【作者】 陈向坚；

【导师】 续志军；

【作者基本信息】 中国科学院研究生院（长春光学精密机械与物理研究所） ， 机械电子工程， 2012， 博士

【摘要】 微型飞行器（MicroAircraft Vehicle-MAV）具有研制及生产成本低、携带便捷、噪声小、重量轻、隐蔽性强、后期维护比较容易等诸多优点。本文主要以姿态稳定控制、位置导航控制内外双回路中的内回路（姿态稳定控制回路）为研究对象。由于MAV姿态控制系统受到各种外界扰动及参数变化、建模误差等原因引起的不确定性等因素的影响，本文以自适应控制技术、滑模控制技术以及模糊神经网络等作为理论研究工具，设计稳定性好、鲁棒性强的MAV飞行姿态控制系统。主要做了如下几方面的工作：本文首先分析了MAV的惯性特性，研究了MAV在飞行过程中机体动力与动力矩以及MAV质心的线运动与机体绕质心旋转的运动方程，得出MAV的动力学模型，给出用于姿态控制的基础数学模型。其次，研究了区间二型模糊神经网络的网络结构及学习算法，通过对动态系统辨识的仿真表明其相对于一型模糊神经网络更高的辨识精度。因此，本文基于区间二型模糊神经网络理论，依据对MAV姿态模型的认知程度不同，针对MAV姿态控制系统受到各种外界扰动及参数变化、建模误差等原因引起的不确定性等因素的影响，研究了三种控制器：第一种：微型飞行器姿态的区间二型模糊神经网络鲁棒自适应控制器。该控制器由传统的PD控制器与区间二型模糊神经网络控制器两部分构成。其中，区间二型模糊神经网络控制器由两个区间二型模糊神经网络组成，其一用来在线学习微型飞行器姿态的逆模型，而另一个是前一个的复制，用来在线学习，实时补偿跟踪误差。第二种：微型飞行器姿态的区间二型模糊神经网络滑模自适应控制器。该控制器协调了MAV姿态的自身结构知识及控制知识，采用一个加权因子来结合间接模糊神经网络控制器和直接模糊神经网络控制器，通过输出反馈控制律和自适应律对其进行在线调整网络参数。第三种：微型飞行器姿态的区间二型模糊神经网络间接自适应控制器。该控制器充分利用微型飞行器姿态的标称模型，利用区间二型模糊神经网络用来在线逼近系统不确定性，鲁棒补偿器用来实时补偿区间二型模糊神经网络的逼近误差以及外界扰动。最后，对三种控制器进行仿真对比分析。无论是否考虑不确定性及外界扰动的存在，三种控制器均满足微型飞行器姿态控制的性能要求。然而，从控制精度、响应速度两个方面出发，得出区间二型模糊神经网络间接自适应控制器相对于其它两个控制器来说，是控制MAV飞行姿态的最佳控制方案。更多还原

【Abstract】 MAV (Micro Aircraft Vehicle) has the advantages of low cost, convenient,lightness, small noise, strong concealment, easier maintenance, and so on. MAVcontrol system was divided into two parts: inner loop and outer loop usually, theinner loop represents the attitude stable control; the outer loop represents the positionnavigation control. This paper focuses on the inner control loop (attitude stablecontrol) of MAV control system, that is to say, this paper takes MAV attitude controlsystem as the research object. But, the MAV attitude control system is influenced byvarious kinds of outside interference and uncertainty, which is caused by variousparameters change, modeling error and other factors. So this paper takes adaptivecontrol technology, the sliding mode control technology, and fuzzy neural networktheory as a research tool to design stable and robust attitude control system of MAV.The main works in this dissertation are arranged as follows:Firstly, the inertial characteristics were analysized, the effects of dynamic forcesand moments on airframe, the dynamical and kinematical equations of centroidmovement and encircling the centroid movement of MAV were also studied, themathematical model of MAV attitude was presented for control.Secondly, the structure and learning algorithms of Interval Type II Fuzzy NeuralNetwork were studied, which has higher precision than Type I Fuzzy NeuralNetwork for dynamic system identification. So, this paper based on the Type II Fuzzy Neural Network theory and MAV attitude model cognitive degree, threecontrollers are studied here for ensuring the robustness and stability of MAV attitudecontrol system:The first kind of controller：robust adaptive controller based on Interval Type IIFuzzy Neural Network. This controller was composed of traditional PD controllerand Interval Type II Fuzzy Neural Network controller, which was constituted by twoInterval Type II Fuzzy Neural Networks, one of Interval Type II Fuzzy NeuralNetworks was used to study the inverse model of MAV attitude, the other one wasthe copy of the first one, tuned online and used to compensate tracking error in thereal-time.The second kind of controller：adaptive sliding mode controller based onInterval Type II Fuzzy Neural Network. A weighting factor, which can be adjustedbased on the trade-off between plant knowledge and control knowledge, wasincluded when combining the control efforts of the indirect adaptive Interval Type IIFuzzy Neural Network controller and the direct adaptive Interval Type II FuzzyNeural Network controller. The free parameters of which can be tuned on-line by anoutput feedback control law and adaptive lawsThe third kind of controller：Indirect adaptive controller based on Interval TypeII Fuzzy Neural Network. This controller makes full use of the nominal model ofMAV attitude. The uncertainties of system were approached by Interval Type IIFuzzy Neural Network online, the approximation error of Interval Type II FuzzyNeural Network and external interference were compensated by robust compensator.Finally, the performances of three controllers were compared and analyzed inthe simulation. The performances of three controllers could meet the performancerequirements of the MAV attitude control nomatter whether under the considerationof system uncertainties and external interferences. However, the indirect adaptivecontroller based on Interval Type II Fuzzy Neural Network is the optimal controlscheme relative to the other two controllers from control accuracy aspect andresponse speed aspect.更多还原