Hanshow Technology and X-Era Lab are bringing wheeled inspection robots into real supermarkets, using electronic shelf labels and digital twins to build a “physical-world coordinate system” that reduces edge computing pressure and enables inventory counting. Compared with highly structured factories and fully unstructured homes, the retail store—falling between the two—is becoming a touchstone for embodied intelligence to validate a commercial loop.

Inside an operating supermarket, a wheeled robot moves slowly along the shelves. It must avoid customers and shopping carts, verify prices against electronic shelf labels, and adjust its camera angle to see inventory hidden behind front-row products. During replenishment, a robotic arm must reach into shelf layers only 30–40 centimeters high, pick up soft, puffed-food bags, and place them neatly back on the shelf.

ehind these tasks is a collaboration between Hanshow Technology and X-Era Lab (Tuoyuan Intelligence). Hanshow serves more than 550 retail customers globally, with products covering over 80 countries and nearly 70,000 stores. Its electronic shelf label network itself forms a readable layer of product and slot information. X-Era Lab focuses on self-developed world action models, using 4D spatiotemporal representations to help robots understand occlusion, lighting changes, and product states.

Retail’s pain points are concrete. Full-store inventory counts at large stores can only be scheduled after closing. During business hours, customers and products are constantly