Six AGIBOT G2 Robots Complete 1,730 Tasks in 24-Hour Shanghai Convenience Store Trial

Real-World Retail Trial

According to AGIBOT, six G2 wheeled humanoid robots were put to work in an active convenience store in Shanghai's Xuhui District for a continuous 24-hour period. The trial covered various shifts, including the lunchtime rush, evening peak hours, overnight operations, and the morning breakfast rush. During this time, the robots executed 1,730 tasks, ranging from customer service interactions and product pickups to coffee preparation, shelf inspections, and store cleaning.

Testing Precision and Adaptability

Operating in a live store presented everyday challenges that controlled laboratories cannot fully replicate. The G2 robots were subjected to practical evaluations, such as distinguishing between regular and sugar-free Coca-Cola, returning misplaced chewing gum to correct positions, and navigating around obstacles like people and boxes blocking the aisles. AGIBOT notes that four out of the six robots passed the final evaluation, which assessed duty time, task completion, customer feedback, and response times.

Genie Master Framework

To coordinate the fleet, AGIBOT utilized its Genie Master embodied AI agent framework. The system links task planning, execution, and feedback through an Embodied Execution Protocol, allowing the robots to learn from successful and unsuccessful attempts. AGIBOT plans to continue testing its robots in additional real-world environments to improve reliability and task execution.