Imagine walking into a supermarket where robots quietly cruise the aisles at night, scanning every shelf to know exactly what's missing or in the wrong spot. Then robotic arms in the back room sort incoming goods and roll out to refill those empty spots — no human clerk needed. REFILLS built exactly that: a system of mobile robots and smart arms that can inspect shelves, recognize thousands of different products, and physically place them where they belong. They even tested the whole thing in a real supermarket.
Supermarkets lose revenue every time a shelf is empty and a customer walks away. Restocking shelves is one of the most labor-intensive, repetitive, and costly operations in retail — staff spend hours walking aisles, checking what's missing, going to the back room, sorting goods, and refilling displays. At the same time, inventory accuracy is poor because manual shelf checks are infrequent and error-prone.
The project built mobile robots that autonomously scan supermarket shelves and create semantic maps of store layouts, robotic arms for sorting goods in back rooms and refilling shelves, and sensors for grasping and collision avoidance. The system was demonstrated end-to-end in a real retail store, producing 25 deliverables including the final store demonstration and specialized grasping sensors.
If you are a supermarket chain dealing with high labor costs for shelf restocking and frequent out-of-stock situations — this project developed mobile robots that autonomously scan shelves, detect missing or misplaced products, and physically refill them. The system was demonstrated in a real retail store with 8 consortium partners across 5 countries contributing to the solution.
If you are a distribution company struggling with sorting accuracy and speed in back-room operations — this project built robotic arms that autonomously sort articles varying in shape, surface, fragility, stiffness and weight. The system uses knowledge bases about product properties to handle items correctly, reducing sorting errors and freeing up workers for higher-value tasks.
If you are a technology provider looking for proven robotic components to offer retail clients — this project created a modular, cost-efficient robotic platform with semantic mapping, product recognition, and manipulation capabilities. With 25 deliverables including sensors for grasping and collision avoidance, there are licensable components ready for integration into commercial retail automation products.
The total EU research investment was EUR 3,692,850 shared across 8 partners, which covered R&D, not unit costs. Based on available project data, the system was designed with a modular approach for cost-efficient robotic units, suggesting the intent to keep commercial pricing accessible. Exact per-unit pricing would need to be discussed with the consortium.
Yes — the project specifically addressed products varying in shape, surface, fragility, stiffness and weight. The system builds semantic environment maps of entire store layouts and uses knowledge bases to identify shelves and recognize missing or misplaced articles. A final demonstration was conducted in a real retail store environment.
The project produced 25 deliverables across 8 partners in 5 countries. IP is likely shared among consortium members including 5 industry partners. Licensing arrangements would need to be negotiated with the coordinator C.R.E.A.T.E. Consorzio in Italy or the relevant technology-holding partner.
The project specifically tackled manipulation of articles varying in shape, surface, fragility, stiffness, and weight. They developed dedicated sensors for grasping and collision avoidance, and use perception-guided reactive control to handle execution errors and uncertainty during manipulation.
The robots generate semantic environment maps for layout identification and store monitoring. Built-in reasoning allows robots to cope with changing task requirements and contexts, meaning the system can adapt when shelves are rearranged or new product categories are introduced.
The project ran from 2017 to 2020 and completed a final demonstration in a real retail store. The technology progressed through three integration and evaluation phases. While demonstrated in operational conditions, moving to full commercial deployment would likely require further engineering and customization for specific store formats.
The REFILLS consortium brings strong commercial credibility with 5 out of 8 partners coming from industry (62%), spanning 5 countries (Switzerland, Germany, Ireland, Israel, Italy). This is not a university-only research exercise — the heavy industry involvement signals that the technology was built with real deployment in mind. The coordinator is C.R.E.A.T.E. Consorzio, an Italian research consortium focused on energy, automation, and electromagnetic technologies. The mix of 2 universities and 1 research organization provides the scientific foundation, while the industry majority ensures practical, market-oriented development. The international spread across key European robotics hubs (Germany, Switzerland, Israel) adds depth in both robotics engineering and retail technology.
C.R.E.A.T.E. Consorzio di Ricerca per l'Energia l'Automazione e le Tecnologie dell'Elettromagnetismo, Italy — contact via SciTransfer for introduction
Want to explore licensing this retail robotics technology or connecting with the REFILLS team? SciTransfer can arrange a direct introduction and help evaluate fit for your operations.
