补偿模糊神经网络在模糊规则训练中的应用  

Application of Compensatory FNN in Training of Fuzzy Rules

作　　者：田八林[1] 李华星[1] 张中荃[2] 

机构地区：[1]西北工业大学,陕西西安710072 [2]解放军西安通信学院,陕西西安710106

出　　处：《计算机仿真》2006年第10期11-14,共4页Computer Simulation

摘　　要：目前,自治式水下机器人(Autonomous Underwater Vehicle,AUV)、自动导引驾驶小汽车、轮船等领域应用模糊规则控制已经受到许多人的关注,模糊规则的制定与训练是其中之关键所在,该文将模糊规则控制应用在无人机自由编队飞行控制中。在训练模糊规则过程中,常规的BP神经网络法存在学习速度慢、无法结合专家知识以及容易陷人局部最小等缺点,为了克服上述不足,文中引入了补偿模糊神经网络,它是一个结合了补偿模糊逻辑和神经网络的混合系统,由面向控制和面向决策的神经元组成,其模糊运算采用动态的、全局优化运算,学习速度快、学习过程稳定。将其用于无人机自由编队飞行的模糊控制规则进行训练,结果表明用补偿模糊神经网络对模糊规则的训练效果良好。In recent years, many people pay attention to fuzzy rule and its application in control of AUV（Autonomous Underwater Vehicle）,automatic guided vehicle and ship. Establishing and training of fuzzy rules is the key. Fuzzy rule control is adopted in UAV free formation flight in this paper. The general BP neural network has the shortcomings such as low learning speed, being not able to combine with the expert knowledge and liable to fall into a local minimum point etc. In order to overcome these disadvantages, this paper introduces the Compensatory Fuzzy Neural Network （CFNN）, which combines the compensative fuzzy logic with neural network, and is composed of control oriented cell and decision - making cell. Its fuzzy computation is dynamic and global optimized, therefore its speed is high. Fuzzy control rules of UAV free formation flight are trained by CFNN, and the result indicates that the CFNN is effective in training of fuzzy rules.

关 键 词：模糊控制规则 补偿模糊神经网络 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]