【作者】 陈子印；

【导师】 王宏健；

【作者基本信息】 哈尔滨工程大学 ， 模式识别与智能系统， 2013， 博士

【摘要】 欠驱动水下航行器（underactuated underwater vehicle, UUV）由于未配置横向和垂向的辅助推进器，使得系统的控制输入数目小于系统的运动自由度维数，不满足Brockett定理的必要条件，无法通过设计连续时不变的反馈控制律实现对欠驱动自由度状态的镇定；UUV在水下三维空间的六自由度运动形式，使得模型具有较高的非线性特性和自由度之间的耦合作用，同时受到周围流体作用导致模型参数的不确定性，都成为UUV运动控制技术研究的难点；本文针对欠驱无人水下航行器的变深控制和三维路径跟踪控制方法开展研究，主要进行了以下几方面研究工作：首先，针对UUV变深控制问题，基于反馈增益反步法设计变深控制器，通过合理选择控制器参数消除部分非线性耦合项，简化虚拟控制量的形式，得到具有PID增益调节的控制器形式；结合李雅普诺夫稳定性理论，分析UUV存在建模不精确时闭环系统的鲁棒性，并给出系统跟踪误差的收敛域范围；然后，为解决UUV变深控制中的模型的不确定性，提出基于神经网络的动态面控制方法，通过一阶滤波器获得虚拟控制量的导数信号，避免了直接对虚拟控制量解析求导所导致“计算膨胀”的不足，基于李雅普诺夫稳定性理论设计神经网络控制器实现对模型不确定性的补偿；其次，针对欠驱动UUV三维直线路径跟踪问题，根据离散路径点坐标信息计算得到参数化的空间直线路径方程，提出基于“虚拟向导”跟踪策略建立载体坐标系下的UUV三维路径跟踪误差方程；基于反步法将控制器设计分解为运动学控制回路和动力学控制回路，进而简化控制器设计步骤，基于李雅普诺夫稳定性理论设计鲁棒控制项保证闭环系统的稳定性；仿真实验实现了对分段直线路径的跟踪控制；再次，为简化三维曲线路径跟踪控制器，针对欠驱动UUV三维路径跟踪运动学误差方程，提出基于反馈增益的反步法设计三维曲线路径跟踪控制器，避免了基于“视线法”设计视线角导引律时所导致控制器的复杂形式；通过引入跟踪误差的积分信号，改善控制系统的稳态跟踪性能；最后，针对UUV三维曲线路径跟踪控制问题，区别于动态面中利用数值微分的方式计算虚拟控制量的导数，提出一种基于二阶滤波器的反步设计方法；将虚拟控制作为滤波器的参考输入，通过积分而非微分过程获得虚拟控制量的滤波值及其导数值，避免了对虚拟控制解析求导的繁琐过程，同时减小测量噪声对控制器的影响，基于李雅普诺夫稳定性理论设计滤波误差补偿回路保证滤波器对输入信号的跟踪精度；通过仿真实验验证滤波器和控制器的有效性。更多还原

【Abstract】 Due to the lack of additional thrusters in sway and heave dirctions of underactuatedunderwater vehicle(UUV), the numbers of control inputs for the system is less than thedegrees of freedom related to three-dimensional motion, which doesn’t satisfies the Brocekettnecessary condition, and can’t achieve stabilization through designing contious time-invariantfeedback control law. The model of UUV exhibits severely nonlinearities and couplingcharacteristic between each degree of freedom when moving in three-dimensional space,meanwhile the model also has parameter uncertainties induced by the surroundingviscousfluid, which all brought the difficulties to the research on control technology for UUV.The main research content of this thesis focuses on the solutions for diving control andthree-dimensional path-following control problems, and the main works has been carried outas follows:Firstly, towards the diving control problem of an UUV, backstepping method based onfeedback gain is developed for the controller design, part of nonlinear terms can be eliminatedby properly selecting the controller parameters, which leads a simplification of virtual control,and the resulted controller has PID gain tuning function. Robustness analysis is carried out forthe closed-loop system in the presence of model uncertaintieswith the Lyapunov stabilitytheorem and the boundary of the convergence region of tracking errors for the system is givenout.Secondly, to deal with the model uncertainties in diving control problem for UUV, aneural network based dynamic surface control method is proposed, the derivative of virtualcontrol is obtained through first-order filter, which can avoid the “computation explosive”when directly computing the analystic derivative of virtual control. A neural network adaptivecontroller is designed to compensate the model uncertainties and external disturbance basedon Lyapunov stability theorem.Thirdly, to deal with the three-dimensional straight-line path-following control problemof underactuated unmanned underwater vehicle (UUV). The parameterized path equation iscomputed for space straight-line with the given coordinate informations of discreteway-points. Then, the three-dimensional path tracking error equation is established inbody-fixed frame based on ‘virtual guidance’ for UUV. The controller design process isdivided into kinematic and dynamic two parts with backstepping method, which leads asimplication for the design procedure. Robust control terms are designed based on Lyapunov theorem and asympotic stability can be guaranteed for the closed-loop system. The pathswitching strategy is used to drive the UUV to follow discontinuous straight-line path.Forthly, to simplify the three-dimensional path-following controller, given thethree-dimensional path tracking error equation, the backstepping methodsbased on feedbackgain is used for the controller design which avoids the complex form of the virtual controlwhen designing the guidance laws based on line-of-sight method. Additional integrator termsare introduced to enhance the tracking performance for the control system in steady phase.Finally, be differ from obtaining the derivative of virtual control through numericaldifferentiation process in dynamic surface control, towards the three-dimensionalpath-following problem, The backstepping method based on second-order filter is proposedfor the controller design. The virtual control is selected as the reference input of the filter, thefiltered signal of virtual control and its derivative is obtained through the integration processrather than differentiation process, which avoids the the complexity caused by needing tocompute the derivatives of the virtual control and also reduces the measurement noiseinfluence that act on the controller. Additional tracking error compensation loop is designedbased on the Lyapunov stability theory to gurarantee the tracking precision of the filter for thereference input. The numerical results are performed to illustrate the effectiveness of theproposed filter and control scheme.更多还原