Distributed Wireless LEarning:
Edge AI for a better wireless network
Problem Statement:
So many different Iot devices trying to communicate and everyone of them working with standard settings in an unknown, dynamic environment..see what could go wrong? In this project we want to design a distributed intelligence mechanism in which various IoT nodes are able to choose the best local parameters to optimise the Quality of Service and adapt their behaviour to compensate for network dynamics in order to be the best possible network it can be!
What are we looking for?
- Distributed Machine Learning (esp. Federated and Multi-Agent Reinforcement Learning)
- IoT communication simulation
- Low-computational AI
Tasks:
The students will receive an IoT wireless network simulator and develop their own flavour of Edge AI algorithms. Specifically they will:
- Model wireless communication performance based on local IoT devices capabilites (sensor node, gateway, actuators ...)
- Build multi-agent solutions in which each IoT device is able to steer its own behaviour towards a (global or local) optimum
- A feedback mechanism which teaches the node if they are improving the whole network's quality of service or not.
Work:
40% Theory, 40% Simulations, 20%Writing.
Contact:
Alessandro Chiumento (a.chiumento@utwente.nl)