Generating and Analyzing Collective Behavior in a Robotic Swarm by the Use of Deep Reinforcement Learning and Deep Neuroevolution

This study proposes a method to apply deep neural networks to controllers of robotic swarms. In a typical approach to design controllers, the designer has to define the features extracted from sensory inputs in advance. By applying deep neural networks with convolution layers, it can automatically e...

Full description

Saved in:
Bibliographic Details
Published inShisutemu Seigyo Jouhou Gakkai rombunshi Vol. 33; no. 5; pp. 163 - 170
Main Authors Morimoto, Daichi, Hiraga, Motoaki, Ohkura, Kazuhiro, Matsumura, Yoshiyuki
Format Journal Article
LanguageJapanese
English
Published Kyoto THE INSTITUTE OF SYSTEMS, CONTROL AND INFORMATION ENGINEERS (ISCIE) 15.05.2020
Japan Science and Technology Agency
Subjects
Artificial neural networks
Computer simulation
Controllers
Convolution
Deep learning
deep neuroevolution
deep Q-network
deep reinforcement learning
Feature extraction
Machine learning
Neural networks
swarm robotics
Online AccessGet full text

Cover

More Information
Summary:This study proposes a method to apply deep neural networks to controllers of robotic swarms. In a typical approach to design controllers, the designer has to define the features extracted from sensory inputs in advance. By applying deep neural networks with convolution layers, it can automatically extract features from sensory inputs. We applied two methods to train the deep neural networks, i.e.,deep reinforcement learning and deep neuroevolution. The controllers were tested in a path-formation task using computer simulations. Compared with deep reinforcement learning, deep neuroevolution was able to generate collective behavior even in sparse reward settings.
ISSN:1342-5668
2185-811X
DOI:10.5687/iscie.33.163