Machine learning in distributed networking environments leveraging the network function virtualisation (NFV) paradigm
External reference persons Prof. Falko Dressler (TUB, Germania)
Description Next generation 5G/6G solutions will be based on network function virtualization (NFV), heterogeneous communication, and mobile edge computing. These new paradigms are indeed very promising to achieve ultra-low latency and ultra-high reliability, thus, enabling new services for mobile users. At the same time, vehicles are, and will ever be, equipped with new ICT capabilities and a multitude of communication interfaces.
In this context, we are integrating the ICT capabilities of vehicles with those of the network infrastructure for the creation of innovative services in this domain. Here, distributed machine learning, most importantly federated learning, is one of the most relevant applications.
This thesis addresses the deployment of machine learning architectures in distributed networking environments, leveraging the network function virtualisation (NFV) paradigm. A network of virtual machines implementing containers will be set up, in order to realise a federated learning scheme. The single containers are considered to run in fixed as well as mobile node, whose connectivity is simulated through a network simulator (Veins). In the relevant case of automated connected cars, the nodes’ mobility is modelled through the well-known SUMO simulator.
The thesis will include the following main steps:
- Building competences on Docker containers, virtual machines and specific simulation tools (SUMO, Veins)
- Realization of a testbed in virtual computing environment with virtual machines and containers
- Deploy a Federated Learning scenario
- Experimental test for validation and performance assessment
Required skills good programming skills (C, or C++, or Python), basic understanding of telecommunications, basic understanding of simulation tools, basic understanding of virtual machines
Deadline 09/11/2022 PROPONI LA TUA CANDIDATURA