2A1-D20 Rapid Behavior Adaptation for Human-centered Robots based on Integration of Primitive Confidence on Multi-sensor Element

概要

論文の詳細を見る
This paper presents a rapid learning method of behavior policy for mobile robots teleoperated by an operator. Rapid policy adaptation cannot be achieved when data from every process cycle is used for learning because significant data, which have a major effect on learning, are not differentiated with insignificant data. We propose a method to solve the problem by selecting significant data for the learning based on change in degree of confidence for each sensor element. A small change in the degree of confidence can be regarded as reflecting insignificant data for learning, so that data can be discarded. Accordingly the system can avoid having to store too much experience data and the robot can adapt rapidly to changes in the user's policy. In this paper we discuss the experimental result of an experiment in which significance evaluation is carried out on each proposition of each sensor. And in the experiment user policy changes between 'avoid' and 'approach' on a mobile robot.
一般社団法人日本機械学会の論文