Project Title: Quality of Service for the Internet Things in Smart Cities via Predictive Networks
This project is a Marie Skłodowska-Curie Individual Fellowships project funded under the Horizon 2020 research and innovation programme by the European Commission (GA No: 846077).
Non-Technical Exposition (for the General Audience)
Fifth Generation (5G) cellular systems are already being deployed all around the world, which will make an enormous impact on our lives: These systems enable a variety of applications, such as remote surgery (as part of the Tactile Internet), augmented (virtual) reality, industrial and vehicular automation, autonomous driving, enhanced mobile broadband as well as Internet of Things (IoT). These applications will become an important part of our digitally connected lives and form the new infrastructure of the smart cities of the new future. Among these applications, IoT appears as a key enabler of technologies that will be deployed in smart cities that will range from smart bins that indicate to the municipality when they are full, smart lamp posts that adjust their lighting in response to the needs of pedestrians and cars for maximum benefit while reducing energy consumption, and a plethora of other services that smart cities will provide to their dwellers. While 5G provides the necessary initial infrastructure world-wide for these diverse services to take off, the evolution towards Sixth Generation (6G) networks will bring in new enabler technologies, such as Artificial Intelligence (AI), which will become part of the digital landscape in smart cities.
The main goal of this project is to enable the plethora of applications in smart cities via the development of AI technology for IoT. We achieve this by concentrating on a key measure of communication network performance called “Quality of Service (QoS)”. For example, on a video call, QoS can be measured by examining how long it takes for your video and audio to reach the other party, as well as the image quality during the call and whether you have any freezes and interruptions. Similar QoS metrics can be identified for each application that we have mentioned above. The Internet must be able to accommodate a variety of services while achieving the QoS of each such application. The main idea behind this project is to utilize AI to enable the operation of smart cities of the near future by delivering the QoS required for each such application.
Technical Exposition (for Specialists)
Over the past ten years, Artificial Intelligence, or more specifically Machine Learning (ML), has been applied to communication networks: The applications range from traffic prediction, congestion control, traffic routing, Quality of Service (QoS) and Quality of Experience (QoE) management to fault localization and network security. In particular, predictive communication networks have gained popularity over the past ten years, where predictions are enabled by ML techniques. These predictions enable the network to foresee the future conditions of the communication network and act accordingly and autonomously. Furthermore, the rise of Software-Defined Networking (SDN) over the past ten years has enabled “network virtualization”, where network resources are no longer allocated at the level of hardware but are allocated at the level of software. For example, in 5G systems, it is possible to open end-to-end “QoS flows’’, which can be visualized as end-to-end tunnels, e.g. from a surgeon who is operating on a patient in another country during remote surgery. Applications such as remote surgery require ultra-high reliability as well as very low latency (URLLC). These QoS flow tunnels enable the surgeon to operate on the remote patient in the presence of a highly-reliable and low-latency end-to-end virtual connection. These URLLC connections must co-exist on the Internet with Mobile Broadband (MBB) traffic, such as video streaming, as well as massive Machine Type Communication (mMTC), such as small IoT or sensor devices that generate very small amounts of data. Building the communication infrastructure of smart cities must successfully accommodate this diversity of traffic types.
In this project, we build Machine Learning based algorithms in order to enable Quality of Service (QoS) delivery, which will become the building blocks of Sixth Generation (6G) networks and beyond. Our work leverages Software Defined Networking (SDN) and builds upon the network virtualization that has been achieved both at the Radio Access Network (RAN) and Core Network (CN) levels. The main difference between our work and existing 5G networks is that we heavily utilize Artificial Intelligence in determining QoS flows over the RAN ad CN and develop novel communication protocols that have the potential to achieve a QoS performance beyond existing 5G networks.
Copyright © Volkan Rodoplu
Links to Conference and Journal Papers
Comparative Study of Forecasting Schemes for IoT Device Traffic in Machine-to-Machine Communication