the starting and ending hubs can also cover the 10-mile sections near them - and that's assuming 60mph travel (typically exceeded in TX)
I think this gets complicated pretty quickly. Someone at a particular exit will be able to cover 10 miles of the northbound lane to the north and 10 miles of the southbound lane to the south. And they'll be able to cover some of the "opposite direction" lane by going past the point you want to reach, exiting, turning around, and coming back in the other direction. How much opposite-direction coverage you can get will depend on how close together the exits are and how long it takes to do the turnaround. But I agree that in principle one person can probably cover more than 10 miles of freeway.
On the other hand, there might be stretches of freeway where the exits are spaced inconveniently, such that you have to have some overlap to get full coverage. Also, it presumably wouldn't be safe for people to be waiting in their car on the freeway. They'll need to be pre-positioned in a parking lot off the freeway, which might add a minute or two to the time they need to get to the freeway entrance. And you've gotta leave some amount of time for the person to get a notification that they're needed and respond. And of course some areas of the freeway are going to be congested, so a company might want to leave itself a bit of a buffer.
Ultimately I think it would make more sense for Aurora to send trucks out in convoys once or twice a day, in which case they could have a single car following along behind the group ready to place the warning triangles. But that seems to contradict Sterling's "no escort" comment so I'm not sure what they're planning.
I think a key issue is that trucks might be stopped or pulled over for traffic reasons, or they may create traffic backup by stopping. In those cases--when the triangles are most needed--no one could reach the site in ten minutes by regular car.
Ques: What about the technical challenge of driving on the highway? How good is Aurora's truck 🚛? Why did highway truck driving not take off before since highways are relatively easy to drive on (more interested in knowing the technical challenges rather than legislative or other challenges)?
Highway driving is easier on average, but the higher speeds mean that the worst-case scenarios are much worse. You can't just slam on the brakes if you get into trouble, and a crash is much more likely to be deadly. This is why Waymo is only starting to test driverless operation on freeways six years after they started doing driverless service on surface streets. Aurora also says that they have a superior lidar sensor that allows them to see further ahead, which is important whe you're traveling down the freeway at 70 mph.
Interesting topic. On the commercialization side, I think truck driving is a no brainer for AVs and the 'triangle issue' may explain why it's taken so long. You can have human drivers on both ends to help with loading/unloading, drop and hook, etc. And then AVs are perfect for long barren stretches of highways inbetween cities. My Tesla 'FSD' excels on freeways, so I don't see any reason why better/more advanced tech (ie lidar) couldn't do the same. Waymo's CEO also said in an interview at the All In Summit that their vehicles can do freeways no problem, but they learn more on city streets (I think there's also major negative PR implications if they get into a deadly accident on freeways whether at fault or not).
Traditionally, red states have been more receptive to business interests but the automation debate may be different. I'm all for advancing automation in a safe/thoughtful way (ie I think we should consider a soft landing for Uber/truck drivers displaced by AVs to avoid catastrophe like the NYC cabbie suicides) but it seems like today's news with Trump banning automation at ports, and Florida's Ron DeSantis banning lab grown meat (let the free market decide, as long as it's safe to eat. People say it tastes terrible though..), etc. there is a clear shift in the old guard's thinking. Will be interesting to see how that jives with the new conservative tech leaders like Musk and others who are definitely pro-automation and see the benefit that will outweigh any short term pains for small groups of workers.
Did any cabbie actually commit suicide? If so it’s probably a medallion owner that spent a fortune for a government monopoly token that dropped in value, not just the loss of a driving job. I imagine independent semi driver-owners will be impacted, but I doubt a self driving truck takeover would be rapid enough to cause great losses in the lifespan of a typical semi.
Yes I spent 2 seconds and found this story on Google Search - https://www.nytimes.com/2018/12/02/nyregion/taxi-drivers-suicide-nyc.html
That’s from 2018 and says, as I expected they were owners of medallions that lost value, so you can blame Uber and Lyft, or you can blame the NYC government trying to over regulate small business with price controls. No one is blaming the drivers.
I'm not blaming Uber and Lyft but disruption has consequences is all. Wouldn't have been hard to set up a soft landing for some cabbies that owned medallions and maybe avoided some suicides. Could care less about the sleazy guys like Michael Cohen that owned a bunch.
Same principle applies with disrupting truck driving jobs (blue collar and full time).