衝突回避ニューラルネットワークシステム : 教師データ収集分析支援システムの構築

概要

論文の詳細を見る
The authors had reported a circumstantial judgment hierarchical neural networks as an initial step, an approach to solving of collision avoidance problem based on a new concept. They pointed out the bottlenecks during collision avoidance neural network system developing. Most of the bottlenecks are concerned with acquisition of ship operators' decision know-how. In order to overcome the bottlenecks, the authors have built a supervised training data acquisition support system. The support system is composed of three function blocks, which are data acquisition, data store and data analysis. And these function blocks make use of computer graphic technology, data-base management technology and knowledge base system technology, respectively. The support system is written in C-language, using GKS graphic software and Shell of UNIX operating system, and run on an engineering workstation. According to a set of evaluation criteria, the support system is evaluated by a ship operator and system developers. And satisfying evaluation results are obtained. Further investigation subjects which establish a neural network topology decision support tool system are pointed out.
社団法人日本航海学会の論文
1992-09-25