• Classified by Topic • Classified by Publication Type • Sorted by Date • Sorted by First Author Last Name • Classified by Funding Source •
Learning Real-world Autonomous Navigation by Self-Supervised Environment Synthesis.
Zifan
Xu, Anirudh Nair, Xuesu Xiao, and Peter
Stone.
In IROS Workshop on Photorealistic Image and Environment Synthesis for Robotics (PIES-Rob) , January
2023.
Machine learning approaches have recently enabled autonomous navigation formobile robots in a data-driven manner. Since most existing learning-basednavigation systems are trained with data generated in artificially createdtraining environments, during real-world deployment at scale, it is inevitablethat robots will encounter unseen scenarios, which are out of the trainingdistribution and therefore lead to poor real-world performance. On the otherhand, directly training in the real world is generally unsafe and inefficient. Toaddress this issue, we introduce Self-supervised Environment Synthesis (SES), inwhich, after real-world deployment with safety and efficiency requirements,autonomous mobile robots can utilize experience from the real-world deployment,reconstruct navigation scenarios, and synthesize representative trainingenvironments in simulation. Training in these synthesized environments leads toimproved future performance. in the real world. In our experiments, theeffectiveness of SES in synthesizing representative simulation environments andimproving real-world navigation performance has been verified by a large-scaledeployment in a high-fidelity, realistic simulator
@InProceedings{zifan_xu_iros_2023,
author = {Zifan Xu and Anirudh Nair and Xuesu Xiao and Peter Stone},
title = {Learning Real-world Autonomous Navigation by Self-Supervised Environment Synthesis},
booktitle = {IROS Workshop on Photorealistic Image and Environment Synthesis for Robotics (PIES-Rob) },
year = {2023},
month = {January},
location = {Detroit, Michigan},
abstract = {Machine learning approaches have recently enabled autonomous navigation for
mobile robots in a data-driven manner. Since most existing learning-based
navigation systems are trained with data generated in artificially created
training environments, during real-world deployment at scale, it is inevitable
that robots will encounter unseen scenarios, which are out of the training
distribution and therefore lead to poor real-world performance. On the other
hand, directly training in the real world is generally unsafe and inefficient. To
address this issue, we introduce Self-supervised Environment Synthesis (SES), in
which, after real-world deployment with safety and efficiency requirements,
autonomous mobile robots can utilize experience from the real-world deployment,
reconstruct navigation scenarios, and synthesize representative training
environments in simulation. Training in these synthesized environments leads to
improved future performance. in the real world. In our experiments, the
effectiveness of SES in synthesizing representative simulation environments and
improving real-world navigation performance has been verified by a large-scale
deployment in a high-fidelity, realistic simulator
},
}
Generated by bib2html.pl (written by Patrick Riley ) on Tue Nov 19, 2024 10:24:46