项目名称： 自组织递归二型小波模糊神经网络的研究及在微型飞行器姿态控制中的应用

项目编号： No.61502211

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 其他

项目作者： 陈向坚

作者单位： 江苏科技大学

项目金额： 20万元

中文摘要： 针对微型飞行器姿态控制系统中存在不确定性、外界干扰等因素影响控制精度的问题，提出了基于自组织递归二型小波模糊神经网络的微型飞行器姿态混合控制策略。首先，设计自组织递归二型小波模糊神经网络模型结构，该模型结构由前件和后件两部分构成：前件为区间二型模糊集模型，其将每个规则的激活强度反馈到自身构成内反馈回路，后件为小波神经网络；其次，网络初始规则为零，研究适用于自组织递归二型小波模糊神经网络的结构学习算法和后件参数学习算法，在线构造组织递归二型小波模糊神经网络。在此基础上，为了解决模糊集合冗余和模糊规则冗余两个问题，设计相应的模型结构修剪策略；最后，设计基于自组织递归二型小波模糊神经网络的微型飞行器姿态混合控制器，验证其跟踪精度、稳定性、鲁棒性以及实时性。项目的研究成果有利于微型飞行器智能控制问题研究的深化。

中文关键词： 二型小波模糊神经网络；模糊C均值聚类算法；参数自适应学习；微型飞行器姿态控制

英文摘要： Hybrid control strategy based on self-organizing recurrent type 2 wavelet neural networks is proposed to handle Micro Aircraft Vehicle attitude control system which mastered the nominal system mostly, but the nonlinearity, uncertainty, external disturbances are unknown. Firstly, design self-organizing recurrent type 2 wavelet neural network model structure, which is composed of two parts: the antecedent parts take the type-II fuzzy-set model and form the feedback-loop internally by feeding the acting strength of each rule, using an algorithm of gradient-descent method for parameter learning; the consequent parts take the wavelet neural network model; secondly, the initial rules is zero, structure learning algorithm and the parameter learning algorithm are applied to self-organizing recurrent type 2 wavelet neural network, and then the model is structured on-line. On the basis, design the corresponding model structure pruning strategy in order to solve the problems of fuzzy set redundancy and fuzzy rule redundancy; finally, design hybrid control strategy for attitude control of MAV based on self-organizing recurrent type 2 wavelet neural network, verify its tracking precision, stability, robustness and real-time performance. The results of this research is helpful for the deepening of research problems in intelligent control of MAV

英文关键词： type 2 recurrent wavelet neural network;fuzzy c-means cluster algorithm;parameters adaptive learning;attitude control of MAV