Driverless Semis have Arrived

https://insideevs.com/news/787100/aurora-driverless-trucks-1000-miles-update/
Pittsburgh-based Aurora is on track to triple its network of driverless delivery trucks.

Aurora, the Pittsburgh-based autonomous trucking company, wants to supercharge its fleet of driverless big rigs. The company said the latest software update to its driverless hardware suite will enable its trucks to drive nonstop from Phoenix to Fort Worth.

That trip is roughly 1,000 miles long and takes 15 hours to complete, but here’s where trucks without drivers behind the steering wheel have the upper hand, because they’re not subject to the same mandatory rest breaks as human drivers. By contrast, a human driver would have to take a rest after 11 hours of driving.

“Being a carrier is a game of margins, and if autonomy can work around the clock, it will be key to growing our customers."

This is Aurora’s fourth software release since deploying driverless trucks on America’s roads back in April 2025.

The company currently has five autonomous trucks carrying cargo without safety monitors between Dallas, Houston, Fort Worth and El Paso.

5 Likes

There have been discussions about this, though I don’t recall which board. I, and some others, always thought this would be a first step. Freeways are free of pedestrians, stop signs, cross-traffic, and the like. Road construction would be a challenge. And I recall one creative thinker coming up with a method to rob such trucks (basically, box it in and come to a stop forcing the truck to stop, then loot the trailer).

I would think it challenging to get approval from all states the vehicles would pass through. There may also be city ordinances along the way which would require changing. They did choose a fairly easy route…not a lot of rain or snow, which makes it easier for the sensors to “see”.

Waymo is able to operate in Phoenix. No major incidents yet.

Though I have become skeptical of this. I saw a documentary about driverless vehicles a few years ago, and it is much more daunting than most people realize. We humans make it look easy, but it’s actually a very complicated task. Tesla was the first to implement “autopilot” on a large scale, and there have been several deaths attributed to people relying on the car driving itself. I don’t own a Tesla, but I believe they had to modify their software so people couldn’t do that (i.e. let the car drive itself without intervention). The first time a driverless semi barrels into a minivan with mom, dad, and a few kids, the project may come to a screeching halt.**

**Plus, there are dust storms -at least in Arizona-, which can result in pile-ups. A driverless semi barreling into THAT would cause mass casualties. Unless the driverless truck knows to pull off on the shoulder, stop/park, turn off the lights and take the foot off the brake (so to speak). That’s dust storm protocol. (Yeah, I live in Arizona.)

I mean, that’s probably not correct, though it depends on what you mean by “Autopilot.” They were the first to call their Level 2 ADAS system something dramatic like “Autopilot,” but they weren’t the first to market with a Level 2 system. Mercedes had their Level 2 system available in 2013, and BMW had theirs shortly thereafter. Tesla activated the Autopilot functionality in late 2015. That was the same year as Honda and Ford brought their Level 2 systems to market.

Returning to the thread topic, many folks did not anticipate the very long amount of time needed to get to AV long-haul trucking that exceeds permissible human drive times. I, for one, didn’t recognize the challenge that the article describes - that although the driving environment might be less complicated, the physics behind a fully-loaded truck moving 65 mph present braking and obstacle avoidance challenges that are more complicated than a slower, lighter passenger vehicle. Also, that LIDAR lets the vehicle see farther than vision - an important factor when stopping distances are so much longer.

4 Likes

Can we derive from that statement that Tesla semi has been leap frogged?

My quick bit of googlefu suggests that they have zero accidents since they began operating such vehicles.

Tesla semi is a prototype hauling mostly air (cheetos or such).

So, did Aurora leap-frog nothing? Yes.

I suspect the Teamsters (and thus unionized companies) will not be supportive.

DB2

1 Like

That part didn’t disturb me much. Perhaps I assumed too much, but I figured they would have programmed the weight/mass into the computer before it set off. But I didn’t think of maneuvering. Semis can jack-knife if the maneuvering is too sudden/extreme. Seems obvious now, it just didn’t occur to me at first. They also can be prone to tipping if they turn too sharply. (1poordad was an office manager/dispatcher for a diesel wrecking company.)

Tesla claimed to have nearly-full autopilot. Their system didn’t prevent people using like that. I remember seeing video of drivers asleep at the wheel of their Tesla while it was cruising down the freeway. I don’t know about Mercedes and BMW, but I don’t recall hearing/seeing any evidence you could be so reckless with theirs. So, I believe Tesla really did the first large-scale fielding of unrestrained autopilot, with some predictable results.

And, as the other poster mentioned, perhaps Tesla has now been leap-frogged. Though I still doubt it is ready for prime time. And I still question the hoops they would have to jump through for different states and municipalities to legally operate these things.

It sounds like it’s also basic stopping distance. That’s one of the things the company mentioned in the recent article. A loaded semi takes about 200 feet longer to come to a stop than a passenger car, even without worrying about jack-knifing, because it’s got more mass and because their braking systems aren’t as quick:

Stopping Distances - Truck Smart

I can see how that might increase the degree of difficulty in getting your AI driver to a sufficient level of safety. The environment is simpler, but the vehicle is much less responsive and agile. Not something that had occurred to me, either - but it makes a certain amount of sense.

1 Like

And I can see the opposite. An AI driver is never distracted, isn’t trying to eat a ham sandwich while driving, isn’t smoking a cigarette with ash falling in his lap or a joint with predictable results. It isn’t speeding and it’s running slow and gentle 100% of the time.

Which isn’t to say it won’t encounter a sudden traffic backup, a pothole to be avoided, or some jerk wanting to brake-check it just because, but those are all things that are going to happen anyway.

I would note that hauling a load of potato chip bags is a very different thing than hauling a load of bottled water, and running between 10pm and 6am on interstates is likewise very different than doing it during rush hours.

If we are all to be crash test dummies in this experiment (as we have been with Tesla’s so called Full Self Driving and/or Autopilot), I hope they do it in other states than where I live, on roads I don’t drive, with trucks I never see. But I suspect other companies will be more responsible about it than Tesla was. At least I hope so.

2 Likes

I don’t disagree, but that wasn’t really my point. I had fully expected that long-haul trucking AV drivers would come much faster than robotaxi services. I thought that solving AV truck driving on freeways would be easier than solving AV city driving for passenger pickup and drop off - and that the economics of AV long-haul trucking would provide a better return than robotaxis, because of the drive-time limitations noted in the original article.

I found it surprising that it’s taken so much longer. The article noted that although interstate driving seems like it would be a much simpler problem to solve, there are some things about interstate driving (higher speeds and heavier/less responsive vehicles) that increase the degree of difficulty. Those things hadn’t occurred to me, but in retrospect I can see why they would present challenges to AV trucks.

2 Likes

I remember. And even agreed with you at the time. Then I watched NOVA’s “Who’s Driving Now”. Totally shattered my illusions. Driving is enormously complicated, and I agree with Gh that I hope they don’t allow auto-drive vehicles until they work out everything (which means it may never happen).

I occasionally see a Waymo in the east valley here (mostly Chandler). We have 300+ days of sunshine, never any snow or ice, the streets are mostly a cartesian grid…that’s why they are testing that here. They are conservative “drivers”, I’ve never seen them at night (they probably recall them before sundown), and we’ve never had a major incident that I’m aware of. I wouldn’t trust them in bad weather, and I still can’t get out of my mind the AI in the NOVA program that didn’t identify a pizza delivery guy as a pedestrian (because he had the pizza on his shoulder).

I like the driver assist features of my ID.4. They’re great. They also don’t allow me to ignore the driving, and start yelling at me if I take my hands off the wheel for more than a few seconds. If I ignore that, they’ll disengage entirely and the car will coast to a stop.

1pg, I seem to remember you writing after the Volkswagen emissions scandal that you’d never own a VW. What changed?

DB2

Not sure I said exactly that. I did lament that I was giving up my TDI. That car was wonderful. But what changed was that they had the best vehicle (EV) within my budget. Japan didn’t have a significant presence in EVs yet**, and I’m not buying a US car. Though I did drive the Bolt. It was ‘ok’, but the ID.4 just felt better driving.

**Except for the Leaf, but that was only good for about 80 miles without air conditioning…with air, even less miles. They’ve made them better since, of course.

1 Like

While this is indeed true, they do have mandatory stops for charging.

How is this any different than with a driver? You box the truck in, the driver is forced to pull over, you make sure the driver sees a weapon, and you loot the truck.

I wonder if driverless semis will eventually towing trailers with solar panels on top of the trailers. That could extend the range of the driverless semis a bit.

Looking on the internet EV Semis have a 500 mile range. How long does it take to charge an EV semi?
A diesel semi has a 2000 mile range. But human driver are required to stop every 8 hours for a 30 minute break. And I believe after 14 hours of driving [so ~1000 miles of range a day] the human driver is require to take a 10 hour break.
Will driverless EV semi be fitted with EREV range extenders? And how much range would those EREV provide?

The amount of juice from a trailer covered with solar panels wouldn’t be worth it, you might get an extra 1 or 2% over the course of a trip.

From my reading, it takes about 30 minutes to charge about 70%, but hours and hours to charge for the “full range” quoted. That’s because the charging rate has to slow down as the battery gets more and more full, otherwise the battery gets damaged.

[Best analogy: the battery has to “park” electrons”. Like a shopping center parking lot that’s empty, you can just zip into a slot. When the lot is 80% full you drive around slowly looking for an opening. At Christmas time you cautiously inch along hoping to find a place that’s not already occupied. Same thing with batteries & electrons.]

So you could charge “full” overnight and get 500 miles, and then “quick stop” for 30 minutes to get another 350 miles (optimistically) and repeat every few hours - but that’s gonna take some extra stations (and pricing) to get there. Diesel infrastructure, by contrast, is well established and pricing is pretty reliable.

None of this means electric semis won’t work; they will - but at least at the beginning it will be special use cases until range, stations, and pricing become more reliable. I can see a case for certain companies with established distribution patterns making use of it; they can charge overnight and run within the mileage limits (or recharge at their own stores [looking at you, Walmart]) but it will be a while before they become a “general use” category.

1 Like

I’ve been wondering how they will handle weigh stations.

The driver would be aware of what’s happening, and bulldoze through it.

1poordad used to manage a diesel wrecking (towing) service. Those guys were not shy about using their size/mass to intimidate. An AI will try to avoid an accident. A driver that realizes someone is trying to hijack him will get aggressive.

Besides, the better place for those sorts of shenanigans are a rest stop while the driver goes inside to relieve himself. Bolt cutters, loot the trailer, and -hopefully- be gone before the driver gets back.

Interestingly enough, there are already plenty of special use cases with diesel trucks. For example on the I-80 corridor in Wyoming, UPS (and presumably others) will pull trailers across the state, then switch tractor and driver, and almost immediately continue on to the destination (or another intermediate destination). Because it’s a “standard” route, the benefits are twofold, one, the trailers get to the sorting location quicker, and two, the drivers do the same route back and forth each day and can sleep at home. Because the total round trip route is less than X hours, the drivers don’t “timeout” and have to sleep somewhere for Y hours. Similar things could be done with electric trucks to avoid requiring huge (expensive) batteries and long charging times.

4 Likes

No way! That driver would lose their job and never be permitted to drive a truck again. Just like store clerks are instructed to not attempt to prevent robberies in progress, truckers are similarly instructed.