ACE ROBOTICS has partnered with KUAIKEDA to establish an Embodied Retail Skill Training Base in Deqing, Zhejiang. The initiative marks the first deployment of the company's "dual-brain collaborative architecture" and Embodied AI RSI (Recursive Self-Improvement) in an instant retail setting. Leveraging a "General Model + Kaiwu World Model" dual-brain architecture alongside Embodied AI RSI, the partners are focusing on high-frequency tasks such as food preparation, product packing, shelf organization, and fine manipulation. Through multi-robot collaborative learning and real-machine feedback, they are continuously generating, optimizing, and sharing transferable retail skills. Specifically, the collaboration targets three core areas: food preparation and retrieval, product packing and store organization, and the fine manipulation of small items. The goal is to advance robot training and capability verification in real-world retail environments, exploring viable paths for embodied intelligence to integrate into daily store operations. Image Credit: ACE ROBOTICS In the food preparation and retrieval sector, robots are training on tasks ranging from picking macarons and serving oden to assembling sandwiches and heating buns in microwaves. The training covers the entire chain—retrieval, processing, and plating. Handling fragile, soft, or skewered ingredients requires fine grasping and sequential assembly skills. They are also mastering complex interactions like preventing spills, operating microwave doors, and setting timers, moving from single-step tasks to full meal prep for convenience store hot food counters. For packing and store organization, the focus is on packing goods, safety checks, handling hanging items, and organizing beverages and snacks. These tasks cover critical links in inventory flow and shelf management. Robots are learning to navigate challenges like mixed bottle and can layouts, soft packaging, and dense shelving. They are refining their ability to identify, classify, and restock items, while mastering continuous workflows such as mixed packing, sealing, labeling, checking expiration dates, and restocking—bringing these capabilities directly to real-world store floors. In the realm of stationery and small items, robots are tackling tasks like organizing transparent stationery, opening eyeglass cases, and handling coffee and tea bags. They are also stamping documents and organizing shopping baskets. The aim is to enhance recognition of tiny objects and refine grasping and tool usage. Dealing with transparent, reflective, or soft items in tight spaces requires high precision and controlled pressure. The robots are practicing combined tasks—sorting, stamping, and shelving—to eventually serve in stationery shops, coffee corners, and checkout counters. Meanwhile, the base is using dozens of robots learning in sync to explore how atomic actions and execution experiences can be shared across units. This approach allows effective skills learned by one robot to be transferred to the fleet. Under the roadmap, ACE ROBOTICS and KUAIKEDA plan to deploy thousands of retail robots to the frontlines of retail by 2027.