Amazon Reportedly Tests Humanoid Robots for Parcel Delivery
- Amazon is developing AI-based humanoid robots to deliver packages from the back of electric Rivian vans, potentially increasing delivery efficiency.
Amazon
wants to bring on new employees—albeit non-traditional ones—to deliver packages.
According to a report from The Information, the e-commerce giant is creating AI-based systems that would allow humanoid robots
to deliver packages to consumers’ doorsteps. The idea is that the robots, which Amazon is reportedly planning to train at a “humanoid park” in one of its San Francisco offices, would ride in the back of its electric Rivian vans. When needed, the robot would pop out of the back of the truck, bring the package to its final destination, then return to the truck for further deliveries.
The Information reported that once robots have been trained in the “humanoid park,” which has one of Amazon’s approximately 20,000 Rivian vans in it, the company intends to take them on “field trips” to real sites where they can make an effort to deliver parcels to residences.
More from Sourcing Journal
The newly reported effort is far from Amazon’s first foray into automated delivery technology; for several years, it has worked to build out its Prime Air business, most recently resuming operations after a voluntary two-month pause that saw it upgrading the software used to power its proprietary MK30 drones.
Theoretically, the humanoid robots could see Amazon delivery drivers being able to increase the number of parcels they deliver within one shift, if both the robot and the driver deliver parcels simultaneously. That kind of technology—though with a different type of robot— is being tested by last-mile delivery company Veho and AI-integrated robotics provider Rivr in Austin, Tex. already.
The Rivr robot rides along with a Veho driver, pops out to make deliveries in line with customer instructions, snaps a photo of the item for proof of delivery and makes its way back to the truck. Today, a human operator follows the robot to monitor how well it handles deliveries, but the idea is that, if the pilot goes smoothly, the partners will expand the technology into other areas. This type of use case for robotics is certainly gaining steam, but humanoids have different capabilities than other robots, which requires them to be trained differently.
SAP, Nvidia and Neura Robotics have committed to a partnership aimed at using Nvidia’s Omniverse digital twin landscape to train humanoid robots with business acumen. As an example, a humanoid working in a store in the future might be tasked with picking up a T-shirt for a customer’s in-store pickup. Rather than solely knowing that it needs to fetch a T-shirt, the humanoid would also understand the business context around that action. That in mind, it might understand that it needs to add more size mediums to the store’s stock, after picking up the last available unit for the consumer’s pick-up. If that same logic can be applied to humanoids delivering parcels in the field, it could increase their productivity and ensure that they meet consumers’ standards.
The Information’s report coincides with other AI-enabled logistics announcements from Amazon. The company announced at an event this week that it would use generative AI to create more detailed maps in an effort to aid delivery drivers on their routes. The updated maps would offer further information about issues a driver could face when working to deliver to a building, the shape of individual buildings and more. That kind of technology has been perpetuated by startups like Naurt , which provides more granular detail about the last 100 meters of a delivery driver’s route, even suggesting where they might park for best access to a building’s primary entrance in apartment buildings and office parks.
According to Reuters, Amazon’s robotics aspirations extend far beyond its delivery processes. The publication reported that Amazon will create a new group inside its Lab126 unit, which will focus on creating and training warehouse robots able to tackle multiple functions, rather than focusing on one repetitive task—which it plans to do through what it’s calling agentic AI. Agentic AI systems have the ability to make decisions and act autonomously on specific problems.
But today, these systems are more widely used in the backend of operations where humans have heavy involvement. For instance, an AI agent trained in logistics might autonomously create a batch of quotes based on inbound customer interest in less-than-truckload shipping. To see this kind of technology deployed within robots’ brains would be an interesting shift.
Reuters reported that Yash Dattatreya, a robotics scientist at Amazon, said the unit will work to create robots that aid human workers, by unloading trailers, among other responsibilities. Dattatreya contends that the machines might relieve some of the burden of heavy lifting or extremely fast processing during busy periods—like the holidays.
The development of the robots is still in early days, Dattatreya reportedly told Reuters.
Amazon did not immediately return Sourcing Journal’s request for comment on its robotics aspirations.
