DIOR · Deep Intelligent Optical and Radio Communication Networks

DIOR: Deep Intelligent Optical and Radio Communication Networks

The DIOR project aims to tackle the complexity and performance limitations in optical and radio parts in telecommunications networks with the aid of machine learning. The key challenges include: 1). increasing demand for the information capacity of the existing telecommunications networks; 2). inefficient radio and optic components in the telecommunications infrastructures (e.g., fiber nonlinearities, such as self-phase modulation, cross-phase modulation, and four-wave mixing, equalization-enhanced phase noise, radio power amplifier nonlinearity of the radio transceivers, mobility-induced beam misalignment, performance degradation in long-haul optical connections); 3). inefficient network resource allocation; 4). lack of joint optimization of radio and optic resources across the access network, transportation network in the traditional signal processing and network management approaches. The DIOR project addresses critical challenges in next-generation telecommunication infrastructure that impact various aspects of modern society, including supporting high-speed and reliable Internet; enabling AI-driven 5G and Beyond Technologies; boosting digital Inclusion and connectivity in both urban and rural areas; improving energy efficiency and sustainability by optimizing the radio and optic components and operational costs and advancing AI applications in telecommunications. The DIOR project sets out to develop Machine Learning (ML) & AI-driven solutions for Communication Networks, focusing on the following objectives: 1. ML-based Digital Pre-Distortion (DPD): Develop ML-based methods to linearize base-station power amplifiers, improving radio signal quality and power efficiency. 2. Develop advanced AI-based methods to model optical fiber channels with power-dependent impairments, incorporating stochastic effects such as polarization mode dispersion (PMD) and amplifier spontaneous emission (ASE) noise, to enhance the accuracy of signal quality predictions. 3. Learn-Based Physical Radio Links: Replace traditional analytical models with deep learning approaches for radio signal processing. 4. Load Demand Awareness Modelling: Utilize high-resolution data to enhance network traffic prediction and optimization. 5. AI-Based Optical Channel Modelling: Create power-dependent optical fiber models that account for stochastic impairments. 6. Integration of optical wireless communications (OWC) with AI-driven optimization techniques and the increasing use of machine learning to enhance system performance in dynamic environments. 7. Topology Optimization for Optical Networks (OCNs): Design AI-driven algorithms for intelligent routing and resource allocation. 8. Joint Optimization of OCNs and radio access networks (RANs): Develop integrated AI mechanisms for hybrid optical and radio networks.

Data: CORDIS, © European Union