JSAI2020

Presentation information

Interactive Session

[4Rin1] Interactive 2

Fri. Jun 12, 2020 9:00 AM - 10:40 AM Room R01 (jsai2020online-2-33)

[4Rin1-64] Analysis of Cooperative Interaction between Creatures Using Neuroevolution

〇Ryosuke Takata1, Yugo Takeuchi1 (1.Graduate School of Integrated Science and Technology, Shizuoka University)

Keywords:Neuroevolution, Cooperation, Neural Network

In order for machines to cooperate, it is necessary to estimate the intention of others. Depending on the cooperative task, it is possible to cooperate smoothly by acting according to roles such as a leader / follower. The purpose of this research is to realize such cooperative behavior of machines automatically and analyze their behavior process. In order to realize cooperative behavior from the bottom up, we used NEAT, a typical method of neuroevolution. As an experiment, a cooperative task in which two creatures lift a heavy box was learned by NEAT. As a result of the learning, the behavior of the creatures that achieved cooperative behavior while adjusting the position flexibly was seen. When we analyzed the neural network, it was a structure that played the role of a leader / follower. In addition, as a result of training the creature in a state where it can send and receive light emission, the light emission emerged to mean as a signal to lift the box. In the future, we will consider the cooperation between humans and AI creatures.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password