Experimental Study on Behavior Acquisition of Mobile Robot by Deep Q-Network
- 20 September 2017
- journal article
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 21 (5), 840-848
- https://doi.org/10.20965/jaciii.2017.p0840
Abstract
Deep Q-network (DQN) is one of the most famous methods of deep reinforcement learning. DQN approximates the action-value function using Convolutional Neural Network (CNN) and updates it using Q-learning. In this study, we applied DQN to robot behavior learning in a simulation environment. We constructed the simulation environment for a two-wheeled mobile robot using the robot simulation software, Webots. The mobile robot acquired good behavior such as avoiding walls and moving along a center line by learning from high-dimensional visual information supplied as input data. We propose a method that reuses the best target network so far when the learning performance suddenly falls. Moreover, we incorporate Profit Sharing method into DQN in order to accelerate learning. Through the simulation experiment, we confirmed that our method is effective.Keywords
This publication has 7 references indexed in Scilit:
- Acquisition of Goal-Oriented Behavior for Snake-Like Robot by CPG and Reinforcement LearningJournal of Japan Society for Fuzzy Theory and Intelligent Informatics, 2017
- Expectations of Robot Field from Artificial Intelligence FieldJournal of the Robotics Society of Japan, 2017
- Mobile robots exploration through cnn-based reinforcement learningRobotics and Biomimetics, 2016
- Human-level control through deep reinforcement learningNature, 2015
- Q-PSP Learning: An Exploitation-Oriented Q-Learning Algorithm and Its ApplicationsTransactions of the Society of Instrument and Control Engineers, 1999
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- Q-learningMachine Learning, 1992