四旋翼飞行器的轨迹跟踪自校正预测控制  

Self-tuning Trajectory Tracking Predictive Control for Four Rotor Aircraft

作　　者：刘丽丽[1] 左继红[1] LIU Li-li;ZUO Ji-hong(Rail Transit Intelligent Control Institute,Hunan Railway Professional Technology College,Zhuzhou 412001,China)

机构地区：[1]湖南铁道职业技术学院轨道交通智能控制学院,湖南株洲412001

出　　处：《控制工程》2020年第10期1838-1844,共7页Control Engineering of China

基　　金：湖南省教育厅科学研究项目(19C1211);校级科研创新团队资助(KYTD201703)。

摘　　要：针对四旋翼飞行器动态运行过程中存在的模型不确定、多变量非线性及外界干扰等问题,采用神经网络高斯径向基函数(Radial Basis Function, RBF)逼近线性自回归模型(Auto-regressive Exogenous, ARX)的方法建立该系统的RBF-ARX混合模型,并基于该RBFARX混合模型设计了一种自校正轨迹跟踪多步向前预测控制器。该法首先构建四旋翼飞行器的RBF-ARX模型结构,辨识并优化模型参数,基于此模型多步向前预测对象输出,计算与实际飞行轨迹偏反馈补偿,有效解决耦合或干扰造成的轨迹跟踪偏移。通过实时控制实验数据分析,验证该方法的有效性和可行性。Aiming at the problems of model uncertainty, multivariable nonlinearity and external disturbance in the dynamic operation of four-rotor aircraft, the RBF-ARX hybrid model of the system is established by using the method of neural network Gaussian radial basis function(RBF) approximating linear autoregressive model(ARX). Based on the RBF-ARX hybrid model, an adaptive trajectory tracking multi-step forward predictive controller was designed. Firstly, the RBF-ARX model structure of the four-rotor aircraft was constructed, and the parameters of the model were identified and optimized. Multi-step forward prediction of object output was made based on the model, and feedback compensation was calculated to compensate the deviation of the actual flight trajectory. The validity and feasibility of the method are verified by real time control experiment data analysis.

关 键 词：四旋翼飞行器 神经网络 轨迹跟踪 预测控制 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]