遠隔操作における操作入力および見まね学習による経路生成との融合による把持支援手法

Abstract

This paper proposes an imitation learning-based approach to assistive teleoperation for a reaching task with multiple objects. First, according to the approaching velocity of the user's input, we estimate the user's intention of the target object. Then, by integrating user input and the output of the demonstrated trajectroy via Dynamic Movement Primitives, we generate the robot endeffector trajectory that will converge to the position of the target object. We present experimental results of an object reaching task to demonstrate the effectiveness of the proposed method in a simulated robot teleoperation system.