Physical AI is being applied to the complex task of automated trailer loading. FedEx Corp. yesterday said that it has expanded its collaboration with Dexterity Inc. to scale Dexterity’s Foresight world model and Mech trailer-loading systems at the FedEx Hagerstown Hub in Maryland.
“FedEx has a long history of using technology to make our network safer, smarter, and more resilient,” said Kawal Preet, FedEx executive vice president for planning, engineering, and transformation. “As we continue evaluating and deploying physical AI in real-world environments, we see opportunities to enhance the speed, reliability, and flexibility of our network, creating long-term value for our business and customers.”
Memphis, Tenn.-based FedEx offers transportation, e-commerce, and business services to customers worldwide. It reported annual revenue of $86 billion and has more than 450,000 employees. In February, the company implemented Berkshire Grey Inc.‘s Scoop autonomous unloading system as part of a multiyear collaboration.
Dexterity takes on physical AI challenges
Dexterity’s Mech dual-armed robot is designed for heavy industrial operations, while remaining compact enough to operate inside trailers. The Redwood City, Calif.-based company said its Foresight world model combines vision, depth, and touch to predict physical actions and make real-time decisions in dynamic environments.
“I’ve spent 18 years trying to understand human motion and translate skills into robotics, with a particular view around physical AI,” said Samir Menon, Ph.D., founder and CEO of Dexterity. “When you close your eyes and pick up something, you’re activating a world model. Once a robot has that understanding of the world, it can have many skills.”
In autonomous trailer loading, Foresight reasons across three spatial dimensions and time to optimize how packages are placed for space, stability, and speed across a wide range of operating conditions.
“Lots of people are doing robotic unloading, but we need a world model to think many steps in to the future,” Menon told Automated Warehouse. “We picked the loading application because it’s 100 times harder than unloading. With boxes of random shapes, sizes, and weights, our system can see the boxes and has the information to go piece by piece, thinking several moves ahead. Our skill models are trained across a variety of robots.”
He asserted that world models such as Foresight will bridge the divide between the digital and physical worlds, enabling robots to do everything from unloading trucks to handling luggage at airports and constructing solar panel arrays.
FedEx is ready to scale automated trailer loading
“Trailer loading remains one of the most physically demanding and challenging workflows in parcel logistics and is fundamental to FedEx operations,” said FedEx. “It requires physical strength, endurance, and real-time problem solving, and is one of the top areas where team members seek additional support.”
The company noted that this process, which involves tens of thousands of trailers loaded daily across its U.S. network, has been difficult to automate. FedEx and Dexterity said they have spent years on joint development, testing, and operational validation.
The partners said the Hagerstown deployment significantly expands the program beyond the pilot site. It enables production at a larger operational scale, with a continued focus on improving safety, consistency, and performance.
Through its collaboration with Dexterity, FedEx said it is establishing not only how physical AI performs in trailer-loading operations, but also how the technology integrates into broader hub operations, including destination planning, trailer assignment, maintenance, and workforce processes.
“FedEx has been an extraordinary partner, bringing deep operational expertise, an unwavering focus on safety, and a clear vision for how physical AI fits into the future of logistics,” stated Menon. “Hagerstown represents an important milestone in demonstrating how physical AI can operate reliably in a high-volume logistics environment, creating a blueprint for network expansion.”