Towards vision-based deep reinforcement learning for robotic motion control
Data(s) |
06/09/2015
|
---|---|
Resumo |
This paper introduces a machine learning based system for controlling a robotic manipulator with visual perception only. The capability to autonomously learn robot controllers solely from raw-pixel images and without any prior knowledge of configuration is shown for the first time. We build upon the success of recent deep reinforcement learning and develop a system for learning target reaching with a three-joint robot manipulator using external visual observation. A Deep Q Network (DQN) was demonstrated to perform target reaching after training in simulation. Transferring the network to real hardware and real observation in a naive approach failed, but experiments show that the network works when replacing camera images with synthetic images. |
Formato |
application/pdf |
Identificador | |
Relação |
http://eprints.qut.edu.au/92332/1/pap168.pdf http://www.araa.asn.au/acra/acra2015/papers/pap168.pdf Zhang, Fangyi, Leitner, Jürgen, Milford, Michael, Upcroft, Ben, & Corke, Peter (2015) Towards vision-based deep reinforcement learning for robotic motion control. In Australasian Conference on Robotics and Automation 2015, 2-4 December 2015, Canberra, A.C.T. http://purl.org/au-research/grants/ARC/CE140100016 |
Direitos |
Copyright 2015 [Please consult the author] |
Fonte |
ARC Centre of Excellence for Robotic Vision; School of Electrical Engineering & Computer Science; Science & Engineering Faculty |
Palavras-Chave | #080101 Adaptive Agents and Intelligent Robotics #090602 Control Systems Robotics and Automation #Robotic Manipulation #Motion Control #Target Reaching #Deep Reinforcement Learning #DQN |
Tipo |
Conference Paper |