LiDAR (laser radar) beams of a Waymo autonomous rotate at eye level while it waits at an intersection in the Mission District of San Francisco in June of 2022. LiDAR systems are how autonomous taxi's and trucks see a three-dimensional environment in real time. Waymo is already operating an entire fleet of fully driverless taxi's in Phoenix, Arizona.
A Crash Course in Autonomous Trucking
By James Year
There are critical information gaps within the trucking community about how and when self-driving trucks or autonomous trucks will arrive. One thing is certain: the profitability of these machines will dictate their arrival. Autonomous trucks are already hauling freight. The first autonomous trucking company will be launching commercially in the fall of 2024. The goal of this article is to condense what drivers need to know, while providing some resources for additional reading and discussion.
Driving Down Freight Rates
The tech developers are telling their investors some interesting things about this technology. According to the investor presentations, an autonomous truck will allegedly be able to cut up to 45 percent of current driver related per-mile freight costs. Analysts have found that the total operating costs could be reduced anywhere between 15 and 45 percent, depending on a multitude of factors. There are no human driven trucking companies that will be able to compete with that capability. Family owned and independently run trucking companies will not be able to afford the equipment and insurances costs to compete in this new high-tech market.
The problem that any type of automation is trying to solve is the human problem. This view-point is consistent throughout autonomous trucking developers.
Aurora Trucking stated in their 2021 investor memo that they believe their technology will create a virtuous cycle. This means that autonomous trucks will create a self-reinforcing chain of events where that technology will effectively create its own demand. For drivers, this translates to if one company adopts it and lowers rates, they will be forcing their competition to follow suit or risk being put out of business.
Aurora expects this cycle to consolidate the self-driving industry with “winner takes most dynamics.” Aurora Trucking even want so far to compare this massive economic potential to the digitalization of advertising where Google and Facebook now dominate over 70 percent of the advertising market.
Aurora believes that autonomous trucks will create a "virtuous cycle." This means that it will create a domino effect in the market where autonomous trucks will create their own demand based on their speed and cost savings.
The other advantage that self driving trucks have is that they won’t have to be restricted by Hours of Service regulations. They are being advertised to opperate around the clock. The utilization rate of autonomous trucks will be up to 2.5 times higher than a human driven truck. The implication here is that 1,000 autonomous trucks could be able to do the work of 2,000 to 2,500 human drivers. This capability, coupled with fuel and labor cost savings, risks flooding the market with cheap freight.
"Transportation is one of the largest sectors of the economy globally and has the potential to create enormous value for investors as it becomes automated," said Aurora Trucking in their 2021 Investor presentation. In this slide Aurora is comparing the monopolization of advertising to the potential of autonomous trucking.
Methods, Regions, and Timeline
Autonomous trucks are targeting the “simplest” aspects of trucking first. Predominantly in long haul and line haul, and especially in the dry van, reefer, and parcel markets. The most likely way that this will work is through a “transfer hub model.” Autonomous trucks will operate in a relay, focusing on the middle mile, from onramp to offramp on interstates. Then human drivers will take it to the final destination. Tech developers have stated that the transfer hub model plays to the strengths of autonomous trucks by having them focus in a much more structured, predictable, and safer environment, compared to urban driving.
But one of the things that isn't being said to regulators, truckers, and the general public is that this transfer hub model is intended to be temporary. In 2022, Chris Urmson, CEO of Aurora Trucking, told investors that this technology will soon become capable of going from warehouse-to-warehouse and eventually dock-to-dock. That statement undermines what Aurora Trucking, The Autonomous Vehicle Industry Association and the American Trucking Association claimed to the House Transportation and Infrastructure Committee in September 2023, that autonomous trucking will allegedly create more, better, and higher paying local driving jobs for truckers. That claim also ignores other tech companies like Nuro and Gatik.AI, who are automating those very same last mile delivery and local driving jobs.
Autonomous trucking developers are targeting the middle mile first to maximize the return on investment and mitigate liability through this "transfer hub model." At scale, this could risk creating a labor surplus in affected markets by pushing more drivers to compete for local driving jobs, while also consolidating trucking jobs into metropolitan areas.
The transfer hub model can also work in tandem with a method called platooning, where a human driver will lead a “platoon” of several autonomous trucks directly behind them. This method also allows for even better fuel economy from very close following distances and reduces wind resistance, despite the inherent safety risks and increased wear and tear on public infrastructure. Both of these methods will restructure the trucking industry, not only in terms of business but also in geography. If the transfer hubs are concentrated near large urban areas, it could effectively cut off rural family-owned trucking businesses, while simultaneously lowering the freight rates they depend on to survive.
Aurora Trucking and several of their competitors are focusing on Texas and the southern United States first. This is mainly due to favorable weather and permissive regulations in those states. They are also targeting high density freight lanes (which are around the most people) to maximize their return on investment. Aurora Trucking and Kodiak Robotics are expecting to launch commercially on I-45 between Dallas and Houston by the end of 2024. Depending on the success of that launch, Aurora has manufacturing partnerships with Volvo and Paccar to scale production rapidly. Based on their 2021 investor report, they expect their autonomous lanes to reach the northern U.S. and New England by 2030. Aurora's 2024 investor presentation claimed that they intent to hit 2 billion driverless miles by the end of 2028.
Autonomous trucks will first appear commercially in Texas, then spread throughout the southern United States due to weather considerations. The developers usually are targeting high density freight routes first to maximize their return on investment. Texas is also ideal because it features those high-density freight routes within a single state and has permissive regulatory policies.
It’s also important to note that Aurora has several competitors, primarily Plus.AI, Torc Robotics and Kodiak. In 2021, Amazon placed an order to Plus.AI for 1,000 retrofit kits to begin upgrading their existing fleet with a “driver-in solution,” that will kick out the driver when those trucks have accumulated enough real-world driver data. According to reporting from Freightwaves, that order was being filled at scale as of 2022 and Plus.AI claims to have over 10,000 orders between the U.S. and China. Aurora may be the first to kick the driver out, but Plus.AI and Amazon could be the first to do it at scale, when the algorithm is deemed "acceptably safe" for business.
Like the space race, autonomous driving technologies are also dual use in nature and can be adapted for military purposes. The Autonomous Vehicle Industry Association, or AVIA, has been lobbying federal legislators to ensure their member's technology is able to be sold in foreign markets. One of their members, Kodiak Trucking, has government contracts to apply their technology to autonomous weapon systems for the US Military. Several other autonomous trucking companies like TuSimple and Plus are also intending to sell their autonomous systems in China.
The low cost ($17,500) and fast installation of Plus.AI's retrofit kits for autonomous trucks gives this company a significant advantage by scaling far faster than their competition. Amazon ordered 1,000 units in 2021. Plus's business model plans to gradually evolve into a driverless system with currently existing truck fleets.
The Driverless Road to Market
Self-driving truck have to overcome significant obstacles to make them commercially viable. The hardware is the easy part. Each self-driving truck uses multiple sets of cameras, radars, and LiDAR’s (laser radar) to work together to perceive a three-dimensional world around them in real time. That data is fed to an onboard computer algorithm which decides how to interpret it and make educated decisions based on the algorithm's training data. Getting that hardware to “think” and act appropriately with the right training data on our interstates is the hard part.
An Aurora autonomouos truck heads south on I-45 from its hub in Palmer, Texas in July of 2023. Autonomous trucks can be identified by large sensor pods that are either above the driver and passenger doors (Aurora) or placed on the front of the rearview mirrors. (Kodiak and Plus.AI)
The developers are close to solving that problem. But there is a famous saying in the tech industry from Tom Cargill, of Bell Labs, who said, “The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time.” Solving the last few percentage points in autonomous trucking is a monumental undertaking. This will only be solved by accumulating gargantuan amounts of data and then distilling the algorithm down to a size that can be run by commercially available computer hardware.
Plus.AI’s strategy to get that volume of data is to mass produce inexpensive retrofit kits and launch a system at scale early to reach their goal of 8 billion real world driving miles. Aurora Trucking’s strategy differs by creating purpose-built machines and getting that data through “simulated miles” with virtual driving programs, similar to video games. As of 2021, they were generating 22 million miles of simulation data per day to train their algorithm. Aurora Trucking claims these simulations give them an advantage by being able to replicate and train their algorithm repeatedly on rare incidents. This method may very well create some advantages. But a computer simulation is a programmer's interpretation of the real world. Reality is excruciatingly difficult to program.
The lateral rotating beams of Aurora's First Light LIDAR (laser radar) as seen reflecting off their trailer during a run to Houston, Texas on I-45 in July of 2023. These systems help build a three-dimensional environment in real time to help the system navigate and percieve threats.
Currently, the leading autonomous trucks developers are getting close to delivering a commercially ready product. Aurora Trucking claimed in their second quarter of 2023 shareholder letter, that 50 percent of their self-driving systems were operating at 100 percent autonomy, with no human intervention. The other 25% of their fleet was operating 99 percent autonomy. The effectiveness of the remaining 25% of their fleet wasn't listed, but they did say they had zero critical safety interventions within that quarter and that their aggregate "autonomy performance indicator" or API metric was at 97 percent autonomy. If true, it appears that this company is getting close to solving autonomous trucking.
This is an imperfect example but even if these systems are 99 percent perfect, over the course of a thousand miles that means that driver intervention would still be required for ten miles along that route. As any driver knows, a lot can happen in that time, let alone in less than 50 feet. Aurora has also stated that these systems will never reach a full 100 percent rate, due to mechanical breakdowns, blown tires, and other unforeseen circumstances. Regardless of all the percentage points and safety statistics, there is no getting around it. Machines fail and Americans will be killed by driverless trucks. The path to autonomous trucking currently lies at reducing risk and liability to become "acceptably safe" for business.
A Waymo autonomous taxi as seen in San Francisco in June of 2022. Waymo is owned by Google's parent company Alphabet and is currently operating a fully driverless fleet of taxi's in Phoenix Arizona. The sign is from Workato and says "Automate the work out of finance. Integrate anything. Automate everything."
The Biggest Picture
Autonomous trucks are a symptom of a much larger problem that Americans are already beginning to face. These same artificial intelligence (A.I.) technologies are being applied to every market simultaneously and will create substantial economic disruptions. And we haven't even touched the similarly transformative and world changing impacts from quantum computing and bipedal robotics, which could also be commercially viable by the end of the decade... Alarm bells have been ringing in academia for years, regarding the rate and scale of this change. Unfortunately, it appears the risks aren't being taken seriously.
Historically, new technologies like the steam engine, typically lead to short term disruptions and economic pain for those displaced. But eventually, often after generations, new jobs were created, and quality of life increased for everyone over the long term. The concern is that the exponential growth in capability of A.I. will prevent those new jobs and opportunities from materializing over the long term. As it further develops, A.I. will increasingly be thumbing the economic scales away from humans and in favor of machines, by being progressively faster, cheaper, and more accurate for a permanently growing set of tasks. When that occurs, large-scale technological unemployment will likely result and overturn 300 years of historical job creating precedent for the first time. This trend will create severe socioeconomic and political problems.
Discussions are already occurring about how to deal with this issue. The most popular idea is a Universal Basic Income, where everyone is paid a set amount of money at regular intervals from the profits from A.I. Elon Musk, Mark Zuckerberg, Bill Gates, Jeff Bezos, Tim Cook, Larry Page, and Sam Altman are part of a growing list of both conservative and liberal advocates for a UBI, or other similarly structured proposals. However, Kai Fu Lee, founder of Google China, criticized the idea in 2019 by saying, “It’s a painkiller, something to numb and sedate the people who have been hurt by the adoption of AI. And that numbing effect goes both ways: not only does it ease the pain for those displaced by technology; it also assuages the conscience of those who do the displacing.”
Industrial support for a Universal Basic Income, or other similarly structured proposals, are something of an admission to a future that warrants serious debate and consent by those who will be negatively affected by artificial intelligence. A UBI isn’t needed if people can pay their bills and buy products to keep the economy going. And there is no guarantee that those displaced will be satiated by a UBI if they still have to accept downward social mobility after being displaced. The future that we’re currently headed towards is expected to exacerbate America’s current social, political, and economic inequalities, while also creating a suite of other expected and unforeseen problems.
In the end, A.I. may be the tech-based "savior" that we need. But having all of us reach that point intact will be the hard part. America needs an honest discussion about the economic risks to people's livelihoods and about how we can avoid the worst aspects of this change.
Hopefully, we can find an outcome that we can all agree to live with.
*This article leaves out other critical issues regarding environmental impact, economic equality, national security, and criminal opportunities with this technology. These potential issues as they relate to trucking are highlighted in the next article titled, “The Virtuous Cycles: A Hypothetical.”