Whatever Happened to the Self Driving Semi?

A once-lively area of automation has largely closed down over the last few years. What happened and why?

A two-truck convoy operated by self-driving truck company Locomation, which closed its doors in 2023. Source: Technical.ly

There are almost three million semi trucks in the United States alone, to the point that trucker is the most common job in 29 states. Most of these are driving 400-600 miles per day along long, straight, predictable highways — a use case that, at a glance, seem perfect for autonomy.

And yet, on-road autonomy looks guaranteed to start not with semis but with taxis, operating over much shorter distances in much less of the United States. Major players like Waymo have shut down their self-driving truck businesses even as they expand self-driving taxi services all across the Sun Belt. And the startup crowd seems to have fared even worse, with once-promising companies like Embark, TuSimple, and Locomation all going under.

However, the news isn’t all bleak. Self-driving truck company Aurora raised around $820 million in new capital, with much of this being from Uber, which has been expanding into logistics with its $20 billion Uber Freight business. So there are active players with substantial funding, even as the field itself is narrowing and self-driving trucks haven’t yet seen their Waymo moment.

So, why did this happen? Will self driving trucks one day fundamentally rewrite our economy for the better, making our roads safer and more efficient, or is something else going on here?

Mounting Challenges

A row of TuSimple trucks. TuSimple wound down its North American operations in 2023. Photo: Cassidy Araiza for The Wall Street Journal

Self driving is hard; ask anyone. Waymo’s taxis and Tesla Full Self-Driving (FSD) are impressive now, but this was not always the case; and it’s been a really long road getting to this point. And trucking is probably worse; incidents abound, like a 2022 crash involving a TuSimple semi.

One thing that you see over and over again in robotics is that there are no shortcuts. A lot of the people trying to do self-driving trucks seemed to think they could make the problem much easier than it really was. Locomation wanted to do convoying, having an autonomous truck follow a human-driven truck; but at the margin, this ends up being just as complicated as full autonomy, since the two vehicles can be separated in heavy traffic. Starsky had perhaps an even riskier plan in their remote teleoperation of semi trucks; teleoperation is hard enough for robots that aren’t moving 80,000 pounds of goods at 65mph down an interstate.

So, to summarize:

• This is a really hard problem, as is all of self driving

• Trucking is uniquely highly regulated compared to other areas of self driving due to the massive risk involved — the vehicles are uniquely deadly.

• Many of the players in this space thought they could simplify the problem; it turned out that they could not.

• The taxi business is very good, and companies like Waymo have decided to focus entirely on it.

For companies like Waymo, it turned out taxis were closer, easier and more lucrative, without nearly as much regulation or hassle. Besides, driving a semi truck is a uniquely hard technical problem, one that’s slightly less suited for the methods we have access to right now.

Let’s Talk Stopping Distance

The core of that problem is vehicle dynamics.

Fully-loaded trucks are massive, with a legally-mandated maximum of 80,000 lbs. This makes everything a truck does notably less responsive. Planning becomes more difficult; learning methods are less effective, too, when there’s not a clear, immediate mapping between input and output.

If we want to discuss how serious a problem this is, we should look at stopping distance; i.e. how long it takes a semi truck to come to a complete stop because, say, there was an accident on the road ahead of it.

Stopping distance for a fully-loaded semi truck traveling at 65 mph is approximately 525 feet to about 600 feet. Even though most US highways have higher speed limits, trucking companies usually limit speed to 65 mph for safety and fuel efficiency reasons; it seems reasonable to expect that autonomous truckers would do the same. But note that this is under ideal conditions; stopping distances can as much as double on icy roads.

Now, a good long-ranged lidar could have 1000 feet of range. Aurora has a particularly good in-house lidar, with about 450 meters (~1500 feet) of range - much farther than many other options. But maximum range isn’t effective range, which is far more important. This is hard to estimate — it varies depending on conditions, on objects, and of course on the quality of the particular classifiers being used to interpret objects. This quantity is notably shorter than the maximum range on practically any sensor, by as much as about half; and we’ll also need to classify if this was a spurious detection (a plastic bag blowing onto the road, a cardboard box) or a serious issue.

And that’s setting aside other concerns: what if there’s a patch of black ice ahead on the road? The lidar can’t detect this at all, and it’s a huge issue for highway driving. There was a famously horrific 133-car pileup in Fort Worth, Texas in 2021, caused by black ice, which led to 65 injuries and six fatalities. If you watch video, you’ll see skilled semi truck drivers carefully bringing their vehicles to a halt through the event, minimizing damage to other drivers as much as possible.

All this is to say, we’re talking about a really important and very high-stakes perception problem. You cannot make any mistakes in this, or trucks will crash, and people will die. With those new 450 meter lidars, Aurora should be fine; but with shorter ranges of 100-300 meters, it’s very easy to run into trouble.

No Shortcuts

Developing any kind of real-world robotics is still something of a slog — it takes a lot of time and money to harden and productionize hardware, to implement data and training pipelines, and build software to handle edge cases.

Many self-driving truck companies, like Embark, just ran out of money during this long process. Many expected there would be an “off-ramp” where they could launch a limited version of the product to make money earlier — convoying for Locomation, for example. Other companies like Plus seem to be pivoting to driver assistance, which is another “off-ramp” which might still potentially pay out.

But many of the potential shortcuts don’t seem to work. Convoying might be easier 95% of the time, but in the remaining 5% it still degenerates to requiring full autonomy, at least for some period of time — what if the robot truck is separated from the lead in inclement weather or heavy traffic? Remote teleoperation is similar; network conditions on a long-haul trucking route are anything but predictable, meaning that you always need to fall back to reliable autonomy (or a human driver).

Finally there are shortcuts that self driving taxis can take that trucks can’t. Trucking routes by necessity cross state lines, meaning that you (usually) have to deal with multiple states’ baroque trucking legislation, and the carve-outs they’ve made for self-driving may end up being different. You can’t just roll out in sunny southern cities, as Waymo and Tesla are for their robotaxi programs.

Looking Forward

Zeem Solutions electric truck. Source: Electrek

And yet, with all these challenges aside, there are still plenty of companies that have kept on trucking. Multiple major players are still in the race, like Aurora. Former leadership from Argo AI recently started a new trucking company, Stack AV, backed by Softbank. And Volvo has recently started testing self-driving semis on Texas roads.

So despite the recent quiet, there’s still some hope for the area. There’s also new technology appearing in trucking in other ways: Zeem, for example, is piloting its new electric truck near Seattle, specifically for drayage — carrying shipping containers locally to distribution centers. While these aren’t autonomous, it shows how at the very least technological innovation can impact the space.

And there are other active players addressing different parts of the problem. Colorado-based Outrider has been working on automating logistics hubs, even if the transit between those hubs can’t be automated just yet. They recently raised $62 million back in 2024.

While the near future looks uncertain, none of the problems here seem fully intractable either; the core issues are all ones that can be solved, they may just take more time and money than expected. It’s fully possible that Waymo or Tesla will get back into the self-driving semi game at some point in the future, leveraging their larger datasets and their manufacturing expertise to deploy at scale, though for now both seem focused on the more lucrative and safer robotaxi market.

For now, Aurora in particular seems like the company to watch; as the last major player in the semi truck space, at least in North America, the area seems to be approaching a make-or-break moment.

---

This is part of a series on robotics in different industries. Previous entries are on construction robotics and military drones.