SUMO Simulations for Privacy-Preserving Federated Learning in Autonomous Vehicles
Eclipse Foundation via YouTube
Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Learn about privacy implications and performance trade-offs in vehicular federated machine learning systems through an 18-minute technical presentation from the Eclipse Foundation. Explore how the Monaco SUMO Traffic Scenario (MoST) is used to assess privacy loss when vehicles share route information for collaborative learning of urban phenomena like parking occupancy. Discover the comparative analysis between individual, federated, and centralized learning approaches, with findings indicating federated systems offer enhanced privacy but reduced performance. Gain insights into implementing SUMO-based learning systems across multiple computers using Docker containerization and client-server architecture. Follow along as presenter Levente Alekszejenkó covers key topics including predictive ML models, learning schemes, scenario testing, performance evaluation, and privacy threat analysis in both spatial and temporal dimensions.
Syllabus
Intro
Motivation
Let us build predictive ML models!
Learning scheme proposal
How to test the learning schemes?
The scenario
Measuring & learning
The trained networks
Centralized learning
Performance evaluation
Privacy threat -space
Federated learning
Privacy threats -space
Privacy threats - time
Conclusions
Taught by
Eclipse Foundation