Project EASY

The challenge 

Modern production facilities generate huge amounts of data that need to be analyzed. A common method is to transfer data to the cloud and process it there using a lot of computing power. However, this is not efficient, as the transfer of large amounts of data requires a lot of energy. In addition, the principle results in a relatively high latency time. The aim of the EASY project is to create a cloud edge continuum that is suitable as a platform for the execution of AI-based analysis methods. The services should always be executed where it makes the most sense. The aim is to improve service quality as well as energy requirements. New algorithms are being adapted for this purpose, such as federated learning for distributed machine learning or a hierarchical diagnostic approach. Execution with less data is faster and more efficient.

Solution approach

EASY takes an innovative approach to creating an edge-cloud continuum. This enables the execution of AI-based analysis and control services in both local edge and cloud environments. To realize this, EASY is developing an open source runtime environment that enables the dynamic execution of compute services. This provides executable and semantically described services that can be orchestrated at all levels of the edge-cloud continuum and are self-configuring. These services can be accessed via a marketplace. In addition, EASY implements intelligent federated learning approaches to improve the monitoring of entire production lines. This enables cost-efficient use of local and global computing capacity. Analysis services are also being developed that can be run on low-performance edge nodes. EASY implements measurement and analysis procedures to balance resource consumption and data transfer in the edge cloud continuum. AI planning processes and digital twins are used for flexible planning and decentralized monitoring. These methods enable resilient control processes at edge level. The methods and systems developed by EASY are evaluated using demonstrator applications in the SmartFactoryKL and SmartFactoryOWL and demonstrated in several reference applications.

Profile