Developers: | Amazon |
Branches: | Electrical and Microelectronics |
Technology: | Robotics, Warehouse automation |
Content |
History
2025: Slow work and spoilage of goods
In mid-May 2025, it became known that warehouse robots Amazon they were not able to completely replace people when performing everyday operations. Machines in many cases spoil goods and work slower than ordinary employees.
Amazon warehouses use Stow and Pick robots. Devices of the first type are equipped with a grip and a retractable bar for working with objects in containers, as well as a visual perception system for assessing the available space. A machine learning model is used to predict the probability of successful packaging based on the current contents of containers. The robot can optimize the space inside the container to add new items.
Data on the movement of more than 500 thousand goods indicate that Stow successfully copes with the task in 85% of cases. According to The Register, 9% of failures recorded damage to objects - mainly due to falling on the floor. In 14% of cases, problems arose when working with books: laying with a robot led to page damage. In terms of speed, Stow machines are inferior to humans: robots lay goods at an average speed of 224 units per hour, while warehouse employees reach 243 units per hour.
At the same time, people show more differences in laying speed - they quickly cope with small items, but spend more time for large items or when they have to squat and climb stairs to get to containers.
In the case of Pick robots, the success rate for tasks reaches 91% (based on 12 thousand selection attempts). At the same time, installations of this type rejected 19.4% of selection requests due to the fact that machine vision did not recognize the object, or due to the risk of damage.[1]