Objective I-To design a novel RAN that supports dynamic scalability, as well as high security and privacy:
The NANCY consortium performed a thorough analysis based on O-RAN standards to pinpoint areas where the project could introduce innovations. The overall architecture of NANCY was outlined, including the placement of each component, the requirements for in-lab testbed and demonstrators, as well as their relevance to the usage scenarios. Furthermore, a theoretical framework based on Markov-chain modelling for B-RAN was created, while various potential security risks faced by B-RAN were identified. Also, the dynamic resources of B5G networks and the capabilities of the end nodes were theoretically modelled with an emphasis on a high-mobility use case. Additionally, significant advancements have been made in the creation of quantum safety mechanisms. These developments encompass simulations, experiments, and demonstrations for Quantum Key Distribution (QKD) with a specific focus on generating and exchanging quantum keys. Moreover, Post-Quantum Cryptography (PQC) has been utilized to ensure secure communication and enable digital signatures. Finally, the theoretical foundation for smart pricing has been set up, based on reinforcement learning and reverse auction theory. The initial version of the reinforcement learning environment has been built and experiments have been carried out to design agents that align with the NANCY objectives.
Objective II-To transform networks beyond 5G to intelligent platforms integrating ultra-reliable connectivity and high-energy efficiency:
The AI-based B-RAN orchestration techniques were developed for deploying services and adjusting resources, as well as effectively overseeing the NANCY service and resource orchestrations. Moreover, a new method for selecting ML models was introduced, utilizing a a DRL-based approach to choose models based on the user's location and environment. To realize the NANCY AI virtualizer, a new method for communication between different slices was shown to greatly reduce inefficiencies and conflicts. In this direction, a novel approach to monitoring unused CPU time in Linux was developed, allowing to identify over provisioned virtual environments. Furthermore, a DRL agent was designed, which efficiently allocates flows for integrated access backhaul (IAB) networks, while particular tools were designed to analyze workloads in real-time. Additionally, an extensive exploration into state-of-the-art semantic communications and networking architectures paved the way for proposing an approach to integrating semantic communications into 6G and future networks. To this end, knowledge extraction techniques were investigated for image-based communications and the establishment of a semantic communications framework enhanced by convolutional neural networks. Also, a goal-oriented approach was developed focusing on vehicle-to-everything (V2X) applications. Finally, cutting-edge explainability algorithms for B5G/6G networks were investigated. In more detail, the SHAP and LIME algorithms were developed, modified, and tested using publicly accessible data sets.
Objective III-To provide almost-zero latency and high-computational capabilities at the edge:
A detailed analysis of cutting-edge self-healing and self-recovery techniques in the domain of B5G/6G networks was conducted. The consortium thoroughly evaluated big data platforms to determine their suitability within the NANCY architecture. Diverse algorithmic solutions were investigated aimed at addressing challenges related to self-healing and self-recovery, taking into account their practicality and integration within the NANCY architecture and test environments. In addition, a detailed workflow for offloading and data caching was created specifically for the NANCY architecture. Also, the NANCY service level agreement (SLA) was defined and a method for overseeing the SLA from start to finish was specified. Finally, the consortium developed methods for handling communication and processing resources at the edge and developed decision-making tools to select the best approach for offloading