Deep Learning for Tra c Signal Control with Variational Inference
Keywords
No Thumbnail Available
Authors
Issue Date
2021-06-18
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Optimizing the traffic signal control problem becomes more relevant as the amount of traffic participants and vehicles in the world increases. Detection of vehicles at intersections may fail which results in vehicles to be unobserved. To cope with incomplete data, in this paper, a Variational Autoencoder (VAE) introduces uncertainty to the scenario. Graph Neural Networks (GNN) are used for efficient communication between agents. Tests and simulations show that the full model (GNNVAE) proposed in this paper, using an existing state of the art reinforcement learning model, could improve performance for partially observed intersections.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen