Objectives
The vision of the smART radIo acceSs with inTegration of user devices (ARTIST) project is a Beyond 5G (B5G) scenario where the User Equipment (UE) is exploited not only to satisfy the specific needs of the UE owner but also to augment the Radio Access Network (RAN) infrastructure as a distributed capability and as a source of network intelligence. In other words, we envision the idea of UEs taking a more active role in network service provisioning as one of the key pillars to ground future mobile network evolution, to the extent that the very notion of a cell will have to be rethought as UEs will be able to actively complement the RAN infrastructure. This constitutes a radical paradigm shift from UEs operating only as Network Service Consumers towards embracing UEs also as Network Service Providers. That is, utilizing UEs as an extended computing, storage and networking element of the B5G RAN infrastructure as well as a central element for the realization of end-to-end connected network intelligence.
Furthermore, in views of the widespread exploitation of Artificial Intelligence (AI) and Machine Learning (ML) in the 5G and beyond era, the integration of the UEs as part of an end-to-end connected network intelligence approach is a cornerstone for the materialization of true context awareness and fully distributed data analytics solutions that do not need to expose to the network the huge amount of raw data collected in the UEs. The ARTIST concept with its intrinsic connected intelligence will contribute to solving and/or mitigating a wide range of problematics that are inherent to wireless networks. Remarkably, the realization of ARTIST paves the way for enhanced service quality (e.g., enhanced coverage, better mobility management, user behavior and demand prediction), higher radio resource utilization efficiency (e.g., more efficient radio resource allocation, congestion control, better beam management), improved system performance (e.g., on device inference reduces network data traffic for more efficient mobility and spectrum utilization, better link adaptation can be attained through position[1]aware interference prediction) and simplified RAN deployment and operation (e.g., more capable Self Organizing Networks for e.g., mmWave network densification, reduced energy consumption).
Objectives
Objective #1: To define a comprehensive set of ARTIST use cases in representative B5G deployment scenarios and extract the system requirements.
Sub-objective #1.1:
To identify, develop and specify the use cases where the ARTIST concept finds applicability in B5G scenarios. Key elements to be specified for each use case are the actors involved, their roles and interactions, the environmental and contextual characteristics (e.g. indoor area, urban
commercial area, event in a stadium, etc.), and the expected behavior, describing the pre-/post-conditions and flow of actions.
Sub-objective #1.2:
To extract the system requirements derived from the selected use cases. It embraces deriving the technical and operational requirements to be fulfilled by the ARTIST architecture and its technical solutions as well as specifying performance requirements with measurable KPIs at global system level (e.g. percentage of coverage area).
Objective #2: To design the functional architecture, information models and procedures for the ARTIST materialization.
The objective is to design and specify the functional architecture of the ARTIST solution where the RAN is augmented through the capabilities of different types of UEs acting as network service providers. It will take the O-RAN architecture as a reference and will enhance it with the inclusion of new functionalities required at the UE, RAN and O&M to efficiently operate and manage the augmented RAN. The design will involve the identification and detailed description of the individual components of the ARTIST architecture, the interfaces connecting them and the interactions between components, which will be specified by means of procedures. Moreover, the objective embraces the specification of the information models that will be used for managing the different components of the architecture.
Objective #3:To conceive, develop and assess the distributed connected intelligence framework across network and terminals to support network operation.
Sub-objective #3.1:
To establish a solid common background in ARTIST by identifying the most adequate candidates to be used in the design of the ARTIST O&M and RRM algorithmic solutions. This will embrace the identification of techniques (e.g., prediction, classification, clustering, reinforcement learning) and tools (e.g., deep learning, neural networks, Q-learning, support vector machines, etc.) that best fit ARTIST’s novel AI-based functionalities.
Sub-objective #3.2:
To generate the datasets to be used when training the ML models of the distributed learning framework. For this purpose, the project intends to use measurements extracted from real mobile networks, expanded with data augmentation techniques and with synthetic data. More details about the dataset generation are given in Section 1.6.
Sub-objective #3.3:
To develop an AI/ML pipeline by combining the specific functionalities of data collection, data pre-processing, model training and inference that will support the decision making for RRM and O&M systems. This will consider the new ARTIST concept of bringing the UE in the process to significantly go beyond existing AI/ML pipelines such as the one considered by O-RAN, through the exploitation of federated learning for training the models.
Sub-objective #3.4:
To assess the performance of distributed connected intelligence framework and federated learning mechanisms by measuring specific KPIs that will guide the evaluations (e.g. prediction accuracy).
Objective #4: To conceive, develop and assess the O&M mechanisms and algorithms for ARTIST network operation.
Sub-objective #4.1:
To design and assess algorithmic solutions for an optimal and automated management of the ARTIST architecture. This involves specifying, characterizing, extracting and organizing the context knowledge about the UEs and their local environment at O&M level. Based on this context knowledge, new decision making solutions will be developed for selecting, activating/deactivating and appropriately configuring the UEs, among those registered in the system as potential network service providers. O&M systems will also need to be extended with new service management functionalities to keep track of the features and capabilities that the UEs can contribute to the RAN including their usability restrictions.
Sub-objective #4.2:
To specify the new UE-to-network management interfaces for configuring the UEs with the proper operational settings and for providing ML models to the UEs for ML inference. Moreover, this sub-objective also embraces to develop the operation of the non-Real Time (non-RT) RAN Intelligent Controller (RIC) of the O-RAN architecture to guide and enforce the required behaviours in the near-Real Time (RT) RIC with regard to the operation of the RRM algorithms, including policy and ML model management and delivery of enrichment information.
Sub-objective #4.3:
To carry out the performance evaluation of the O&M solutions in selected use cases, allowing to demonstrate the fulfilment of the user and system requirements as well as to assess specific KPIs (e.g. activation/deactivation/reconfiguration times).
Objective #5: To conceive, develop and assess the control plane mechanisms and algorithms for ARTIST network operation.
Sub-objective #5.1:
To conceive and develop the RRM algorithms to control the operation of the RAN augmented with the UE capabilities. Within the O-RAN reference framework, the so-called near-Real Time RAN Intelligent Controller (RIC) is established as the central component to handle the RRM logic that guides the operation of RAN control plane. On this basis, ARTIST will conceive and develop the necessary RRM enhancements and extensions to efficiently control the operation of the RAN.
Sub-objective #5.2:
To carry out the performance evaluation of the RRM algorithms for selected use cases, conducted to assess specific KPIs (e.g. dropping probability for handovers).
Sub-objective #5.3:
To establish the holistic operation of the different near-RT RIC RRM solutions in relevant use cases. Only by developing such holistic perspective can the interactions among the various RRM functionalities be identified and synergies consistently exploited.
Objective #6: To carry out experimental proof-of-concept (PoC) of key ARTIST functionalities.
Leveraging on the designed architecture and the developed algorithmic solutions for near-RT RRM, O&M and distributed connected intelligence, the project will experimentally demonstrate selected ARTIST functionalities.