In recent years, warehouses around the world have automated almost everything, except the most difficult part, separating individual items inside boxes, it is the last frontier of logistics automation, a bottleneck so complex that it still depended almost entirely on human hands.
But at ICRA 2025, <a href="
https://www.youtube.com/@ROBOTERA-pg4ew/videos"RobotEra claimed to have crossed that line for the first time, the company presented the Star Act L7, a life-size humanoid robot that combines industrial hardware with a fully integrated language-action vision model capable of interpreting the environment, understanding commands and acting directly on shelves.
According to RobotEra, it is the first real application of a built-in VLA in the logistics sector and not just a laboratory demonstration, the tobot was built to fit perfectly into what already exists in warehouses, the waist with 3 degrees of freedom reaches shelves of up to 2 m.
The Star Act X Hand One uses five fingers and 12 joints to manipulate everything from square boxes to irregular packaging and the cross-shaped wrist avoids blind spots when entering compartments. The entire structure was designed to face the real chaos of inventories, reflected light, slippery surfaces, stacked objects and unpredictable shapes, but what really differentiates the L7 is the brain, the Era-42 with its VLA model does not depend on scripts or rigid coordinates, it interprets images and natural language. and transforms it directly into motor movement, without intermediate stages.
That end-to-end flow reduces latency and allows for quick adjustments if something moves, falls or is misplaced. Although the market is full of bold promises such as companies offering millions of robots for the next few years, Robotera chose a more pragmatic path, attacking a specific expensive and high-impact problem and if this really scales, the last manual part of the logistics could disappear faster than the sector expected.
