Towards zeRo toUch nEtwork and services for beyond 5G

Impact

This project has created, developed, and tested a set of new procedures, algorithms, mathematical tools, and technologies that have contributed to advancing the automation of 5G and 5G+ networks and network services. A subset of these developments has been applied to verticals such as industry 4.0, automotive, surveillance, leisure, and healthcare, among others. The project has holistically addressed automation across all network components, including radio access network (RAN), optical and satellite access network, core network, network infrastructure, cloud and edge computing, and network services. Developments have been carried out in accordance with recommendations, standards, and open frameworks such as 3GPP, ETSI, and IEEE. Some of these organizations have contributed with participation and proposals to working groups. 

The project's advancements have primarily focused on the following areas:

  • Network slicing: Using Network Calculus algorithms and queueing theory, mechanisms and schedulers have been created for dynamic management of slices in RAN and end-to-end, covering access network, core network, and infrastructure. A TSN-5G-TSN gateway has also been modeled and developed to connect TSN islands end-to-end, ensuring quality of service. These results have been applied to verticals such as industry 4.0.
  • Multi-technology, multi-domain service orchestration: Vertical handover between LTE/5G/WIFI technologies has been designed using multi-path TCP governed by artificial intelligence mechanisms. For UAV (Unmanned Aerial Vehicle) environments, a service orchestration framework based on SDN, NFV-MANO multi-VIM, and lightweight container architecture for embedded systems for critical missions with energy consumption constraints have been developed and field-validated.
  • Intelligent management and orchestration in RAN, optical, and satellite access networks: DBA algorithms have been designed for Long-Range EPON for ranges of 100 km. Genetic algorithms have also been applied for dynamic management of the control plane and placement of controllers onboard low Earth orbit (LEO) satellites, extending SDN technology. In IoT networks, automatic enrollment algorithms for digital certificates have been designed. Physically, configurations with P4 have been deployed and validated. Probability blocking models have been proposed for the (O)RAN.
  • Zero Touch: Deep reinforcement learning and multi-agent mechanisms have been applied for dynamic placement of MEC and service requests, resources modeled by VNF, and composed of edge computing and storage capacity. These solutions have been simulated, emulated (creating private 5G, EPON, and TSN networks), and applied to field verticals.

Main project results that represent an advance in knowledge:

  • Design of a service Management and Orchestration platform (MANO) based on cloud native technologies (Kubernetes) and UAVs for environments with resource limitations.

  • Definition of mechanisms for the automatic selection of communication links in hyperconnected 5G UAV applications (e.g. with 5G/LTE capacity, WiFi, line-of-sight radio links, etc.)

  • Requirement analysis and design of enabling solutions for private 5G networks, as well as evaluation of their performance for different configuration options.

  • Design and evaluate an end-to-end 5G slicing architecture to support industrial processes with different bandwidth and latency requirements efficiently.

  • Definition of a formal model for configuring asynchronous TSN transport networks that minimize the probability of flow rejection and guarantee limited delay with end-to-end reliability.

  • Development of an algorithm assisted by artificial intelligence for the configuration of transport networks based on TSN that includes the distribution of end-to-end requirements between the different domains of the network.

  • Development of an SDN controller for TSN networks following the IEEE 802.1Qcc standard.

  • Design of an architecture to support delay-sensitive communications (TSN) in applications implemented under the NFV framework using hypervisor-based or container-based virtualization.

  • Development of a radio resource allocation and optimization algorithm assisted by artificial intelligence in a multi-technology radio access network.

  • Evaluation of the benefits of multi-technology radio access networks (5G and WiFi) in an industrial scenario, considering analytical models for packet loss probability and delay for URLLC and eMBB services.

  • Creation of a 5G prototype network using a real base station and backbone network, as well as different mobile devices.

  • Design of an IoT network architecture for deploying massive sensors, based on LoRaWAN and SDN, that allows optimizing traffic and delay in the backbone network.

  • Analysis and evaluation of parameters for latency reduction in 5G networks, considering the space between subcarriers and the frame structure, achieving 4-5 ms latencies with real equipment (round trip time).

  • Dynamic creation of slices that adapt to the variability of network, storage and computing resources based on a distributed and multi-agent architecture using deep reinforcement learning algorithms (DQN, Rainbow, etc.), applying long-term memory neural networks (LSTM) and Random Forest that are capable of predicting the performance of each slice and adapting the amount of resources needed accordingly.

  • Design of SDN-based solutions to solve the problem of dynamic scaling and load balancing in transparent VNFs, guaranteeing bidirectional flow affinity and avoiding the appearance of loops in the network.

  • Design of a scheduler to deploy virtualized network functions or various tasks, taking into account the energy consumption of the different nodes that make up a Kubernetes cluster.

  • Development of a WIM (Wide-area Infrastructure Manager) with slice-level access control mechanisms for reserving paths with guaranteed bandwidth and redundancy at level 2.

  • Integration within the NFV orchestration framework of heterogeneous resources, different from traditional NFVI, such as LTE and 5G base stations, network hardware accelerators, and TSN network equipment.

  • Design and evaluation of a recovery mechanism in case of link failures in intra-CPD networks or operator networks that connect remote CPDs.

  • Design of a mechanism to improve the efficiency of IoT communications through 5G networks, reducing the volume of traffic generated and allowing faster detection of alarm situations.

  • Design and validation of a system for online validation of digital certificates aimed at improving the efficiency of access of non-3GPP devices to the 5G Core through the N3IWF module.

  • Design of an architecture based on TLS for the security of HTTP/2 communications of the 5G Core based on Service-Based Architecture (SBA).

Research transfer to industry and the society:

  • The project has collaborated with multiple companies and technological centers, including Gestamp, ATOS, ZTE, Euskaltel, Embeblue, Teltronic, SoC-e, Ikerlan, Vicomtech, Tecnalia, i2Cat, CTTC, municipalities of Gavá, Castelldefels and Barcelona, Aguas de Barcelona, CETAQUA, Tinkerers Lab, Fundación ONCE, Mobile World Capital, GSM, Generalidad de Cataluña, etc.

  • Creation of a new spin-off company (Alteraid) by members of the research group of the UPC.

Master's and doctoral training:

  • Dissemination of the project activities and results in the UPV/EHU’s “Cybersecurity 4.0” master’s degree.

  • Creation of a Joint master’s in Communications and Data Science (CoDaS) Erasmus program in the UPC.

  • Creation of the international master’s degree international Artificial Intelligence for Connected Industries (AI4CI), where UPC takes part.