The latest deliverable of the 6GARROW project introduces the initial system architecture of the laboratory demonstrations developed within the fifth work package of the project. The Deliverable 5.1 defines four representative demonstrations for validating AI-native technologies across the device, RAN, and CN domains. The deliverable further establishes the relationship between the demonstrations and the 6GARROW functional architecture and exemplifies, based on one demonstrator, a unifying demonstration environment for collaborative edge inference. The architectural descriptions provide the foundation for the subsequent implementation, integration, and validation of the proposed AI/ML technologies.
Deliverable D5.1 contributes to Objective 4 of the 6GARROW project: Develop and Validate Experimental Testbeds and Proof of Concept Platforms to showcase the key 6GARROW innovations. It presents the initial architectural design of the four laboratory demonstrations of Task 5.1.
The demonstrations assess the proposed AI/ML concepts for 6G networks in a controlled environment. They comprise semantic-aware device–edge co-inference, cross-domain AI coordination across the radio access and core domains, AI/ML-enabled physical-layer processing, and CSI compression. For each demonstration, the objectives, logical architecture, physical/software architecture, interfaces, capabilities and validation targets are specified, establishing a consistent framework for implementation and experimental validation.
Each design is grounded in the AI-native system architecture defined in D2.2 and revised in D2.3, and in the AI/ML methods under development in WP3 and WP4. The deliverable also introduces a unifying demonstration environment for edge intelligence applications based on Demonstration 1, mapped to the layers of the 6GARROW functional architecture. Experimental results obtained on these platforms will be reported in D5.3.