Knowledge4Retail: Artificial intelligence for retail

© Fraunhofer IPA, Rainer Bez
“Service robots can be used to load shelves autonomously”

In short

“Knowledge4Retail” is a strategic marketing tool that links the worlds of analog and digital commerce and makes digital solutions available for individual customer service: The consortium with representatives from business, trade and research develops technologies and an open platform for AI-based applications for stationary retail commerce.

In detail

The aim of the project is to establish a new generation of information systems for retail and its supply chains as a digital innovation platform and ecosystem. The planned platform will contain a powerful database of semantic digital twins of products and entire retail stores, thus laying the foundations for AI and robotic applications from different suppliers.

In the course of the project, four use cases will be developed, implemented and tested as examples to demonstrate the capabilities and economic benefits of the platform for the people involved. The following scenarios are currently being planned:

  • “Intelligent intralogistics”
    For example, AI can be used to sort incoming goods from various sources based on their storage location and requirements in each individual store
  • “Optimal store layout” 
    In order to respond more flexibly to customer needs, which also differ from one region to the next, artificial intelligence is used to calculate the optimal range and placement of products for the stores. 
  • “Service robots”
    Robots can load shelves autonomously and thus relieve the workload on store employees. This case demonstrates the seamless integration of robotics applications into the planned platform. 
  • “Smart fridge”
    To demonstrate the transferability of the platform to similar domains, a pilot application will demonstrate how the platform can interact with a cashless, intelligent refrigerator operating according to the “Go” concept.

The focus of Fraunhofer IPA's development work in the project is on the following topics: 

  • Advancement of “ScanStation” developed by Kaptura and Fraunhofer IPA for digitizing products in order to generate comprehensive semantic digital twins of objects. The multimodal fusion of visual information enables robust high-resolution 3D models of a typically wide range of products to be generated. Elements of products relevant to semantics, such as texts or logos, are also extracted. 
  • Standardization and interfaces for an open platform
  • Design and implementation of the pilot applications in the fields of machine vision and machine learning in order to recognize objects, in particular retail items.

IPA is also involved in the design of the AI-based software platform.