6 Robots work in convenience stores making coffee and sorting merchandise. 2 million people “monitored” online

AI summary

Summary of article:

Shanghai Xuhui convenience store used 6 humanoid robots for 24-hour store operations, completing tasks such as greeting guests, coffee, tallying, cashiering and cleaning. 4 of them were successfully converted into regular employees, demonstrating the potential of the retail landscape.

A convenience store in Xuhui, Shanghai recently welcomed special staff: 6 humanoid robots “on duty” 24 hours a day, responsible for welcoming guests,coffee Production, tallying, cashiering and cleaning, a total of 1,730 tasks have been completed, with over 2 million views online. The test ended at 10 am on November 3, and 4 out of 6 robots were successfully “regularized”, demonstrating the progress of the application of humanoid robots in retail scenarios.

Youth Daily reported that testing began at 10 a.m. on November 2, with six Zhiyuan Elf G2 humanoid robots working together: Ahao welcomes guests, Huobao and Xuanwang do the tally, Kashao makes coffee, Dingxiang is responsible for cashiering and cooked food services, and Sister Leng cleans up. Within 24 hours, the robot visited customers 683 times, picked up and delivered 369 items, made coffee 147 times and completed 277 replenishment inspections. During the live broadcast, more than 2 million people “observed the work” online.

The specially invited experience officer deliberately created difficult problems, mixed sweet and sugar-free colas, and asked to take only the sugar-free version; This includes disrupting chewing gum shelves, testing whether the robot can return to its position, and even standing in the middle of the aisle or placing cartons to block the path. The robot once complained: “You are very brave.” However, it still confirmed the products through voice commands and visual recognition, sorted the shelves one by one, stopped first when encountering pedestrians or obstacles, and then adjusted its route to go around.

A seemingly simple convenience store job is not easy for robots. When lifting the golden cake, if the force is too light, it will slip easily, and if it is too heavy, it may be crushed. When hanging rainbow candies on a hook, if the angle is slightly off, it may catch the products next to it. The display of products and customer flow in the store are constantly changing, and replenishment, cleaning and picking up goods can also interfere with each other, which greatly tests the robot’s adaptability.

According to reports, the six robots share the embodied intelligence framework Genie Master as the “team brain”, which connects task decision, action execution and result feedback through the EEP execution protocol. The system arranges the work according to the on-site situation, and then decides to continue, correct, or delegate to other robots based on the results. Each operation will be reviewed and reviewed, the effective experience will be stored in long-term memory, and the correct actions will be used as training materials to reduce repeated mistakes.

According to the report, the Shanghai Municipal Commission of Economy and Information Technology launched the “Special Action for Real-Scene Training of Humanoid Robots and Embodied Intelligence” in June 2026, listing catering retail as a key scenario, and proposed to create more than 100 high-value application scenarios by the end of 2026, thereby starting the implementation of 10,000-unit scale Gone. This test shows that industry is promoting humanoid robots for generalized business applications through real-environment training.

Six robots work in a Shanghai convenience store. (Taken from Youth Daily)

Summary FAQ


  • Six Ziyuan Genie G2 humanoid robots work continuously 24 hours a day at a Zhuhui convenience store in Shanghai. They are responsible for welcoming guests, making coffee, tallying, cashiering, cooked food service and cleaning and have completed a total of 1,730 tasks.


  • Specially invited experience officers mixed sweet and sugar-free cola, disrupted chewing gum shelves, and deliberately stood in the middle of the aisle or placed cardboard boxes to block the path, to test whether the robot could complete search, homing, and circling by relying on voice commands and visual recognition.


  • Tests have shown that humanoid robots can handle diverse tasks in real retail environments, and continuously adapt through shared team brains and feedback mechanisms. The successful transformation of four robots into full-service robots reflects the official direction of promoting real-life training in catering retail.

Share on Social Media

Leave a Reply

Your email address will not be published. Required fields are marked *