The KIGLIS project, funded under the German BMBF and EU schemes and running from 1 November 2020 to 31 October 2023, set out to design an optical and AI‑supported reference network for future digitalised cities. VPIphotonics GmbH led the effort, providing its well‑known software tools for modelling optical components, transmission systems and communication networks. The consortium included partners such as Atesio, Nokia and TelemaxX, with whom VPI shared prototypes, developed common software interfaces and carried out joint investigations. The project’s technical lead was Dr Shi Li, with Dr Elias Giacoumidis handling technical coordination and Harald Fleick managing administration.

The scientific work was organised into three tightly coupled work packages. First, a comprehensive requirement list for future optical access networks was produced, based on an analysis of 30 prospective applications. Four representative use cases – Remote Assistance, Video Surveillance, Cooperative Adaptive Cruise Control and Road Safety Warnings – were selected to anchor a reference architecture centred on a fibre‑based access network that must support all existing wireless technologies. The network is required to deliver at least 100 Gb/s over a maximum reach of 20 km, with a latency below 1 ms and a reliability of 99.999 %. Using VPI’s link‑configuration tool VPIlinkConfigurator, the consortium built a reference topology for the city of Karlsruhe, which was subsequently published in a peer‑reviewed article.

Second, the project tackled the technological challenges of high‑bit‑rate passive optical networks. VPI’s expertise in transient modelling, non‑linear effects and fast estimation of transmission impairments was leveraged to optimise network configurations and to develop an emulator for non‑linear propagation. The team also produced efficient digital‑signal‑processing algorithms for probabilistically shaped signals, and integrated these into a disaggregated network planning framework that can be managed via SDN controllers.

Third, a dedicated test environment was created to generate realistic training data, train AI models and evaluate their performance. AI methods for signal denoising were developed and validated against the simulated network scenarios. Parallel work produced algorithms for generating monitoring data streams that can be fed back into the optimisation loop. The resulting AI‑enhanced models were integrated into a demonstrator that showcases real‑time network optimisation and fault detection, illustrating the practical benefits of the approach.

Throughout the project, VPI maintained open interfaces to popular simulation platforms such as MATLAB, Python and VBA, enabling seamless co‑simulation with partner tools. The consortium held regular progress meetings, with the kickoff held online on 1 December 2020 and the final meeting on 12–13 October 2023. The project’s outcomes include a validated reference network design, a suite of AI‑enabled optimisation algorithms, a demonstrator platform, and several publications that disseminate the findings to the wider research community. The collaboration model, combining VPI’s modelling capabilities with partner expertise in network deployment and management, demonstrates a scalable pathway for deploying AI‑driven optical access networks in future smart cities.