[
Warehouse robots have traditionally performed best with predictable, repeatable work. Truck loading rarely follows a set pattern.
Boxes come in different sizes, shapes, and weights. Some are damaged, while others sag or have labels telling workers which side must stay up. Heavy parcels cannot always be stacked on top of lighter ones, and the inside of every trailer can look different.
Dexterity Inc. is using physical AI to take on that changing environment with its Mech robot, which the company described as a “superhumanoid.” Beckhoff Automation worked with Dexterity on the control architecture connecting the robot’s motion, safety, navigation, and data.
“It’s just like you’re playing Tetris,” Robert Sun, co-founder and founding engineer at Dexterity, told Automated Warehouse. “This is a weird game that you have to play or learn on the job that can’t really be automated, no matter how hard you think about it.”
Mech combines two robotic arms with a mobile base and AI designed to help it make decisions as conditions change. Sun said physical AI becomes useful when traditional automation reaches a “brick wall” and a task requires decisions that cannot easily be written into fixed rules.
“When the system has to really think about things, there’s a lot of different options,” said Sun.
Dexterity takes on loading, other warehouse jobs
Dexterity started in 2017 after its founders worked together in the robotics lab at Stanford University, where they first worked with EtherCAT, the open fieldbus communication standard developed by Beckhoff. EtherCAT has grown into a global standard for automation equipment, particularly mobile robots.
Initially, the company explored other warehouse applications before focusing more heavily on logistics work involving boxes, bags, mailers, and parcels.
Its main product is Mech, whose arms can work separately or together to handle tasks that might otherwise require two people.
handles packages in warehouse
applications. Source: Beckhoff
Automation
“It’s all about strength and the ability to do things that humans, ergonomically, are going to have very challenging times with,” said Sun.
Truck loading became one of Dexterity’s targets after the company began working with FedEx several years ago. The logistics company initially approached Dexterity about truck unloading, but Dexterity decided to also tackle loading.
Loading is more complex because the mobile manipulator must decide where each package belongs while accounting for weight, shape, condition, and orientation. A damaged box may need special handling. A heavy package should not be placed where it could fall on someone unloading the trailer. Other packages need to remain in a certain position.
Sun said Mech also encounters a wide mix of products, including heavy cases of canned drinks, shrink-wrapped packages, mailers, and bags.
While the robot is primarily used for truck loading and unloading, Dexterity is also applying the technology to order fulfillment, palletizing, and depalletizing.
Mech keeps humans in the process
Dexterity said it does not expect the Mech to solve every problem on its own. A box may arrive partially open, tape may be coming loose, or a product may be damaged or difficult for the robot to understand. When something unusual happens, the robot can ask for help.
Sun explained that the Redwood City, Calif.-based company has trained customers’ employees to act as robot pilots.
“Just like in Japanese anime, you have a pilot driving that can handle a lot of these tasks, but your Mech is augmenting your strength in a way,” he said.
Employees can monitor robots through a tablet, check their status, and step in when needed.
“The people can now use their human ingenuity to solve problems in a way, rather than their body to have to solve them,” said Sun.
Dexterity said it expects that relationship to become more digital over time. Instead of physically entering a trailer to address every problem, an operator could eventually give the robot instructions through a tablet or other interface.
Doug Schuchart, global industry segment manager for intralogistics at Beckhoff Automation, said robotics can also help warehouses manage changes in labor demand. Retail and e-commerce operations often need additional workers during peak periods such as the holidays. Robotics can help companies add capacity without the same swings in seasonal staffing.
“It is about also augmenting with robotics so that your general labor force is steady throughout the year,” said Schuchart.
Beckhoff helps connect AI to the physical world
Physical AI needs a steady flow of information about what is happening around the robot, which helped bring Dexterity and Beckhoff together.
Dexterity was already using the EtherCAT industrial Ethernet protocol as it developed the Mech. The company later worked with Beckhoff on an architecture that could bring programmable logic controller (PLC) functionality, motion, safety, and navigation into a single system while providing data to the AI.
Schuchart said EtherCAT provides the bandwidth needed to move large amounts of sensor and actuator data while still supporting real-time machine control.
“Physical AI, it requires data,” said Schuchart. “It’s all based on mining data and using it in the analytics for actionable insights.”
Beckhoff’s PC-based control platform also allows those functions to run together rather than using separate processors.
Safety across environments is another challenge
Safety adds another challenge because the Mech must operate in different customer environments. Dexterity and Beckhoff use Fail Safe over EtherCAT, or FSoE, to connect safety devices across the system. The technology allows Dexterity to adapt sensors and other safety equipment to the customer site rather than using one fixed setup.
One warehouse may already have gates or scanners in place, while another may require different equipment.
“It’s actually crazy difficult to have a programmable, modular safety system,” said Sun.
Schuchart emphasized the importance of flexibility during development.
“As a system is being developed and matured, it’s even more important to have that kind of flexibility with the safety system because things are always changing,” said Schuchart.
The amount of information available to the AI also matters. Sun said traditional industrial equipment can limit what information outside software can access. Physical AI benefits from access to sensor readings, actuator data, fault codes, and other information about how the equipment is operating.
“Being able to expose all of the information of the physical world is the most important part,” said Sun.
Rather than build every industrial component itself, Dexterity wanted its AI to communicate with established automation hardware.
“What Beckhoff has done with soft PLC, with EtherCAT, has allowed us to expose everything,” said Sun.
That information can help the AI respond when something goes wrong while giving operators and technicians more insight into what needs attention.
Rethinking warehouse operations with Dexterity
The longer-term opportunity may extend beyond automating individual warehouse tasks. Sun compared this shift to the transition from early roads for horses to modern infrastructure built for cars.
“Should we stop building interstates and build mechanical horses instead of cars?” said Sun. “We’re going to have to change up our infrastructure or change up the way we do things.”
Robotics could create a similar shift in logistics.
Warehouses, trailers, and material flow are largely designed around what people can physically do. Robots do not have the same limits. That could allow companies to rethink how products are moved, stacked, loaded, and unloaded rather than simply adding a robot to an existing manual process.
“You can transform the way people do logistics, material handling, material flow across the world,” said Sun.
Schuchart said physical AI offers companies another way to approach warehouse problems that have been difficult to automate.
“It’s easy to point out a problem. It’s a lot harder to solve the problem,” said Schuchart. “When I look at Dexterity, they’re one of the few companies out there that haven’t just identified a problem in the industry, they’re solving it.”