Deep learning of structured environments for robot search

Robots often operate in built environments containing underlying structure that can be exploited to help predict future observations. In this work, we present a framework based on convolutional neural networks to predict point of interest locations in structured environments. The proposed technique...

Full description

Saved in:
Bibliographic Details
Published inAutonomous robots Vol. 43; no. 7; pp. 1695 - 1714
Main Authors Caley, Jeffrey A., Lawrance, Nicholas R. J., Hollinger, Geoffrey A.
Format Journal Article
LanguageEnglish
Published New York Springer US 01.10.2019
Springer Nature B.V
Subjects
Artificial Intelligence
Artificial neural networks
Computer Imaging
Control
Deep learning
Domains
Engineering
Mechatronics
Neural networks
Pattern Recognition and Graphics
Robotics
Robotics and Automation
Robots
Searching
Unknown environments
Urban environments
Vision
Online AccessGet full text

Cover

More Information
Summary:Robots often operate in built environments containing underlying structure that can be exploited to help predict future observations. In this work, we present a framework based on convolutional neural networks to predict point of interest locations in structured environments. The proposed technique exploits the inherent structure of the environment to train a convolutional neural network that is leveraged to facilitate robotic search. We start by investigating environments where the full environmental structure is known, and then we extend the work to unknown environments. Experimental results show the proposed framework provides a reliable method for increasing the efficiency of current search methods across multiple domains. Finally, we demonstrate the proposed framework increases the search efficiency of a mobile robot in a real-world office environment.
ISSN:0929-5593
1573-7527
DOI:10.1007/s10514-018-09821-4