劣性遺伝構造を有するニューロ制御器を用いた段階的GA学習法による多重トレーラの制御系設計法

DOI Web Site Web Site BIBCODE 被引用文献1件 参考文献17件

書誌事項

タイトル別名
  • Control System Design of Multitrailer Using Neurocontrollers with Recessive Gene Structure by Step-up GA Training
  • レッセイ イデン コウゾウ オ ユウスル ニューロ セイギョキ オ モチイタ ダンカイテキ GA ガクシュウホウ ニ ヨル タジュウ トレーラ ノ セイギョケイ セッケイホウ

この論文をさがす

抄録

In a previous study, we proposed a step-up training method for the multitrailer truck control system using neurocontrollers (NCs) evolved by a genetic algorithm (GA) and showed its efficiency. However, the method does not enable the training of NCs for a five-trailer connected truck system. In this paper, we present a new version of the step-up training method that enables the training of NCs for a five-trailer connected truck system. The proposed method is as follows: First, NCs are trained only to avoid the “jackknife phenomenon". Second, NCs are trained for minimizing squared errors starting from easy initial configurations. Finally, NCs are trained for minimizing the squared errors starting from more difficult initial configurations. The difficulty of training steps increases gradually. To improve training performance, we applied a recessive gene model to network weight coding and genetic operations. In this study, we applied the recessive gene model to the classic exclusive-or (XOR) training problem and showed its convergence performance. The GA training of NCs with the recessive gene model maintains diversity in the population and avoids evolutionary stagnation. Simulation shows that NCs with the recessive gene model and the proposed step-up training method are useful in the controller design of the multitrailer system.

収録刊行物

被引用文献 (1)*注記

もっと見る

参考文献 (17)*注記

もっと見る

関連プロジェクト

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ