MT Magazine January/February 2026

JANUARY/FEBRUARY 2026

FEATURE STORY

15

operations generally don’t have people to spearhead these automation projects, which can require a substantial amount of time. So, Kelly said, the cost of implementation needs to come down; the time for deployment needs to collapse; and the overall ease of use of robots needs to improve. He said that he sees a lot of evidence that things are moving in that direction. The Amazon Effect One of the places where Kelly looks is probably a place that most people at shops – top ones and others alike – don’t look: Amazon. Amazon operates more than 300 facilities worldwide that utilize robots – both commercial models you likely recognize and robots that Amazon itself is developing. With more than a million robots, the company claims to be “the world’s largest manufacturer and operator of mobile robots.” Its fleet of robots includes Hercules, which can lift and move 1,250 pounds of inventory, and Proteus, a fully autonomous mobile robot that can move heavy carts full of goods and maneuver around employees while doing so. With so many robots in its facilities, Amazon has developed DeepFleet to coordinate traffic and avoid congestion. The company calculates that this suite of AI-based foundation models improves the travel time of the fleet by 10%. Additionally, Amazon recently announced the development of Blue Jay, which it describes as a “next-generation robotics system that coordinates multiple robotic arms to perform many tasks at once.” While this may not be useful in a physical shop – after all, Amazon fulfillment centers have high-speed conveyors that move thousands of items – conceptually, it is: Blue Jay consolidates three robotic workstations into one. Another example of what is sometimes referred to as “physical AI” (AI that works with machines, such as robots or vehicles, to perform tasks in the real world, as opposed to the digital space) is Vulcan, a robot with a sense of touch. Its end effector includes force feedback sensors, so it knows when it makes contact with an object and how much force is being applied. Vision allows it to see items in a container; a rod allows it to move the items, creating space for new items or to allow accessibility for removal, which is performed by a suction cup on the end effector. While Vulcan can handle about 75% of the items processed in Amazon fulfillment centers, it is smart enough to know when it can or can’t perform its required tasks. When it encounters something it can’t handle, it alerts human employees for help. Although these systems are developed for warehouse situations, they are likely to lead to a proliferation of operations, including in manufacturing. Architecturally, the robots being used at Amazon are either like the ones being used in manufacturing operations (e.g., 6-axis arms) or are flat and have wheels (for carrying things on their surfaces).

Optimus under its “skin.” It probably isn’t coming to a shop near you anytime soon. (Image: Tesla)