DeepGoal: Learning to drive with driving intention from human control demonstration

Ma, H; Wang, Y; Xiong, R; Kodagoda, S; Tang, L

DeepGoal: Learning to drive with driving intention from human control demonstration

Ma, H Wang, Y Xiong, R Kodagoda, S

Tang, L

Permalink

Publisher:: ELSEVIER
Publication Type:: Journal Article
Citation:: Robotics and Autonomous Systems, 2020, 127
Issue Date:: 2020-05-01

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

The embargo period expires on 1 May 2022

Adobe PDF

Download full textAdobe PDF (6.89 MB)

View on publisher's site

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Ma, H
dc.contributor.author	Wang, Y
dc.contributor.author	Xiong, R
dc.contributor.author	Kodagoda, S https://orcid.org/0000-0001-5175-9138
dc.contributor.author	Tang, L
dc.date.accessioned	2020-11-27T21:59:05Z
dc.date.available	2020-11-27T21:59:05Z
dc.date.issued	2020-05-01
dc.identifier.citation	Robotics and Autonomous Systems, 2020, 127
dc.identifier.issn	0921-8890
dc.identifier.issn	1872-793X
dc.identifier.uri	http://hdl.handle.net/10453/144435
dc.description.abstract	© 2020 Elsevier B.V. Recent research on automotive driving has developed an efficient end-to-end learning mode that directly maps visual input to control commands. However, it models distinct driving variations in a single network, which increases learning complexity and is less adaptive for modular integration. In this paper, we re-investigate human's driving style and propose to learn an intermediate driving intention region to relax the difficulties in end-to-end approach. The intention region follows both road structure in image and direction towards goal in public route planner, which addresses visual variations only and figures out where to go without conventional precise localization. Then the learned visual intention is projected on vehicle local coordinate and fused with reliable obstacle perception to render a navigation score map that is widely used for motion planning. The core of the proposed system is a weakly-supervised cGAN-LSTM model trained to learn driving intention from human demonstration. The adversarial loss learns from limited demonstration data with one local planned route and enables reasoning of multi-modal behaviors with diverse routes while testing. Comprehensive experiments are conducted with real-world datasets. Results indicate the proposed paradigm can produce more consistent motion commands with human demonstration and shows better reliability and robustness to environment change. Our code is available at https://github.com/HuifangZJU/visual-navigation.
dc.language	English
dc.publisher	ELSEVIER
dc.relation.ispartof	Robotics and Autonomous Systems
dc.relation.isbasedon	10.1016/j.robot.2020.103477
dc.rights	info:eu-repo/semantics/embargoedAccess
dc.subject	0801 Artificial Intelligence and Image Processing, 0906 Electrical and Electronic Engineering, 0913 Mechanical Engineering
dc.subject.classification	Industrial Engineering & Automation
dc.title	DeepGoal: Learning to drive with driving intention from human control demonstration
dc.type	Journal Article
utslib.citation.volume	127
utslib.for	0801 Artificial Intelligence and Image Processing
utslib.for	0906 Electrical and Electronic Engineering
utslib.for	0913 Mechanical Engineering
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Strength - CAS - Centre for Autonomous Systems
pubs.organisational-group	/University of Technology Sydney
utslib.copyright.status	open_access	*
utslib.copyright.embargo	2022-05-01T00:00:00+1000Z
dc.date.updated	2020-11-27T21:58:59Z
pubs.publication-status	Published
pubs.volume	127

Abstract:

© 2020 Elsevier B.V. Recent research on automotive driving has developed an efficient end-to-end learning mode that directly maps visual input to control commands. However, it models distinct driving variations in a single network, which increases learning complexity and is less adaptive for modular integration. In this paper, we re-investigate human's driving style and propose to learn an intermediate driving intention region to relax the difficulties in end-to-end approach. The intention region follows both road structure in image and direction towards goal in public route planner, which addresses visual variations only and figures out where to go without conventional precise localization. Then the learned visual intention is projected on vehicle local coordinate and fused with reliable obstacle perception to render a navigation score map that is widely used for motion planning. The core of the proposed system is a weakly-supervised cGAN-LSTM model trained to learn driving intention from human demonstration. The adversarial loss learns from limited demonstration data with one local planned route and enables reasoning of multi-modal behaviors with diverse routes while testing. Comprehensive experiments are conducted with real-world datasets. Results indicate the proposed paradigm can produce more consistent motion commands with human demonstration and shows better reliability and robustness to environment change. Our code is available at https://github.com/HuifangZJU/visual-navigation.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/144435