Federated Learning in Smart-Bikes
Source: Accell Group
Problem Statement:
We aim to maintain road safety of bike riders. As of 2018, about 1/4th of the daily mobility in The Netherlands is by bikes. Additionally, the number of road fatalities of bike riders has increased in 2020 (CBS Survey 2021). This throws light on the need for road safety of bike riders.
At the Pervasive Systems Group, we aim to address this problem by embedding intelligence in bikes using sensors and communication units. We would like to study the use of Federated Learning in Smart-Bikes which aid in improving the road safety. We are interested in Federated Learning because of its privacy preserving nature.
To put things in context, we want to investigate what insight could be gained from a distributed set of smart-bikes, automobiles and other infrastructural elements (e.g. traffic signals). However, there are quite some challenges (e.g. inconsistent data, aggregation issues, etc.) in such a distributed learning system.
Tasks:
The prospective student is expected study the literature on recent developments in road safety and Federated Learning. Specifically, the focus should be on studying the challenges in Federated Learning and proposing a solution. The student is also expected to be familiar with Python language and be motivated to work on Vehicular Networks and Machine Learning methods.
WORK:
60% Theory, 20% Practical, 20%Writing
Contact
Deepak Yeleshetty - d.yeleshetty@utwente.nl