Periodic Reporting for period 1 - IPOSEE (Intelligent and Proactive Optimisation for Service-centric Wireless Networks)
Okres sprawozdawczy: 2023-03-01 do 2025-02-28
Fast and reliable wireless connectivity is essential for businesses and industries in a broad range of sectors, from education and finance to healthcare, transport, utilities, logistics, mining, and manufacturing. As the demand of wireless coverage and services continues to rise, current wireless networks will fact new challenges, in terms of cost and resource efficiency, and reliability. The IPOSEE project, funded within the Horizon Europe Marie Skłodowska-Curie Actions programme, targets new concepts and solutions for automating and optimisation of service-oriented wireless networking, with delivering services as the focus, rather than just providing basic connectivity.
The objectives of research activities of the IPOSEE project are:
1) To devise and implement effective machine learning algorithms to extract, analyse, and interpret spatial-temporal the spatial-temporal patterns in mobile network data traffic. These algorithms are able to quantify uncertainty in service-demand prediction.
2) To identify and define representative subset(s) of services and their corresponding demand patterns for radio access network (RAN) deployment scenarios, by studying and analysing the correlation among the spatial-temporal patterns of service demand, to simplify the training for service demand forecasting.
3) To gain fundamental understanding of the achievable performance gains/limits, in particular capacity, via network densification (i.e. adding more network elements) and optimization of radio propagation environment optimisation with AI-aided traffic forecast, and to benchmark the performance for demanding scenarios such as requirement of low latency.
4) To develop AI-enabled RAN optimisation applications in conjunction with a probabilistic optimisation engine, based on AI predictions and RAN optimisation to cost-effectively meet service requirements. These optimization applications will be deployed as containerized microservices specifically designed for operation within the RAN Intelligent Controller (RIC) of the Open RAN (O-RAN) architecture.
The objectives of research activities of the IPOSEE project are:
1) To devise and implement effective machine learning algorithms to extract, analyse, and interpret spatial-temporal the spatial-temporal patterns in mobile network data traffic. These algorithms are able to quantify uncertainty in service-demand prediction.
2) To identify and define representative subset(s) of services and their corresponding demand patterns for radio access network (RAN) deployment scenarios, by studying and analysing the correlation among the spatial-temporal patterns of service demand, to simplify the training for service demand forecasting.
3) To gain fundamental understanding of the achievable performance gains/limits, in particular capacity, via network densification (i.e. adding more network elements) and optimization of radio propagation environment optimisation with AI-aided traffic forecast, and to benchmark the performance for demanding scenarios such as requirement of low latency.
4) To develop AI-enabled RAN optimisation applications in conjunction with a probabilistic optimisation engine, based on AI predictions and RAN optimisation to cost-effectively meet service requirements. These optimization applications will be deployed as containerized microservices specifically designed for operation within the RAN Intelligent Controller (RIC) of the Open RAN (O-RAN) architecture.
The following activities addressing the project’s technical and scientific objectives have been performed with a number of achievements.
1) AI-Driven Traffic Analysis/Forecasting: The project has developed new schemes that take advantage of artificial intelligence (AI), in particular advanced unsupervised learning algorithms to separate the datasets into clusters and identify the spatial-temporal traffic patterns, and thereby enable to forecast hotspots and service classes. The research has also addressed the use of forecast in performing optimisation of RAN deployment, configuration, and resource allocation.
2) Mathematical Modelling for Network Densification: For network densification and optimisation, the research efforts have followed a mathematical modelling approach that is designed to characterize the effect of the physical radio propagation environment, more specifically reconfigurable intelligent surface (RIS), and to study performance improvement with the deployment of RIS. The modelling approach has low computational complexity, to enable to efficient RIS deployment and configuration.
3) Analysis and demonstration of Use Cases and Applications: Based on the results above, the project has conducted in-depth analysis on a selection of representative use cases and applications. These analyses serve as concrete validation and demonstrations, showcasing the real-world effectiveness and applicability of the techniques, with insights into network behaviour under different service and traffic conditions.
1) AI-Driven Traffic Analysis/Forecasting: The project has developed new schemes that take advantage of artificial intelligence (AI), in particular advanced unsupervised learning algorithms to separate the datasets into clusters and identify the spatial-temporal traffic patterns, and thereby enable to forecast hotspots and service classes. The research has also addressed the use of forecast in performing optimisation of RAN deployment, configuration, and resource allocation.
2) Mathematical Modelling for Network Densification: For network densification and optimisation, the research efforts have followed a mathematical modelling approach that is designed to characterize the effect of the physical radio propagation environment, more specifically reconfigurable intelligent surface (RIS), and to study performance improvement with the deployment of RIS. The modelling approach has low computational complexity, to enable to efficient RIS deployment and configuration.
3) Analysis and demonstration of Use Cases and Applications: Based on the results above, the project has conducted in-depth analysis on a selection of representative use cases and applications. These analyses serve as concrete validation and demonstrations, showcasing the real-world effectiveness and applicability of the techniques, with insights into network behaviour under different service and traffic conditions.
By the progress made thus fat within the project, the research delivers tangible potential to transform RAN planning, deployment, and operation, for systems beyond the current 5th generation networks. More specifically, the results of AI-driven spatial-temporal traffic forecasting and low-complexity mathematical models and algorithms for optimising RIS deployment for improving the radio propagation environments come with the potential impact of providing network operators with practical performance-engineering tools. The ongoing research and development work on integrating the new techniques as containerised microservices within the O-RAN RIC framework contributes to the industry's shift towards open, intelligent, and programmable wireless networks.
Further research and development are needed in order to accomplish the project objectives. Specific activities include the development of AI models that are able to provide high-resolution traffic prediction and uncertainty quantification, implementation of time-efficient algorithms for large-scale optimisation of RIS deployment and configuration to become part of a commercial RAN planning tool, as well as the expansion of the tool with AI-enabled optimisation engine and RIC microservices and associated application programming interfaces (APIs).
Moreover, via the project’s two industrial participants, continuous efforts are needed for identifying potential IPR, conducting commercialisation (i.e. integrating the results in the RAN planning tool), as well as interaction with relevant standardization bodies (e.g. Open RAN Alliance).
Further research and development are needed in order to accomplish the project objectives. Specific activities include the development of AI models that are able to provide high-resolution traffic prediction and uncertainty quantification, implementation of time-efficient algorithms for large-scale optimisation of RIS deployment and configuration to become part of a commercial RAN planning tool, as well as the expansion of the tool with AI-enabled optimisation engine and RIC microservices and associated application programming interfaces (APIs).
Moreover, via the project’s two industrial participants, continuous efforts are needed for identifying potential IPR, conducting commercialisation (i.e. integrating the results in the RAN planning tool), as well as interaction with relevant standardization bodies (e.g. Open RAN Alliance).