2013 Fiscal Year Final Research Report

Human Action Model Based on Multiple Agent for Remote Communication

Research Project

Project/Area Number	24700148
Research Category	Grant-in-Aid for Young Scientists (B)
Allocation Type	Multi-year Fund
Research Field	Intelligent informatics
Research Institution	Osaka City University
Principal Investigator	TAKUTO Tomohito 大阪市立大学, 大学院工学研究科, 准教授 (80397695)
Project Period (FY)	2012-04-01 – 2014-03-31
Keywords	強化学習 / データマイニング / 行動予測
Research Abstract	To solve the communication delay via network, the human action model which can replace a human is proposed. The model is structured by the two-layer. The first layer assumes reactive behavior, and it is modeled by the multiple agents. The agent consists of a linear quadratic controller and a linear predictive function. The second layer assumes predictive action with accumulative reward. The second layer indicates commitments to agents on the first layer to act strategical. The proposed human action model is implemented to simple games on PC and the control habit of the human skilled player is modeled well. Furthermore, the acquired model can play the same game instead of the human player. The human operation model is expected to make the communication via internet smooth.