The EXPLOR project provided comprehensive analysis of methodologies in cellular and optical network modeling, resource management, and ML optimization, identifying key procedures for functional splitting, scheduling in C-RAN, SG and OFH modeling, and ML-driven resource management. The project then went beyond state-of-the-art methodologies across these core areas. Namely, EXPLOR pursued novel SG frameworks based on Graph Convolutional Networks (GCN) and Graph Attention Networks (GAT), as well as independent Poisson Point Processes across multiple spectrum bands, providing a basis for analysing coverage, rate, and delay performance. Artificial Neural Network (ANN) and Deep Neural Network based equalization methods was developed for impairment mitigation in radio-over-fiber (RoF)-based OFH systems. The project produced a novel resource management algorithm using Software-Defined Networking (SDN) and a Simulated Annealing heuristic to reduce Grade-of-Service and power consumption within the fronthaul network. An accurate analytical model to calculate Admission Ratios in converged optical-wireless networks was developed, optimizing resource allocation across network slices. Moreover, developed PHY OFH modeling for RoF systems, featuring advanced MMW generation and detection techniques, facilitates the complete simulation of end-to-end optical-wireless links. A configurable, SDN-based transceiver (SDT) architecture was also developed, allowing for coexistence of optical transmitter and subcarrier pools for reduced redundancy. Power consumption optimization in OFH was enabled through Diagnostic Algorithm (DA) and Optical Virtualization Algorithm (OVA), allocating computational resources for specific accuracy requirements with the flexibility to adapt dynamically based on compatibility and C-RAN demands. Rigorous testing validated the OVA as a critical component for optimizing OFH resources in conjunction with the C-RAN orchestrator, showing success in optimizing system performance. The final stages involved framework integration. C-RAN Virtual Resource Management (VRM) framework was integrated with OVA, enabling dynamic configuration of the MMW RoF SDN-based OFH architecture. This setup maximized admission rates and minimized power consumption and latency by leveraging inter-module feedback mechanisms. Link parameters were dynamically adjusted based on SG and OFH module limitations, optimizing transmitter drive power while maintaining Quality of Service. A feedback mechanism linking SG and OFH modules was established, with SG data informing OFH transmission dynamic range requirements. The VRM-OVA feedback mechanism supports OFH reconfiguration, while SG-OVA feedback serves to confirm the support for cell distances. The network dimensioning process established through this integration enables the use of parameters for any converged optical-wireless setup. An ANN-based equalization technique at the receiver end of the MMW RoF link supports training and testing, while a Small Cell Deployment module using GCN and GAT validates SG approaches. Finally, comprehensive performance evaluations comparing resource management algorithms with alternatives, alongside experimental validation of PHY/SG models, demonstrated power savings achieved through OVA-driven OFH reconfiguration.
Efforts to disseminate these outcomes leveraged a combination of digital platforms, conference contributions and talks, journal publications, and the organization of scientific workshop, seminar and open-day events. Feedback from these event sessions informed further development of the EXPLOR simulation platform. To enhance visibility in the academic community, EXPLOR project outcomes were shared through a book chapter, eight peer-reviewed journal articles, and ten conference papers. These publications highlighted enabling methodologies across the project's four core research areas, alongside novel approaches to resource management, ML-driven optimization, and advanced RoF solutions.
