2325 ロボット足先の最適着地点列に基づく歩行のニュートラルネットワーク学習 : 教師データの選定に関する検討(機構と設計(1),OS・一般セッション講演)

概要

論文の詳細を見る
The purpose of this study is the development of obstacle avoidance algorithm for quadrupedal walking robot in response to environment. Stepping points are generated by using neural network. Training data for neural network are optimum stepping points generated by using the genetic algorithm based on the mechanism of robot model and the obstacle. A large number of training data are required because obstacles have various sizes and shapes. We compare optimum stepping points generated by using the genetic algorithm to make a search for starting position of stride adjustment. Some numerical results of the stride adjustment are shown for several obstacle environments.
一般社団法人日本機械学会の論文
2011-10-08