Vision Transformer for Learning Driving Policies in Complex and Dynamic Environments

Loading...
Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Conference article in proceedings
Date
2022-07-19
Major/Subject
Mcode
Degree programme
Language
en
Pages
7
1558-1564
Series
2022 IEEE Intelligent Vehicles Symposium, IV 2022, IEEE Intelligent Vehicles Symposium, Proceedings, Volume 2022-June
Abstract
Driving in a complex and dynamic urban environment is a difficult task that requires a complex decision policy. In order to make informed decisions, one needs to gain an understanding of the long-range context and the importance of other vehicles. In this work, we propose to use Vision Transformer (ViT) to learn a driving policy in urban settings with birds-eye-view (BEV) input images. The ViT network learns the global context of the scene more effectively than with earlier proposed Convolutional Neural Networks (ConvNets). Furthermore, ViT's attention mechanism helps to learn an attention map for the scene which allows the ego car to determine which surrounding cars are important to its next decision. We demonstrate that a DQN agent with a ViT backbone outperforms baseline algorithms with ConvNet backbones pre-trained in various ways. In particular, the proposed method helps reinforcement learning algorithms to learn faster, with increased performance and less data than baselines.
Description
Publisher Copyright: © 2022 IEEE.
Keywords
Other note
Citation
Kargar , E & Kyrki , V 2022 , Vision Transformer for Learning Driving Policies in Complex and Dynamic Environments . in 2022 IEEE Intelligent Vehicles Symposium, IV 2022 . IEEE Intelligent Vehicles Symposium, Proceedings , vol. 2022-June , IEEE , pp. 1558-1564 , IEEE Intelligent Vehicles Symposium , Aachen , Germany , 05/06/2022 . https://doi.org/10.1109/IV51971.2022.9827348