適応トラスの動作規範獲得 : モジュール形ニューラルネットによる実現 (<小特集>ファジィ/ニューラルネットのダイナミックスと制御への応用)

概要

論文の詳細を見る
Motion of the adaptive truss brings about a change in its geometry and mechanical properties, and it is evaluated from multiple aspects. Some aspects are definite and mathematical, while others are hazy or indefinite. They conflict, in general, and the overall criterion is the result of a compromise among them. Although definition of the overall criterion is critical in motion planning of the adaptive truss, the manner of compromise to define it is vague and is conventionally accomplished through intuition. This article discusses criterion acquisition for optimal motion planning by means of the multilayered neural network. The modular representation of the required neural network is proposed by paying attention to the multiple aspects of evaluation of the motion and the structural formation of the adaptive truss. The feasibility of the proposed approach is demonstrated through simulational case studies.
一般社団法人日本機械学会の論文
1992-06-25