This project aims at developing theoretical foundations for in-network data management to exploit different types of network resources, enhance the performance of the distributed machine learning models and corresponding applications. Distributed machine learning leverages resources at the network edge, near the data sources, for model training and inference, thereby offloading from these tasks the centralized cloud platforms and satisfying demands of applications for low latency, scalability and privacy. Training ML models at realistic 5GIoT network scales requires processing enormous volumes of data generated by the end-user devices, thus leading to huge amount of data communications between network users and the central cloud platform. As a consequence, the data uploading and training procedure in real-time applications (e.g., ARVR) causes significant performance degradation. Finally, since the training data are collected by network users, preserving the privacy of network users with affordable accuracy loss is an important issue in distributed ML. Developing data aware resource allocation algorithms in distributed ML systems is a complex and open problem. Existing works on distributed ML can be classified into two categories: (i) one focusing on how to run the ML tasks (model training and inference) at the edge devices in a collaborative and resource-optimized manner, and (ii) another focusing on how to exploit ML to find a better solution to the optimization problems in the edge network. It has been illustrated that data quality is a critical factor in machine learning algorithms, and data importance-aware network resource allocation and training algorithms can improve ML systems performance. Therefore, a unified treatment of data selection, communication and computing is required to improve the performance of distributed ML systems. More importantly, protecting against data leakage in the distributed resource-sharing process is an unresolved practical issue in protocol realization.
The Industrial Distributed Edge Architecture over Machine Intelligence for Local Learning (IDEA-MILL) research project addresses the key challenges in future hyper-connected industrial systems to deliver innovative solutions, which are prepared to handle the extremely large data volumes collected from multiple heterogeneous sources, make adequate inference, and provide timely response. Our proposed distributed edge architecture diffuses machine intelligence across the network by bringing more critical and demanding applications closer to smart machines while moving lightweight data and local learning outcomes to the cloud for enterprise-level analytics. Our expected key scientific results are in developing efficient, flexible, and scalable methods that automate model training processes as well as control storage, computing, networking, and radio resources to minimize the associated costs and latencies.
One of the key use cases of 6G networks will be massive twinning, which is an extension of the digital twin concept of industrial processes introduced in 5G to different type of physical objects. In this perspective, a system based on the use of a digital twin for optimized sensing and communication in Personal Area Networks (PANs) is currently an under researched and not commercialized area. This project aims to start filling the gap in this respect.
The idea of Personal Area Network (PAN) has largely evolved in recent years as result of the increasing number of smart devices that can be interconnected within an individual person’s space, as well as from the multiple type of connectivity technologies that can be used to connect the devices (i.e., short range and long range). Wearable devices such as smartwatches, smart-bands (activity trackers), smart-glasses, and smart portable sensors (e.g., air quality sensors) represent only a limited subset of new devices that can operate within a PANs, along with more conventional computing platforms such as personal computers, smartphones, and tablets. Typically, the set of these devices simultaneously execute multiple applications, which are in turn developed by separate providers, operating according to custom-made APIs and recommended set-up (e.g., type of connectivity to be used, frequency of the sensor data readings, etc.). The entire PAN generates a large amount of data that is typically processed through dedicated services hosted in cloud environments in a multi-tenancy fashion.
Challenges. The data generated by each application running in the cluster of PAN devices is uploaded to the different back-end applications hosted in cloud facilities in a non-orchestrated way. The lack of such orchestration mechanisms, together with the software heterogeneity and the inconsistent setup of the applications executed in each device, can generate various inefficiencies from the perspective of: (Challenge I) overall energy efficiency of the PAN, (Challenge II) poorer sensing data quality due to data redundancy and data inconsistency, and (Challenge III) inefficient offloading and data upload scheduling policies.
Project Goal. Considering the set of challenges and requirements described above, we have identified two key areas that will be driving the R&D activities of our project. The first relates to the development of a smart application "SmartSense Mesh", which can be interoperably installed and executed in a wide set of smart devices belonging to a PAN, for performing actionable data-driven insights aimed to ensure optimized sensing and communication. The second is linked to the development of an innovative digital twin framework "SmartSense Mesh-DT", which can assess—using AI-powered analytics—the adequacy and efficiency of a corresponding physical PAN system.
6G Finland is an active coalition of Finnish 6G R&D organizations to advance the impact of Finnish 6G expertise globally, build new international partnerships, and intensify national 6G development efforts towards sustainable and data-driven society enabled by instant and unlimited wireless connectivity.
Finland has a long and successful history in the field of wireless mobile technologies. Numerous successes have shown us the power of cooperation. It has – once again – led us to a global leading position, now in 6G research and development. The world’s first large-scale 6G research program, 6G Flagship was launched in Finland in 2018.
Currently, Finland also leads the European 6G flagship initiative, Hexa-X-II funded by EU and plays a significant role in other 6G measures of the EU as well. The motivation for the Hexa-X-II project stems from the ongoing Hexa-X project, which has successfully developed a vision for 6G wireless networks and positioned itself at the centre of 6G development on a global scale, particularly in Europe. The key motivating factors for Hexa-X-II include technology push, society pull, and strategic autonomy.
6G Finland is a national contact point of Finnish 6G know-how, and actively participates in 6G discussion both nationally and internationally. 6G Finland operates as a network to which new members are invited on a content basis.
The Nokia Center for Advanced Research (NCAR) is a joint center with University of Helsinki, Aalto University, and Nokia. NCAR has the mission of developing and exploring radical new ideas and techniques for scalable, robust, and efficient software for future networks and mobile solutions. The NCAR team combines expertise from multiple areas, such as distributed systems, network science, and machine learning. The team comprises of researchers from the University of Helsinki, Aalto University, and Nokia.
MegaSense: Scalable real-time 5G air pollution sensing as a service for megacities.
MegaSense utilizes 5G network and city reference monitoring stations for precise atmospheric real time readings and Artificial Intelligence platform for a spatially distributed network to field calibrate low cost sensors in operational mode for fixed monitoring stations and mobile air quality monitoring stations to reduce high cost of air quality infrastructure.
This Special Interest Group (SIG) is aimed to harness the synergy between AI and Edge computing and improving the interaction between two flagships: the 6G flagship, and the FCAI flagship. Edge computing is revolutionizing communication networks and is expected to be a key component of next generation network. AI solutions can benefit by harnessing the potential of Edge Computing, and edge computing processes can benefit by AI. This page highlights how each of our research program interacts with Edge Computing within FCAI and lists the groups currently involved.
This SIG will help organize events that improve the synergy between the 6G and FCAI flagships. It will provide a platform for discussing the impact edge computing and AI have on each other.
This SIG is aimed to harness the synergy between AI and Edge computing and improving the interaction between two flagships: the 6G flagship, and the FCAI flagship. Edge computing is revolutionizing communication networks and is expected to be a key component of next generation network. AI solutions can benefit by harnessing the potential of Edge Computing, and edge computing processes can benefit by AI. This page highlights how each of our research program interacts with Edge Computing within FCAI and lists the groups currently involved.
This SIG will help organize events that improve the synergy between the 6G and FCAI flagships. It will provide a platform for discussing the impact edge computing and AI have on each other.
We also have a mailing list fcai6g-sig-edgeai@helsinki.fi. You can subscribe to this mailing list by sending an email to majordomo@helsinki.fi with the following command in the body of your email message: subscribe fcai6g-sig-edgeai. This mailing list will be primarily used to notify the upcoming events.
We also request you to join the slack workspace edgeai-oulu.slack.com
Coordination: Professor Sasu Tarkoma (University of Helsinki), Dr. Ella Peltonen (University of Oulu), Dr. Ashwin Rao (University of Helsinki)