2足歩行運動を生成する神経系構造の自律的獲得

概要

論文の詳細を見る
A model to generate autonomous bipedal walking is proposed; a walking motion is generated through the mutual entrainment between neural oscillations and the pendulous movement of body dynamics. However, in the past model, the initial structure of neural systems had to be determined by trial and error. In this paper, based on the process of human motor training, the structure and parameters of the neural system were determined autonomously. The structure of the neural network and the weights of the connections were encoded by a length-variable chromosome in genetic algorithms. They were searched so as to acquire a walking pattern gradually through some stages from supported walking (i.e., infant walking aided by parents) to independent walking. As a result, we could autonomously acquire the structure and parameters for a neural system generating bipedal walking without any artificial trial and error. Therefore, we improved the practical use of this movement generating model.
社団法人日本機械学会の論文
1998-09-25