流体内マイクロ移動ロボットに関する研究 : 第3報, 教師なし学習形RBFファジィ・ニューロによる水中マイクロ移動ロボットの泳法の獲得

概要

論文の詳細を見る
This paper deals with a micro mobile robot in water that acquires swimming ability through unsupervised learning. The robot utilizes PZT [Pb (Zr,Ti) O_3] as an actuator and moves fins for propulsion. The motion of a fin is subject to dynamics, thus, acquisition of swimming ability indicates creation of appropriate input to the PZT. We used an unsupervised self-tuning RBF Fuzzy Neuro based on the radial basis function with an insensible region to generate the input signals. The method of learning is based on a genetic algorithm (GA). In other words, a string in the GA stands for a fuzzy system. We show the effectiveness of the method through simulations. We stress that the leaning method does not require any dynamic or physical parameters of the robot. Therefore, the method is effective for leaning in a complicated dynamic system and an unknown environment.
一般社団法人日本機械学会の論文
1996-05-25