The Elusive Dream of Fully Autonomous Construction Vehicles

Khari Johnson

Just a few years ago, the promise seemed limitless: Automate cars and bring an end to traffic accidents, the biggest killer in the United States. Automate construction, with robot dozers, excavators, and other heavy machinery, and US housing and infrastructure shortfalls could be solved.

Built Robotics began testing autonomous excavators in 2017 with the goal of training machines to do more on construction sites. At the time, CEO Noah Ready-Campbell predicted that fully autonomous equipment would become commonplace on construction sites before fully autonomous cars hit public roads.

But after nearly seven years of digging trenches with autonomous excavators, Built Robotics last month announced plans to shift its focus from general construction projects to installation of solar farms. For that purpose, it rolled out RPD-35, a robotic pile driver that performs a single straightforward task, using a dull metal head fitted onto a mechanical arm to smash steel beams into the ground.

“It’s all solar all the time for us now,” Ready-Campbell says. “It may feel like we’re narrowing our focus, but I think solar over the next 10 to 20 years is going to become the story of our electrical grid as a country.” He says the shift is aimed in part at riding the coattails of infrastructure and climate-change funding passed by the US Congress last year, and that Built will now devote less time to other projects.

Some fully autonomous vehicles are on the road in China and the US, but just as the dream of driverless cars and robotaxis has yet to come to true, the promise of AI in construction appears to have fallen short.

Major companies like Caterpillar in the US, Doosan in Korea, and Volvo in Europe started experiments with fully autonomous heavy-duty machinery for construction sites a few years ago, but the idea has yet to advance beyond the prototype stage or become widely available.

Construction sites pose a difficult challenge for the developers of AI and robotics technology. Construction tasks often involve manipulating objects in 3D, and they take place on sites in a state of continuous change, while automation is most successful carrying out repetitive tasks with predictable outcomes. Making safe self-driving vehicles that travel on well-mapped public roads, which change more slowly, is in some ways easier.

Caterpillar is the world’s largest construction equipment manufacturer and has deep AI experience. The company won Darpa’s self-driving car challenge that involved a 142-mile race through the Mojave Desert in 2007, and it began experiments with autonomous trucks in the 1990s. Caterpillar customers currently operate nearly 600 autonomous trucks from the company in mines around the world, but the firm has yet to commercialize automated dozers or excavators for construction.

That’s despite setting a goal in 2020 to increase sales of software for control of autonomous machinery and make up for a decline in revenue from heavy-duty equipment. CTO Karl Weiss says that while Caterpillar has no semi-automated machinery routinely at work, it is testing heavy machinery automation on construction sites with a small number of partners, although he declines to share details.

Joseph Winters

David Nield

Adrienne So

Andrew Couts

When will those trials lead to autonomous construction machinery doing real work on construction sites? “We'll get there,” Weiss says, but Caterpillar’s partners need to be comfortable with the technology’s maturity. “There are risks involved, and we're on that journey, learning with them as we go, so that when we are commercially ready, they’re ready and comfortable with the product.”

Weiss says that Caterpillar started work on automating mining operations and construction sites around the same time, more than a decade ago, but automation happened quicker in mines for a few reasons.

First, mines have semi-permanent roads, and being underground lets you safely secure the area. And since mines are generally in remote places where it’s tough to house and feed people, automation can be more attractive. By contrast, construction sites are often short-lived and in a state of constant change, without permanent roads.

Caterpillar, along with startup Teleo, argues that the road to fully autonomous construction sites must first go through a phase where semi-automated equipment is operated remotely, by workers elsewhere. In this development stage, people with the necessary training can work with semi-autonomous machines anywhere in the world using an interface that resembles a video game, potentially even working from home. In parallel, AI experts will identify repetitive tasks suitable for automation.

Heavy machine operators today can choose to use some limited automation features, such as automatic grading to make surfaces flat when using a dozer. But the goal, Caterpillar chief engineer Michael Murphy says, is to allow one person to simultaneously operate four or five machines at a time by having algorithms take on much of the work.

Caterpillar equipment in automation experiments today resembles conventional machinery. However, Volvo and Bobcat parent company Doosan, which pledged to commercialize its autonomous Concept-X project by 2025, are already designing machines without cabins where a human operator sits.

Volvo Autonomous Solutions head of communications Ceren Wende says the company has a single cabin-less hauler at work in a limestone quarry in Switzerland and seven autonomous trucks in a mine in Norway but no autonomous heavy equipment operating on construction sites.

An excavator without a cabin for a human operator looks striking, says Anthony Levandowski, CEO of startup Pronto.ai, but he predicts that such machines are still “very, very far away” from widespread use.

Levandowski was once a founding member of the pep squad predicting that self-driving cars would soon take over. Before pleading guilty to taking confidential information from Google’s Waymo autonomous driving division (and receiving a pardon from former president Trump), he helped catalyze the automated driving industry when, in 2008, he programmed a self-driving Prius to cross the San Francisco-Oakland Bay Bridge (with a police escort) to deliver a pizza.

“I was like, ‘I think we're about two years away from having this commercializable,’” Levandowski says. “That was 15 years ago.” Today, he judges self-driving cars to have fallen behind. Pronto, like Caterpillar, focuses on automating trucks that travel on predetermined routes in mines and quarries.

Though the trucks can weigh more than 100 tons, it’s significantly easier than autonomous driving on public roads, because the vehicles operate on simpler, privately owned road networks. Employees are trained on how to behave and what to expect around the autonomous machines.

Levandowski says Pronto isn’t working on automating construction. He expects progress will be modest in the next few years, taking on simpler tasks such as automatic grading using a dozer and water trucks for dust suppression.

Built CEO Noah Ready-Campbell says his company’s research and development efforts are now focused on the robotic pile driver, despite the company’s history with automating dozers, skid steers, and excavators. Although the company showed it was possible to dig trenches with automated excavators, it ran into roadblocks when trying to convince customers to embrace automation. “You have to solve a big enough pain point to spur adoption,” Ready-Campbell says. “People will only change behavior if it's worth it.”