Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Deployment & production· 日本語で読む

ASICS cuts missing-inventory recovery time by about 60% with Corvus autonomous drones

ASICS America Corporation is operating Corvus One at its distribution center in Byhalia, Mississippi. At two high-risk storage areas, it increased inventory counts to monthly and cut the time to recover missing inventory by about 60%. It is redirecting about 200 hours of monthly counting work to inventory research and discrepancy resolution.

Key facts

  • Corvus Robotics announced on November 10, 2026, that ASICS America Corporation is operating Corvus One at its distribution center in Byhalia.
  • Inventory-count frequency at two high-risk storage areas increased from twice a year to monthly, a sixfold increase.
  • ASICS says the time to recover missing inventory fell by about 60%, and about 200 hours of monthly counting work were redirected to inventory research and other tasks.
  • The drone flies seven times per night and operates in aisles less than 6 feet wide without additional infrastructure inside the building.
  • Corvus Robotics says the drone detected 26 cases of missing inventory, including more than 200 pairs of shoes, over three months.

What happened

Corvus Robotics announced on November 10, 2026, that ASICS America Corporation is operating its autonomous drone, Corvus One, at its distribution center in Byhalia, Mississippi. The center handles about 3 million items across about 44,000 SKUs. The drone increased inventory counts at two high-risk storage areas from twice a year to monthly. ASICS says the time to recover missing inventory fell by about 60%, and about 200 hours of monthly inventory-counting labor were redirected to other inventory-management tasks.

Why it matters

ASICS says that moving to monthly inventory counts has made it about 60% faster to recover items whose locations are unknown, and that it now lists products on online marketplaces that it previously excluded from inventory information. For warehouse operators, this is an example of how increasing inventory-count frequency can affect not only the labor involved in inventory research but also how sellable inventory is handled.

Analysis

The central point in this case is less the drones' flight than the operational shift in inventory-counting cycles and the link between inventory discrepancies and information on stock available for shipping and sale. The two areas were previously counted twice a year, but are now counted monthly. ASICS says the time to recover items whose locations are unknown fell by about 60%. The operational change appears to be a shift from waiting for the next scheduled count to find an item to a system that detects discrepancies more frequently. The counting method also includes measures tailored to the site. In areas holding returned goods and individual items, multiple units of the same SKU may be in the same location, so scanning UPCs alone could lead to a mismatch with the boxes. Corvus One records UPCs and counts boxes on the same flight, using the box count as a basis for reconciliation. Staff also review images and video from nighttime flights the following morning, while many discrepancies are handled onboard. This is therefore a setup that automates repetitive physical checks and redirects people to discrepancy research, rather than one that fully replaces human labor. The reported redirection of about 200 hours a month is presented as an operational result. The connection to sellable inventory is another point. Before deployment, ASICS excluded 12 to 15 items per 100 from inventory information for online marketplaces because of fulfillment risk. After deployment, it says it lists all inventory information and sells the relevant products at regular prices. This is an example of how improvements in inventory counting can affect not only inventory accuracy but also the product selection shown on sales channels. The facility's inventory is managed in a Manhattan Associates warehouse management system, but the announcement does not say how the drone's count results are linked to that system. We covered the funding and leadership changes in Corvus Robotics raises $20 million, names Kabir as new CEO. This announcement offers a look at the company's business through a specific inventory-counting task at a logistics facility. It also raises the question of how robots connect with operational systems, as covered in Geek+ introduces integrated software platform for warehouse robots. The methods and performance conditions for integration in actual operations still need to be confirmed. Points to watch include integration of count results with the warehouse management system, how detection accuracy is evaluated, and the operational burden if monthly counting is expanded to more areas. The announcement does not provide figures or conditions on these points.

Details

The facility holds 2.3 million pairs of shoes and 700,000 apparel and accessory items. Inventory is managed in a Manhattan Associates warehouse management system. Corvus One operates in the "singles area," which handles returned goods and individual items, and in "case reserve." On each flight, it scans UPCs while counting boxes. In the singles area, it compares box counts with scan counts, including when multiple units of the same SKU are present. ASICS schedules seven flights per night, and the following morning staff review the captured images and video, with many discrepancies handled onboard. The aisles flown by the drone are less than 6 feet wide. The aircraft uses onboard computer vision and spatial reasoning, and operates without additional localization infrastructure inside the building. Corvus says that over three months the drone detected 26 cases of missing inventory, including more than 200 pairs of shoes.

Sources

Translated from the Japanese original: /news/asics-corvus-no-jiritsu-hikou-doroon-de-zaiko-kaifukuji-kan-3c8529. The Japanese version is authoritative if they differ.