1A1-J02 Agent with Intensified Intelligence and Inclusive Strategy for Kendo: "I"s : 1st Report: Attack Motion Prediction(Evolution and Learning for Robotics)

Abstract

In this article, we propose a method of attack motion prediction for Kendo agent. The attack motion is modeled by Gaussian Mixture Model (GMM). Then, parameters of the GMM illustrate the motion as a ordinary differential equation. In order to predict the motion, the ordinary differential equation is solved by using the Euler' method. Finally, to evaluate the proposed method, verification experiment is carried out, and the efficiency of the method is described.