摘要
本文描述了一个简单的基于神经网络的小型帆船自适应舵的控制方案。该方案采用了两个简单的双层神经网络:一个使用在线的BP算法来学习未知的帆船动态;另一个则利用前一个所学到的知识来调整它的连接权系数和偏差权系数以产生控制信号。仿真结果表明,与仔细调整过的PID控制器相比较,神经网络控制器在航向保持上与PID近乎相同,但却具有好得多的鲁棒特性。
A simple adaptive control scheme for a yacht based on neural networks has been described. The design concerns two simple two-layered neural networks. The first one learns the unknown yacht dynamics by using an online common BP algorithm, while the second one uses the current knowledge of the first to adjust its connection weights and bias weight in order to generate the control signal. Simulation results of tests carried out for comparing the NN control with a well tuned PID autopilot show that it is capable to provide a good robust stability even if the NN control can not provide a better performance in terms of track keeping.
出处
《中国造船》
EI
CSCD
北大核心
2004年第1期25-32,共8页
Shipbuilding of China
