Self-driving trucks

[skluth]

I think Self-driving vehicles of all types could be discussed again as it's now happening commercially. It probably signals the end of the truck-driving profession at least as we know it. We also had a Senator candidate in our last primary who made a big deal about Tesla's self-driving cars. (He lost.) Article linked in headline.

The future is now as driverless trucking hits nation's highways

Markie Martin, Tom Palmer

Posted: APR 21, 2022 / 07:23 AM CDT | Updated: APR 22, 2022 / 11:20 AM CDT

(NewsNation) – Many Americans might not realize that driverless tractor-trailers are currently navigating the nation's highways, hitting the open road with absolutely nobody behind the wheel.

Many of us have ridden in a smaller car – like a Tesla – that has a driverless feature, but to be in a large freight truck that is maneuvering through cities and highways is a completely different ballgame.

It's the future of the industry, but the future is already here.

Autonomous driving technology company TuSimple was founded in San Diego in 2015 with a mission to improve the safety and efficiency of the trucking industry.

TuSimple is a developer of heavy-duty, self-driving trucks and the autonomous startup has already created a freight network along the Sun Belt from Phoenix to Houston.

"This is better, no doubt about it,"  said Jim Mullen, chief administrator and legal officer for TuSimple. "That's not a slight on the human driver whatsoever, it's just factual."

TuSimple's driverless truck is built to blend in, but equipped with high-definition cameras, radars and artificial intelligence. It can "see"  a thousand meters ahead and 360 degrees around the vehicle.

According to Mullen, it's far more aware and advanced than human detection.

"It doesn't get mad, there's no road rage, it's in no hurry,"  Mellen said. "It doesn't have the characteristics that we as humans sometimes have."

But seeing it and experiencing it are two different things.

NewsNation correspondent Markie Martin recently traveled to Tucson, Arizona inside one of the autonomous trucks.

Martin and her crew hopped inside a driverless semi and rode along for an hour as it drove itself across an Arizona highway.

"Autonomous driving started,"  the system chimed as it took the helm during their test ride.

While in autonomous mode, the truck is driving itself.

Koabi Brooks, TuSimple's general manager, has been working on the company's "driver-out"  system for four years and no longer gets nervous when taking his hands off the wheel.

The "driver-out"  system drove for an hour up and down the Arizona highway, switching lanes, taking exit ramps and maneuvering around traffic – all while nobody touched the wheel.

"The technology's here ... we have to show the world that it works,"  Brooks said. "And really, for over-the-road trucking, we do think that "˜driver-out' is the next step for this. When people talk about the driver shortage, they're really talking about over the road. There are plenty of drivers out there, but living on the road ... is less and less popular."

NewsNation also spoke to David Gaber, product manager with Waymo Via, an autonomous trucking company, about the safety precautions of driverless trucking.

"We take safety very seriously,"  Gaber said during a Thursday appearance on "NewsNation Prime."  "We have a rigorous safety framework that we published. It's available at waymo.com. That framework contains both methodologies and principles and plans for testing our technology both on closed course as well as simulation.

"We think this technology is most appropriate for long-haul, over the road routes, today. These are the same routes that truck companies have a very hard time recruiting for. We want to deploy this technology on routes that are hundreds if not thousands of miles, because these are the routes where truck drivers have to spend days, or weeks, away from friends and family.

"Our trucks do not get lonely, they do not get bored, they do not miss home. And so we see this technology as a natural compliment to serve these ultra long-haul routes. And that would allow existing trucking fleets to focus on regional routes, with the added benefit of truckers getting ... to go home every month,"  Gaber added.

The driver shortage is a real issue facing the industry. It's projected by the American Trucking Associations (ATA) that the shortage will surpass 160,000 positions by 2030.

TuSimple says the company is trying to help solve for that, not push truckers out of a job.

"If you're a truck driver today or even thinking about being a truck driver, you will continue to have a full career as a truck driver,"  Mullen said.

And the benefits of driverless freight are numerous – as it not only reduces fuel consumption and carbon emissions but also works to alleviate supply chain delays.

Mullen uses the specific example of getting fresh produce from Arizona to Oklahoma, saying, "We were able to get those watermelons from Nogales to Oklahoma City in 14 hours. If you did that with a human driver, that's about a 24-hour run."

Providing the consumer with fresher produce and everyday deliveries at a faster rate is the company's focus.

So far, TuSimple has completed hundreds of miles of autonomous deliveries in Arizona and plans to start commercializing in 2024.

It's a date when everyday drivers can expect to frequently share the nation's highways with semitrailers without a driver at the wheel.

[wanderer2575]

Will the trucks be programmed to block passing lanes while micropassing on uphills, and tailgate cars on downhills?

[Chris19001]

There is only a miniscule percent chance the standard commercial insurance market will insure driverless trucks at any point in the immediate future.  Why would an insurance company underwrite a risk where the truck 's software is the probable cause of loss?    We will get to a point someday, but its not right around the corner so to speak.

[Max Rockatansky]

There is only a miniscule percent chance the standard commercial insurance market will insure driverless trucks at any point in the immediate future.  Why would an insurance company underwrite a risk where the truck 's software is the probable cause of loss?    We will get to a point someday, but its not right around the corner so to speak.

Someday might be pretty far off.  Even cars don't seem to be close to being able to operate autonomously without the aid of a supervisory driver.  There is so much variation on roadways that it isn't likely full automation is coming any time soon. 

Speaking of automation, has there been any serious attempts to advance the technology in more controlled forms transportation?  Specifically I'm thinking rail, sea borne shipping and even air freight might be better venues to explore more automation over anything on a road/highway.

[kalvado]

Insurance companies are certainly used to dealing with the risk of drowsy, drunk or sick driver. There are certainly ways to test automation and get some statistics to estimate risks.



Now, some criticism:
company web site has no word about commercially available solution. they planned to get 40 trucks before covid.  In February 2022, company reported a grand total of 550 miles of fully autonomous driving in Arizona. Not sure if that was in Phoeinx or I-10 outside of city with an escort car...  Show me how they navigate I90/I94/I65 cluster in Chicago...
Safety report (1 year old)  doesn't have hard numbers, at least at a first glance:
https://www.tusimple.com/wp-content/uploads/2021/03/TuSimple-Safety-Report.pdf 
Some numbers on the web site look just made up to me.


My impression is that self-driving - like fusion reactors - is always just a few years away at most.  There are successful tests, and then radio silence.

UPD: 2022 document: https://ir.tusimple.com/static-files/8b58d513-5193-4ea4-b2d9-1c353248ebbe
So far, driverless trucks passed 28 traffic lights and did 101 lane change. Well.... that's a good start....

Upd2: they also claim 160k miles of autonomous driving - without any details on what was automated, and if someone was behind the wheel, no statistics shown. Color me skeptical, until they just used cruise control to maintain speed.   

[jeffandnicole]

I think Self-driving vehicles of all types could be discussed again as it's now happening commercially. It probably signals the end of the truck-driving profession at least as we know it.

You're about 50 years ahead of the game.  We'll still have another generation or two of truck drivers to go.

[SectorZ]

Can't wait to watch the truck pull itself over and put tire chains on in the snow.

[MoiraPrime]

You should be wary of the marketing around any self driving tech right now. They've been known to use sketchy marketing to make their tech look more developed than it actually is.

Relevant video about a different self driving vehicle for reference: https://www.youtube.com/watch?v=CM0aohBfUTc

[triplemultiplex]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

[kalvado]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

[mgk920]

Heavy freight and passenger carrying vehicles that can steer themselves?  We used to call those 'trains'.



Mike

[kalvado]

Heavy freight and passenger carrying vehicles that can steer themselves?  We used to call those 'trains'.



Mike

one word: engineer

[vdeane]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

I also remember reading about the possibility of "truck trains", where multiple self-driving trucks follow a human-driven truck.  I could see something like tandem lots where people drop off trucks, line them up in a train and put them in self-driving mode, and then they follow the leader until they get to the destination area, where they pull off, take them out of self-driving mode, and wait for human drivers to pick them up and take them down the last mile.

[Rothman]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

I also remember reading about the possibility of "truck trains", where multiple self-driving trucks follow a human-driven truck.  I could see something like tandem lots where people drop off trucks, line them up in a train and put them in self-driving mode, and then they follow the leader until they get to the destination area, where they pull off, take them out of self-driving mode, and wait for human drivers to pick them up and take them down the last mile.

And then the human slams on the brakes...

[kalvado]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

I also remember reading about the possibility of "truck trains", where multiple self-driving trucks follow a human-driven truck.  I could see something like tandem lots where people drop off trucks, line them up in a train and put them in self-driving mode, and then they follow the leader until they get to the destination area, where they pull off, take them out of self-driving mode, and wait for human drivers to pick them up and take them down the last mile.

And then the human slams on the brakes...

Which is the easiest part of it, with interconnected vehicles and instant electronics response.
There is also a technique of very close following -  I believe "battalion driving" is the name - where speed differential due to sharp braking has no time to build up and is small enough not to damage anything.

Those trains may be an issue for the traffic, but that is a totally different story.

[Rothman]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

I also remember reading about the possibility of "truck trains", where multiple self-driving trucks follow a human-driven truck.  I could see something like tandem lots where people drop off trucks, line them up in a train and put them in self-driving mode, and then they follow the leader until they get to the destination area, where they pull off, take them out of self-driving mode, and wait for human drivers to pick them up and take them down the last mile.

And then the human slams on the brakes...

Which is the easiest part of it, with interconnected vehicles and instant electronics response.
There is also a technique of very close following -  I believe "battalion driving" is the name - where speed differential due to sharp braking has no time to build up and is small enough not to damage anything.

Those trains may be an issue for the traffic, but that is a totally different story.

Instant response on ice and snow...

[Dirt Roads]

^^^^
There is very little in the way of real-life experience using "platooning" software described above.  One of the main issues that we have experienced in the AGT (automated guideway transit) world is that the software delay between vehicles operating in a "platoon" gets greater for each path.  For instance, if the control software has a 1/2-second filter for "data chatter", the best case response for the first vehicle is 500 msec plus brake response; 1.1 sec plus brake response for the second; 1.815 sec plus brake response for the third; 2.662 sec plus brake response for the fourth; and so on.  The worst case may be significantly worse as you get down the line.  You've got to build this into the physical distance separating the vehicles.

An interesting twist raised by kalvado is that the bumpers can be designed to fully interlock, which would allow the vehicles up front to assist with the brake effort while a control package downline in the "platoon" is trying to catch up.  I'm not sure that the current Motor Vehicle Carrier Safety Regulations would allow such a change in trailer bumper technology, but those kind of legislative issues seem to be easier to solve when "job loss" is not the driving issue.

[kalvado]

Just like passenger automobiles, the autonomous features are likely going to remain in the realm of adaptive cruise, lane keeping, and emergency braking.  The type of things that will be mostly relevant and easiest to utilize on freeway travel during non-adverse weather conditions.  The kind of things that will make up for lapses in concentration on the part of human drivers to avoid crashes and save a few lives.

If anything, major warehouse to warehouse route is the most plausible use scenario IMHO
For one, NYS Thruway has tandem trailer parking lots at (almost?) every exit, and local businesses use shuttle trucks to move second trailers around.
Having autonomous trucks go from lot to lot with human taking care of last mile seem pretty plausible in my world - highway is one of easiest driving modes..

I also remember reading about the possibility of "truck trains", where multiple self-driving trucks follow a human-driven truck.  I could see something like tandem lots where people drop off trucks, line them up in a train and put them in self-driving mode, and then they follow the leader until they get to the destination area, where they pull off, take them out of self-driving mode, and wait for human drivers to pick them up and take them down the last mile.

And then the human slams on the brakes...

Which is the easiest part of it, with interconnected vehicles and instant electronics response.
There is also a technique of very close following -  I believe "battalion driving" is the name - where speed differential due to sharp braking has no time to build up and is small enough not to damage anything.

Those trains may be an issue for the traffic, but that is a totally different story.

Instant response on ice and snow...

Let's put it so: human drivers somehow manage to navigate that, and there is probably a finite number of techniques to manage those problems. Those may, most likely, be translated into computer strategy.  Driving a truck doesn't require college degree, it is a skill which may - and is - be taught.  Slamming brakes on ice on 18-wheeler is an interesting way to get disability and unemployment benefits at the same time - and I bet that is part of what they teach.  I don't know what would they do on unplowed snow (not that human trucker would be comfortable in that case), but estimating braking distance and keeping enough of a gap is definitely within the possibility.   

Whatever you think about these self-driving computers, they definitely have a good idea of situation. 1000+ feet scanning range, 30 s situation analysis horizon - that is probably better than an average human driver has.
Areas where humans are still superior are less obvious. Distinguishing pedestrian from the bike rider, for example, is one of those problems. You need that distinction to predict future motion, though. Once those are removed...  Looks like things become much easier:
https://www.australianmining.com.au/news/rio-tintos-autonomous-trucks-pass-one-billion-tonne-milestone/

[kalvado]

^^^^
There is very little in the way of real-life experience using "platooning" software described above.  One of the main issues that we have experienced in the AGT (automated guideway transit) world is that the software delay between vehicles operating in a "platoon" gets greater for each path.  For instance, if the control software has a 1/2-second filter for "data chatter", the best case response for the first vehicle is 500 msec plus brake response; 1.1 sec plus brake response for the second; 1.815 sec plus brake response for the third; 2.662 sec plus brake response for the fourth; and so on.  The worst case may be significantly worse as you get down the line.  You've got to build this into the physical distance separating the vehicles.

An interesting twist raised by kalvado is that the bumpers can be designed to fully interlock, which would allow the vehicles up front to assist with the brake effort while a control package downline in the "platoon" is trying to catch up.  I'm not sure that the current Motor Vehicle Carrier Safety Regulations would allow such a change in trailer bumper technology, but those kind of legislative issues seem to be easier to solve when "job loss" is not the driving issue.

Again, if we're talking platoon (thanks for the word!) or convoy driving, instant messaging about braking are a no-brainer. I don't think such convoy is a good idea, though; 2 trucks following each other can create a merging problem; set of 10 trucks unwilling to split... 

[kalvado]

This article looks to me as a good technical review. It is is 2 years old, but seem current enough to me.  I specifically recommend looking at chapter 7 for the state of the art summary.
https://eprints.qut.edu.au/198107/1/56922561.pdf

The key "problem" for autonomous vehicle developers here is that cars are already quite safe: about one fatality for every 100 million miles of driving

For comparison, California has 340 billion miles per year and and 3600 traffic fatalities in 2019; so autonomous cars are about as safe as average US driver (which is not the highest possible bar, frankly speaking)