【作者】 朱亮；

【导师】 姜长生；

【作者基本信息】 南京航空航天大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 空天飞行器(ASV)是各国正在大力发展的新型航空航天飞行器,它们在运行中表现出的多任务、多工作模式、大范围高速机动等特点使得控制系统设计成为一项极具挑战的研究课题。围绕这一基础科学问题,本文在空天飞行器建模与分析、不确定非线性系统控制和自主控制系统设计三个方面开展了较为深入的研究。首先,根据国内外公开发表的文献资料建立起ASV超声速和高超声速飞行条件下6自由度数学模型。该模型包含完整的动力学方程和运动学方程,其中气动力系数和力矩系数是迎角,马赫数及气动舵面偏角的函数,发动机模型为吸气发动机和变推力火箭发动机的组合推进装置,飞行器的质心、惯性矩是飞行器质量的时变函数。开环分析表明整个模型能够体现出ASV复杂的非线性、耦合性以及快速时变性等特点,具有一定的代表性,可以满足未来ASV先进制导和控制等问题的理论研究和仿真验证需要。其次,基于此平台研究了ASV的飞行控制系统设计问题。轨迹线性化控制(TLC)是一种新颖有效的非线性跟踪和解耦控制方法,文中首先对它的设计思想和理论基础进行回顾。然后根据奇异摄动理论,将ASV的飞行控制系统分成快慢回路,并分别为它们设计TLC控制器。最后在ASV高超声速飞行条件下进行仿真验证,以检验该控制系统的有效性和鲁棒性。接着,通过理论分析找出系统中存在的不确定对当前TLC方法产生不利影响的机理,指出随着不确定的增大,TLC方法性能会不断降低直至失效。为了解决这个问题,本文基于补偿思想提出一种新的TLC控制策略和控制结构,并首先利用非线性干扰观测器对于不确定的估计能力加以实现,给出一种基于非线性干扰观测器的轨迹线性化控制新方法。基于Lyapunov理论证明出闭环系统所有误差信号均以指数形式收敛至零。数值例子和ASV的仿真验证结果表明,新方法不仅有效而且能够大大提高不确定条件下TLC的控制性能和鲁棒性。随后,利用神经网络对未知非线性函数的逼近能力,提出一种新的鲁棒自适应轨迹线性化(RATLC)控制结构。首先利用径向基神经网络来实现这一控制策略,设计出神经网络自适应调节律,采用Lyapunov方法严格证明出在自适应调节律作用下闭环系统所有误差信号最终有界。最后将所得结果推广至单隐层神经网络。高精度地逼近系统中存在的不确定,可大大提高控制效果,为此本文研究了两种新的神经网络干扰观测器技术。该技术具有广泛的适用性,可与已有的很多控制方法结合来提高它们在不确定条件下的控制性能,因此该研究结果对于发展不确定非线性系统逼近策略具有重要的借鉴意义。在此基础上,提出新的基于神经网络干扰观测器的RATLC控制结构,设计出合适的自适应调节律并利用Lyapunov稳定性理论进行了严格的证明。上述所有RATLC方案均在ASV高超声速飞行条件下进行仿真验证,仿真结果表明这些控制算法不仅有效,而且可以获得非常优异的控制性能。因此RATLC策略使得当前TLC方法获得很大的发展。最后,研究了ASV自主式控制系统体系结构的设计问题。通过对多智能体技术的简单回顾指出该技术用于ASV自主控制系统设计的可行性,对ASV自主控制系统做出需求分析,采用多智能体技术设计出ASV的自主控制系统,并对整个系统的工作原理和各个主体的具体功能做出详细讨论。通过与深空一号自主系统的对比研究,指出新系统在智能性,鲁棒性、移植性和扩展性等方面的优势。更多还原

【Abstract】 The current research in developing next generation reusable flying vehicles is focused on aerospace vehicles (ASVs). The control systems of the ASVs pose several challenges due to their multi-mission profiles、large attitude maneuvers and complicated flight conditions. In this dissertation, three relative problems, i.e. modeling and analysis of a conceptual ASV,uncertain nonlinear system control and autonomous control system design, are studied.First of all, a six degree-of-freedom simulation model of a conceptual ASV is presented, which includes the whole of kinetic equations and motion equations. Aerodynamic force and moment coefficients are given as functions of angle of attack, Mach number and control surface deflections. The propulsion system covers a combination of an air-breathing engine and a variable thrust liquid propellant rocket engine. Rigid-body mass moments of inertia and center of gravity location are functions of time-varying vehicle weight. Open-loop dynamics and stability characteristics demonstrate that the proposed model can be used to allow research, refinement and evaluation of advanced guidance and control methods.Next, this thesis provides the design of a flight control system for the ASV based on Trajectory Linearization Control (TLC) method. The TLC method is a novel nonlinear tracking and decoupling control technique, which can be viewed as the ideal gain-scheduling controller designed at every point on the flight trajectory. In Chapter 3, its background and theoretical basis are reviewed at first. Two control laws are provided to the outer loop and inner loop respectively in terms of singular perturbation theory. Then, a simulation is presented for the ASV under hypersonic cruise condition. The simulation results demonstrate the good performance of the controllers.Theoretical analysis illustrates that the TLC method may exhibit poor performance in the presence of uncertainties. A novel control structure is developed by combining the current TLC method with the well-known compensation idea. A nonlinear disturbance observer enhanced TLC approach is implemented at first. The NDO is used to estimate the uncertainties, and then integrated with the TLC method. Stability and performance analysis of the composite closed-loop system is conducted using Lyapunov’s direct method. By applications to control of a numerical example and the ASV, effectiveness of the proposed method is validated.Then, utilizing the universal approximation property of neural networks, several robust adaptive TLC (RATLC) approaches are established. The NN outputs are introduced to estimate the uncertainties, and robust adaptive terms are used to overcome the reconstruction errors. By Lyapunov’s direct method, rigorous proofs demonstrate that the provided adaptive laws can guarantee ultimate boundedness of all the signals in the integrated system. Furthermore, we expand the idea of fuzzy disturbance observer technique and investigate two types of neural networks disturbance observers (NNDO). Subsequently, more RATLC schemes are obtained through integrating the current TLC with the NNDO outputs. Conditions are derived which guarantee ultimate boundedness of all the errors in the combined system. All the above RATLC algorithms are illustrated for the flight control application of the ASV. Simulation results show that each of the proposed algorithms can significantly improve uncertainty attenuation ability and performance robustness of the current TLC. So our philosophy is a promising way and greatly extends the current TLC method.At Last, architecture of autonomous control system for the ASV is designed based on the multi-agent system (MAS) thechnique. The architecture integrates several autonomous agents and completes the desired mission through coordination and cooperation. By analyzing work principles of the entire system, it can be seen that the provided architecture can improve the autonomy of the ASV, enhance its capability to solve complex problems and satisfy requirements of autonomous operations for long periods of time. Compared with the remote agent architecture of the deep space 1, the system has higher intelligence, better robustness, transplantation and expansibility.更多还原