【作者】 许能飞；

【导师】 曹云峰；

【作者基本信息】 南京航空航天大学 ， 导航制导与控制， 2005， 硕士

【摘要】 本课题源于国防重大演示验证项目——微型飞行器演示验证样机研制,它具有很强的型号工程背景。 系统辨识( System Identification)作为现代控制论和信号处理的重要内容,是近几十年发展起来的一门学科,它研究的基本问题是如何通过实验数据来建立控制对象的数学模型。本文主要研究有关微型飞行器辨识建模方法及其实现途径。 文中首先介绍了国内外微型飞行器的研究现状以及相关的系统辨识的基础知识;接着探讨了微型飞行器实验建模的方案,包括数据采集装置组成、角速率陀螺等部件的安装和标定误差分析。 课题接着研究了微型飞行器辨识建模的试验设计过程,根据固定翼微型飞行器的气动性能,将微型飞行器建模分为纵向和横向两个通道,并在试验信号的输入输出设计,野值点识别、剔除与补正等方面进行了较为深入的探讨。 最后,本文采用两种辨识方法——无偏最小二乘法和神经网络方法分别辨识出微型飞行器的参数模型和非参数模型,并用交叉检验法对这两个模型进行验证。 研究结果表明,通过系统辨识的方法建立微型飞行器的模型在工程上是可行的。更多还原

【Abstract】 As an important content of modern control theory and signal process, system identification is a developed subject in several decades recently. It focuses on how to find a control mathematics model for the object of experiment using the data of experiment. The thesis studies the modeling method of MAV using system identification. Firstly, the paper introduces the background of abroad and domestic about Micro Air Vehicles. The basic knowledge of system identification is also introduced. Then the hardware design for modeling of MAV is discussed, it include the structure of data collect device, effect of errors in installation on calibration of angle velocity gyroscope etc. In follow, the paper discusses the experiment design of MAV. Based on the analysis of the proper aerodynamic characteristics of fixed-wing MAV, we can build the longitudinal model and the lateral model for MAV. It deeply discusses the experiment signal design of input and output, identification and elimination of outliers in measurement data. Two identification methods are used in the paper. They are Unbiased least square estimate method and neural network method. The correctness of the model for MAV will be proved by cross test method of the experiment data. The result indicates that the method of system identification can be used for MAV.更多还原