Foundational Learning: Bedrock Robotics is Building Something Big 

“Our opinion is that this thing's ready to go,” says Kevin Peterson says of Bedrock Robotics’ autonomous driving for heavy equipment. The CTO adds, however, that while the technology is mostly where it should be, the Bay Area-based construction startup is approaching commercialization deliberately. 

“We're going to take time,” he tells me. “It's a business and it's a technology and you have to make sure it's really, really robust, performant, and useful. All of that polish takes a lot of time, but it's not really a science project in the way that some of these things have been.” 

The company’s first offering is Bedrock Operator, a hardware system that retrofits standard heavy construction equipment with autonomous navigation. This arrives courtesy of eight cameras, GPS, and LiDAR, that are distributed along a roof rack that is mounted atop the vehicle’s operator cab. A combination of on-board edge and cloud computing gets the system where it needs to go. 

While eventual plans will cover various different machines, the current work is focused on excavators. Their pivoting bodies and earth-moving claw arms present an added challenge for autonomous systems navigating in uncertain terrain. “It’s a big arm, essentially,” says Peterson. “I think of it as five-degree freedom arm with, it has tracks, you know, tracks wheels. It's fine. And so all the techniques that you're seeing around doing manipulation apply.” 

While it’s not shipping, Bedford’s Operator system has been doing a good deal of work out in the field, according to the CTO.  

“If I were to go visit one of your test sites right now with one of these systems in an excavator, it would be working pretty closely to what you would expect it to be in its final version,” says Peterson. “We have machines that have been running on customer sites autonomously. We finished up a job that we like to talk about last fall where we did 10% of the work for the customer. And big job. It's a big manufacturing facility. We did 65,000- 70,000 cubic yards of work.” 

In its present iteration, Bedrock Operator requires a human in the loop for supervision to assure it’s operating as intended. While the safety concerns aren’t quite as high as city streets, obstacles and unknowns continue to be a concern in a construction site. He cites one specific example that has become a well-known edge case among long-haul truckers traversing the United States  

“One of the things that we're doing currently is validating our detection system, collision system, collision avoidance system,” Peterson adds. “You want to do that very, very carefully in self-driving. There's a famous dog walker that walks between Dallas and Houston and on the side of the highway, every day. You talk to people at any of the trucking companies, they know exactly who I'm talking about. There's one of them.” 

While maintaining caution, Bedrock has developed with remarkable speed. The startup was founded in May 2024, coming out of stealth with an $80 million funding announcement last July. It made waves again earlier this month with a $270 million Series B raise, bringing its to-date funding up to $350 million. The team’s pedigree no doubt played a large role in stoking early consumer confidence. Both Peterson and cofounding CEO Boris Sofman left high-profile engineering roles at Waymo to found the startup.  

Peterson adds that hardware, software, and training technologies have matured considerably since he first began working in self-driving as a PhD at Carnegie Mellon, back in 2005.  

“When you start these things, you're not always certain who's going to join you and who's going to come along for the ride,” he says. “We just have insanely talented people who are very passionate, and they work super hard. I think that technology really is ready. It's not just a story.”