For the past few years, the Delivering the Future event has showcased the latest technology that powers Amazon operations. The company's 2023 event in Seattle featured updates on its pharmacy products and drone delivery.
This year in Nashville, Amazon discussed AI updates to the shopping experience and how the company is using computer vision to further reduce package delivery times. While the company didn't showcase any new robotic systems during the two-day event, it did provide important insight into how it integrates with existing products.
Immediately after the event, TechCrunch spoke with Tye Brady, principal engineer at Amazon Robotics. This conversation has become an annual event for several years now, giving us the opportunity to dig deeper into how Amazon's robotics story has changed in the past 12 months and how it will evolve in the next 12 months. He gave it to me.
According to Amazon's internal statistics, the company currently has more than 750,000 robots in its U.S. fulfillment centers. This is the same public persona the company promoted in 2023. But this is not all. This 750,000 number only includes the company's autonomous mobile robots (AMRs).
These are the familiar wheeled systems that have been roaming Amazon's warehouse floors since the company acquired Kiva Systems in 2012. These tote robots (which the company also calls drivetrain systems) include a variety of models, including the autonomous Proteus, which was unveiled at Re:Mars in 2022.
Image credit: Amazon
While ARM makes up the majority of Amazon's fleet, other form factors also have their own space on the floor. The next largest category is Amazon's robotic arms, which currently include Robin, Cardinal, and Sparrow, each tasked with sorting and stacking objects.
The latest entry into the Amazon Robotics family is Sequoia, which was announced at the 2023 Delivering the Future event. The name, borrowed from the giant sequoia trees of Northern California, may be an homage to the size and scope of this system. Sequoia is an automated storage and retrieval system similar in principle to those offered by companies such as AutoStore.
The first Sequoia systems came online in 2023 at fulfillment centers in the Houston area. Amazon announced Wednesday that a system five times its size forms the heart of a massive warehouse in Shreveport, Louisiana. While the fulfillment center itself is not new, it has been significantly expanded to over 3 million square feet.
Amazon certainly has the resources to build new greenfield robotic fulfillment centers from the ground up. Instead, the large retailer is focusing on renovating existing brownfield warehouses. This method is less resource intensive, but requires the company to avoid the existing delivery task of “fixing the plane while it's in flight,” as Brady puts it.
Image credit: Amazon
Amazon's first “Gen 12” building, the Shreveport center, will eventually have 10 times as many robots as its previous facility, although the company has not yet disclosed that number. Brady added that in addition to these new robots, more robot-centric jobs will also be created. This means that the role of RME (Reliability Maintenance Engineering) will be 25% more than before.
The company added that it plans to hire 2,500 people once its fulfillment operation, which is equivalent to 55 football fields in Shreveport, is up and running. Automation fans will say that these technologies will allow humans to focus on things that robots can't do. Brady is a big proponent of this concept.
When asked what jobs humans are still suitable for, he replies: “Problem solving, common sense, thinking rationally, understanding the big picture, understanding context. There's some physical work as well.”
Agility's Digit robot made headlines at the 2023 event, but Amazon didn't talk much about the humanoid aspect. To be sure, the company has been exploring the role bipedal robots could play in its fulfillment centers, including with the Agility pilot it announced last year. However, since the completion of the pilot, things have been uneventful.
Image credit: Amazon
“We’re still learning,” Brady says of the agility partnership. “It's slow and steady. 'Research and development' is the best way I can capture it. ” This slowness, he explains, is a result of finding ways to incorporate such technology into existing workflows.
“Start with the problem you're trying to solve,” Brady says. “When you get a piece of technology and you think, 'How can I apply this?' and you try to force things, that's a dangerous path. In fact, our distribution centers have concrete There's a lot of great flooring to drive through. The wheels are pretty good. But there are also stairs and uneven terrain once you start getting out.
Brady confirmed that the partnership is still in effect, but had no additional information to share with TechCrunch.
One partnership that has become more apparent in recent months is Covariant, a spinoff of the University of California, Berkeley. Amazon announced in August that it had hired founders Peter Abbeer, Peter Chen and Rocky Duan, as well as about 25% of its workforce. This move is an effort to expand the role of foundational models in industrial environments.
Image credit: Amazon
Amazon, for example, says its Sparrow robotic arm “can now handle more than 200 million unique products of all shapes, sizes, and weights.” But edge cases always exist. These will be handled by both human employees and better-trained AI systems. Covariant plays a key role by working with these types of large datasets to fine-tune product selection, placement, and more.
“We're on track to start working on some very deep, very applied machine learning problems,” Brady said of the deal.