Periodic Reporting for period 1 - BANYAN (Big dAta aNalYtics for radio Access Networks)
Período documentado: 2019-12-01 hasta 2021-11-30
With the increasing diversity and heterogeneity of mobile services used by people and machines, 4G/5G networks need to meet the growing variety of quality of service (QoS) requirements.
Due to the limitation of current mobile technologies, promising new technologies were emerged to tackle such issues. Network slicing, technologically enabled by Network Function Virtualization (NFV), is a promising paradigm. It improves the flexibility and elasticity of mobile networks by dynamically forming and combining logic slicing to adapt to the fluctuation of different mobile service demands. A key enabler for network slicing is accurate data-driven models and the prediction of the Spatio-temporal dynamics of the mobile service traffic, which allow discovering knowledge relevant to the orchestration of slices and anticipating the need for their reconfiguration.
Another important aspect that shall be considered is the increasing indoor traffic demand that more than 90% of mobile data traffic occurs within buildings; Therefore, most 5G RAN shall be deployed indoors to provide access points to the users. In the proximity of the indoor wireless access network (RAN), the demand for effective data-driven slice management is particularly critical. RAN must adapt to most of the capacity and demand changes related to each mobile service, and its performance is very important to users' QoS.
Objectives:
• Develop algorithms based on multivariate analysis and deep learning (DL) to forecast macroscopic spatial-temporal mobile service demands caused by user traffic and mobility;
• Develop data analytics based geolocation algorithms to geo-locate and characterize in-building mobile traffic demands at a high level of detail (e.g. including the exact time, building, and floor where traffic is generated); and
• Develop data analytics-driven mechanisms to proactively optimize the orchestration of virtualized 5G RAN resources, coordinate outdoor and indoor networks, and coordinate multi-RAT indoor networks.
In addition to the above research objectives, the project also has the following (B) doctoral training objectives:
• Train a group of 5 outstanding early-stage researchers (ESRs) for both academia and industry.
• Establish a virtual European center of excellence for data-driven 5G RAN research that will exist well beyond the end of this project, reducing the fragmentation and facilitating long-term transnational and inter-sectoral collaborations.
Due to the limitation of current mobile technologies, promising new technologies were emerged to tackle such issues. Network slicing, technologically enabled by Network Function Virtualization (NFV), is a promising paradigm. It improves the flexibility and elasticity of mobile networks by dynamically forming and combining logic slicing to adapt to the fluctuation of different mobile service demands. A key enabler for network slicing is accurate data-driven models and the prediction of the Spatio-temporal dynamics of the mobile service traffic, which allow discovering knowledge relevant to the orchestration of slices and anticipating the need for their reconfiguration.
Another important aspect that shall be considered is the increasing indoor traffic demand that more than 90% of mobile data traffic occurs within buildings; Therefore, most 5G RAN shall be deployed indoors to provide access points to the users. In the proximity of the indoor wireless access network (RAN), the demand for effective data-driven slice management is particularly critical. RAN must adapt to most of the capacity and demand changes related to each mobile service, and its performance is very important to users' QoS.
Objectives:
• Develop algorithms based on multivariate analysis and deep learning (DL) to forecast macroscopic spatial-temporal mobile service demands caused by user traffic and mobility;
• Develop data analytics based geolocation algorithms to geo-locate and characterize in-building mobile traffic demands at a high level of detail (e.g. including the exact time, building, and floor where traffic is generated); and
• Develop data analytics-driven mechanisms to proactively optimize the orchestration of virtualized 5G RAN resources, coordinate outdoor and indoor networks, and coordinate multi-RAT indoor networks.
In addition to the above research objectives, the project also has the following (B) doctoral training objectives:
• Train a group of 5 outstanding early-stage researchers (ESRs) for both academia and industry.
• Establish a virtual European center of excellence for data-driven 5G RAN research that will exist well beyond the end of this project, reducing the fragmentation and facilitating long-term transnational and inter-sectoral collaborations.
The amendment that regulated the BAYNAN consortium update, with the addition of IMDEA as the third partner, was prepared by the partners and informally approved by the project officer during Jun 2020. The new consortium began the ESR hiring after October 2020 when the amendment was approved.
1. Multiple types of social media accounts have been created to generate further outreach impact, including the BANYAN website that presents the project's progress, achievements, and other news to the wider public. Meanwhile, the project has also presented the related news via LinkedIn and Twitter accounts.
2. All 5 ESRs were recruited and enrolled in a Ph.D. program. ESRs have been working on their personal development plan (PDP), they also attended post-graduate courses at their PhD hosting university on topics related to their research program, along with a number of local training courses on complementary skills, including IT, research methods, oral and written communications, project management, public engagement, and personal development. ESRs also attended seminars given by local researchers and internationally recognized experts, as well as actively interact with local staff members.
3. Three research reports have been delivered describing the outputs of tasks 1.1 1.2 and 2.1.
-D1.1 Report on mobile traffic demand baseline analytics. The objective of this deliverable is to understand and forecast mobile services demands at a macroscopic scale, that is, citywide. It is motivated by the fact that network slicing, technology that allows the creation of multiple logical network instances on the same physical network, demands a service-oriented approach to managing mobile networks.
-D1.2 Report on mobile traffic demand multi-scale analytics. The deliverable aims at developing analytics that allows characterizing the demand generated by mobile services over time and space. Specifically, the goal is to capture hidden traffic structures emerging within urban areas.
-D2.1 Report on algorithms to geo-localized traffic to buildings. A new deep learning model for time series classification, exploiting the concept of self-attention for the first time in the context of indoor-outdoor detection. It shows that self-attention can significantly enhance the IOD accuracy, as it allows a sequence model to focus on the parts of the sensor reading sequence that are more important for the environment classification.
4. Training plan report of D4.1 was delivered, and the D4.2 personal development plan has been delivered.
5. dissemination and outreach plan D5.1 has been delivered, which pre-planned the dissemination and outreach in the next few years. The postponed training school was scheduled after all ESRs were recruited, and the originally scheduled training school will be held on time.
1. Multiple types of social media accounts have been created to generate further outreach impact, including the BANYAN website that presents the project's progress, achievements, and other news to the wider public. Meanwhile, the project has also presented the related news via LinkedIn and Twitter accounts.
2. All 5 ESRs were recruited and enrolled in a Ph.D. program. ESRs have been working on their personal development plan (PDP), they also attended post-graduate courses at their PhD hosting university on topics related to their research program, along with a number of local training courses on complementary skills, including IT, research methods, oral and written communications, project management, public engagement, and personal development. ESRs also attended seminars given by local researchers and internationally recognized experts, as well as actively interact with local staff members.
3. Three research reports have been delivered describing the outputs of tasks 1.1 1.2 and 2.1.
-D1.1 Report on mobile traffic demand baseline analytics. The objective of this deliverable is to understand and forecast mobile services demands at a macroscopic scale, that is, citywide. It is motivated by the fact that network slicing, technology that allows the creation of multiple logical network instances on the same physical network, demands a service-oriented approach to managing mobile networks.
-D1.2 Report on mobile traffic demand multi-scale analytics. The deliverable aims at developing analytics that allows characterizing the demand generated by mobile services over time and space. Specifically, the goal is to capture hidden traffic structures emerging within urban areas.
-D2.1 Report on algorithms to geo-localized traffic to buildings. A new deep learning model for time series classification, exploiting the concept of self-attention for the first time in the context of indoor-outdoor detection. It shows that self-attention can significantly enhance the IOD accuracy, as it allows a sequence model to focus on the parts of the sensor reading sequence that are more important for the environment classification.
4. Training plan report of D4.1 was delivered, and the D4.2 personal development plan has been delivered.
5. dissemination and outreach plan D5.1 has been delivered, which pre-planned the dissemination and outreach in the next few years. The postponed training school was scheduled after all ESRs were recruited, and the originally scheduled training school will be held on time.
1. BANYAN will produce a new generation of creative, entrepreneurial and innovative ESRs, many of whom will become future leaders in academia and industry and will shape the future of a fully connected, intelligent world.
2. BANYAN will:
(1) develop EU capacity and skills to advance big data analytics (BDA), machine learning, indoor localisation, intelligent resource allocation and proactive network optimisation;
(2) integrate BDA and machine learning more closely with 5G to facilitate efficient NFV;
(3) increase the QoE of 5G users (people and things) in diverse and dynamic services such as eMBB, uRLLC and mMTC;
(4) reduce cost and energy per bit;
(5) increase EU competitiveness in 5G and beyond 5G, smart building/city;
(6) develop lasting academic and industry collaboration which ensures interdisciplinary and cross-sector research training programmes;
(7) create new and exciting career prospects for the fellows.
3. BANYAN will Contribute to structuring doctoral training at EU level and strengthening EU innovation capacity.
4. BANYAN technical results obtained during the project are described in:
-D1.1 Report on mobile traffic demand baseline analytics
-D1.2 Report on mobile traffic demand multi-scale analytics
-D1.3 Report on Mobile traffic demand predictors
-D2.1 Report on algorithms to geo-localise traffic to buildings
-D2.2 Report on algorithms to localise indoor UEs
-D2.3 Report on in-building mobile traffic characterisation
-D2.4 Recommendation to 3GPP on how to rank buildings for phased in-building mobile network deployment
-D3.1 Orchestrators for network slice management at the mobile edge
-D3.2 Report on joint outdoor-indoor optimization based on BDA (to 3GPP, NGMN, Small Cell Forum, etc.)
-D3.3 Report on proactive optimization of indoor networks (to 3GPP, NGMN, Small Cell Forum, etc.)
2. BANYAN will:
(1) develop EU capacity and skills to advance big data analytics (BDA), machine learning, indoor localisation, intelligent resource allocation and proactive network optimisation;
(2) integrate BDA and machine learning more closely with 5G to facilitate efficient NFV;
(3) increase the QoE of 5G users (people and things) in diverse and dynamic services such as eMBB, uRLLC and mMTC;
(4) reduce cost and energy per bit;
(5) increase EU competitiveness in 5G and beyond 5G, smart building/city;
(6) develop lasting academic and industry collaboration which ensures interdisciplinary and cross-sector research training programmes;
(7) create new and exciting career prospects for the fellows.
3. BANYAN will Contribute to structuring doctoral training at EU level and strengthening EU innovation capacity.
4. BANYAN technical results obtained during the project are described in:
-D1.1 Report on mobile traffic demand baseline analytics
-D1.2 Report on mobile traffic demand multi-scale analytics
-D1.3 Report on Mobile traffic demand predictors
-D2.1 Report on algorithms to geo-localise traffic to buildings
-D2.2 Report on algorithms to localise indoor UEs
-D2.3 Report on in-building mobile traffic characterisation
-D2.4 Recommendation to 3GPP on how to rank buildings for phased in-building mobile network deployment
-D3.1 Orchestrators for network slice management at the mobile edge
-D3.2 Report on joint outdoor-indoor optimization based on BDA (to 3GPP, NGMN, Small Cell Forum, etc.)
-D3.3 Report on proactive optimization of indoor networks (to 3GPP, NGMN, Small Cell Forum, etc.)