D20: Use-case Deployment and Demonstrations v2.0

Deliverable D20 – Use-case Deployment and Demonstrations v2.0 documents the work
carried out under Work Package 5, Tasks 5.3 (Lab Experimentation) and 5.4
(Demonstrations), across the full thirty-eight-month duration of the CODECO project. It
presents a comprehensive account of how the CODECO framework was deployed, validated,
and demonstrated across six use-cases spanning smart city infrastructure, urban mobility,
media delivery, energy management, industrial automation, and smart buildings.
The six pilots — led respectively by the University of Göttingen, i2CAT, Telefónica, the
Universidad Politécnica de Madrid, fortiss GmbH, and Almende — each addressed a distinct
real-world challenge requiring dynamic orchestration of computational and networking
resources across the Edge-Cloud-IoT continuum.

The technical results confirm that CODECO delivers measurable value in each deployment
context. In P1, dynamic workload offloading improved LiDAR analytics throughput by 32%
under constrained edge conditions, with CODECO responding effectively to pod failures,
channel degradation, and latency increases. In P2, CODECO's reinforcement learning-driven
orchestration reduced Age of Information by 13.5% and positional tracking Error Distance by
23.6% relative to a centralised baseline at peak vehicular load, while maintaining both
deployments within the 3.5-metre lane-width safety threshold. In P3, CODECO's ALTO-based
network awareness enabled more efficient placement of media delivery caches and faster
adaptation to network disturbances than a static Kubernetes deployment. In P4, the
decentralised energy management system achieved the targeted 5–10% improvements in
energy efficiency and carbon footprint reduction across three UPM campuses, with fault
tolerance validated at the 2-minute recovery target. In P5, CODECO demonstrated reduced
CPU consumption, more stable communication patterns, and successful stateful SLAM
migration on physical AMR hardware, with all operational KPIs addressed or partially
evidenced. In P6, automated deployment and event-driven redeployment — triggered by
topology changes and occupancy-derived BLE statistics — demonstrated a class of adaptive
smart building management that would be practically unachievable through manual operation.
Cross-pilot analysis of CODECO component engagement reveals that ACM and NetMA were
the universal foundation of the framework across all six pilots, while SWM provided workload
migration capabilities in five, PDLC provided learning-based orchestration intelligence in four,
and MDM provided data-flow observability in three. The variation in component engagement
reflects the genuine diversity of the pilots rather than incomplete integration: each partner
engaged the subset of CODECO's capabilities that matched their application's operational
requirements, and in several cases — P4's energy-domain integration, P6's Virtual Kubelet
layer — developed custom bridging components that extended CODECO's reach into
previously unsupported device classes and data modalities.