6llz / federated_learning

Enhance Vehicle Cooperative Perception with Federated Learning

Federated Learning for Cooperative Perception in Vehicular Networks (Ongoing Project)

The goal of this project is to enable vehicular cooperative perception using Deep Learning techniques in Federated Learning paradigm [1]. Technically speaking, this repository provides the modules to generate and manage a vehicular network in OMNeT++ [2], collect the communication statistics in Veins [3] (for now only bandwidth consumed by the message transmission and delay) as well as C++ - Python API [4]. A block diagram of the program can be found below.

Running the simulation for the first time can be a complex thing to do. For this purpose, below you can find a link to the video highlighting the program's main functionality.
https://user-images.githubusercontent.com/49762976/173159771-cc975128-f8bb-4162-b7bf-ea6cb3cc43b2.mov

Program versions used to run the simulation

• OMNeT++ 5.6.1
• Veins 5.2
• SUMO 1.11

It was executed on Ubuntu-20.04 via WSL

References

[1] https://ieeexplore.ieee.org/document/9473581
[2] https://doc.omnetpp.org/workshop2008/omnetpp40-paper.pdf
[3] https://ieeexplore.ieee.org/document/5510240
[4] https://zeromq.org/