Amazon is working on a multimodal identification system (MMID) that allows it to dispense with barcodes in its warehouses and instead implement automatic item identification.
Barcodes, those labels with parallel vertical bars of different thickness and spaced apart that hold information, are not as reliable as Amazon would like, especially in the cataloging process in its warehouses, where handling objects can damage them.
The current process requires employees to scan the barcode of every item arriving at warehouses and throughout various points in the delivery process. “This process is repeated millions of times in a massive catalog of items of different shapes and sizes, and cannot be easily automated,” they point out from Amazon Robotics in a post on their website.
At the moment, the company notes that there is no robot that can “handle any item that might come into a warehouse and then scan it.” But it is determined to replace barcodes, and to do so, it works on multimodal identification.
This system is based on machine learning and uses a conveyor belt and a camera that photographs the item carried by each tray, in order to identify “virtual-physical discrepancies”, that is, the cases in which the items containing the trays do not match those picked up in inventory.
The company details that in the first tests, this system achieved coincidence rates of 75 and 80 percent, and that it currently achieves rates of 99 percent. The goal of this project is to combine this technology with robots to speed up and make package delivery more accurate.
This MMID system is being used in the logistics centers of Hamburg (Germany) and Barcelona, after a trial carried out first in a distribution center in Szczecin (Poland).