JD.com’s Seven Fresh banner is using dual-arm robots on an operating supermarket floor rather than only in a warehouse cell, according to autonews.gasgoo.com.
That report said JD.com and DexTeleop signed a strategic partnership on August 12, 2026 covering product development, technical integration, marketing, channels and industrial infrastructure. The first named site is a Seven Fresh store in Taiyuan, where DexTeleop TA-series machines are assigned food-and-beverage sample delivery, shelf restocking and customer-flow guidance.
Those tasks mix navigation, manipulation and close work around shoppers. Sample delivery is largely a path-and-handoff problem. Restocking requires contact with shelves, packs and fixtures. Guidance puts the robot among people who are not trained operators.
The hardware pairs two arms with lifting equipment and an omnidirectional base, so it is not limited to moving a tote between fixed stations. The store layout is treated as given rather than rebuilt around a robot path.
A live grocery floor is a hard test: mixed packaging, blocked aisles, sudden stops, children near the machine, damaged goods and crowded facings. The Taiyuan site is therefore better read as a data-collection deployment than as proof of labor replacement. Further units are planned for Seven Fresh stores in Beijing and Tianjin to gather more real-world operating data before any broader rollout is described.
For JD, the arrangement is also a chance to see whether one embodied-AI stack can span supermarkets, warehouses and logistics, which impose different safety, speed and manipulation demands. A skill learned in a backroom may not transfer to an aisle; shared data and integration can still lower the cost of trying.
Public material does not give unit counts, autonomous hours, restock success rates or how much teleoperation and human intervention remain. Those figures would show how independent the system actually is.
Safety is equally material. A two-arm machine carrying products has a different risk profile than a slow shelf scanner. Navigation, stopping distance, force limits and behavior when a customer enters the work area are part of the product.
The Taiyuan work is notable because dual-arm robots are performing several store tasks in front of shoppers, and the next step is explicitly more operating data in more cities. Whether that envelope travels with the robot, rather than with one well-supported store, is what Beijing and Tianjin should test.