Real-time medical data in the cloud

The Internet-of-Medical-Things (IoMT) is growing in application number and complexity. The sheer amount of data produced can be further analysed in a centralized way with the more powerful means of the cloud systems. I have been researching the technologies for massive data gathering in hospitals of lung ventilator data, and their back-up/analysis on cloud back end.

Official website

Colosseum

Colosseum is a wireless network channel emulator with hardware in the loop[1] sporting 256 USRP Software Defined Radios designed and released by DARPA in 2019.

  • Despite being a flexible network emulator, standard SDR cannot support 802.11 packet timing requirements; designed, implemented, and deployed an extension using Zylinx System-on-chip boards and OpenWiFi to fully enable Wi-Fi communications [2].
Official website

Dynamic Spectrum Sharing

The increasing demand for wireless network communication has led to an overcrowding of the wireless spectrum. As we are running out of available licensed bands, we can exploit spectrum sensing and machine learning to devise intelligent and dynamic wireless spectrum sharing techniques.

  • Research on deep neural networks applied to raw wireless I/Q samples;
  • Design and implementation of intelligent prototypes for spectrum sharing [1];
  • Registered a patent, Channel-Aware Reactive Mechanism (ChARM), US 63/244,192

Network Generators

NetworkX is a Python package for the manipulation and the study of complex networks.

  • Designed and developed a random graph generator reproducing given global properties, such as modularity and community structure [1][2]
  • Implemented a random graph generator for internet-like networks [3] based on BGP architectures
Official website -- Source code

De-centralized live streaming

PeerStreamer-ng is a distributed application for media live streaming; its core and back-end are written in C to grant a high level of portability and minimal resource footprint, while it uses HTML5 and WebRTC for the graphical user interface and front-end, hence, being usable on any device supporting a web browser.

  • Designed and implemented the full-stack system [1]
  • Tested and used with the Wireless Community Network users of Ninux in Italy and AWMN in Greece
  • Participated in the media streaming development and deployment tasks of the multi-disciplinary, European netCommons project
Official website -- Source code

Content distribution optimization

Delay distribution guarantees and high-probability delivery time reception in unstructured networks.

  • Designed a mathematical model and derived its theoretical results for resource optimization and delay minimization in unstructured network broadcasting. The solution can be easily implemented through neighbour gossiping and the optimized system improves reception delay by 60% and packet loss by half in simulated networks [1]
  • Analytically derived stochastic bounds for content delay reception in a mesh distribution network. These are of particular interest for real-time broadcasting applications [2][3]
  • Analitically derived age of information metrics in unstructured vehicular networks [4]
Source code

Networking emulation

Network Protocol and Application Testing Toolchain (NePA TesT) is a network emulator built on top of mininet and focused on prototyping. It exploits the Linux kernel namespaces to obtain a lightweight virtualisation and emulate networks of thousands of nodes.

  • Designed and implemented the network emulator [1]
  • Emulated live video streaming distributions and the OLSR routing protocol on real topologies [1]
Source code

Network cross-layer optimization

Distributed optimization of resource usage and communication delays in unstructured mesh networks; with a particular focus on Community Networks.

  • Derived a cross-layer optimization technique leveraging link-state protocol information in mesh networks for reducing link bottlenecks and increase network resource usage fairness. Emulated results on real-world networks show reduction of overall link usage up to 66% [1][2]