欠驱动非线性系统控制问题的研究

发布时间：2018-05-12 18:23

本文选题：单神经元PID + 反馈补偿　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：欠驱动系统指的是驱动维数少于自由度维数的系统,区别于二者相等的全驱动系统。在控制理论中,二者均是控制系统的重要分类。欠驱动系统具有许多诸如节省能源、节约成本和提高自由度等优点,同时有些特定的工作环境下工作或者有特定功能的机械装置本身就需要使用欠驱动结构,而且在某些全驱动机构发生故障的之后,全驱动系统也可能变为欠驱动系统,所以欠驱动系统的应用越来越广泛,研究成果也越来越多。但是由于其欠驱动的特性,在控制中的难度相对较大,在控制过程中,既要让全驱动部分得到控制,又要让欠驱动部分能够一直保持平衡,这为控制器的设计增加了难度。因而,在这方面的研究和关注一直很多。为了拓展欠驱动系统的研究成果,针对欠驱动系统开发更多控制方法,本文对欠驱动系统进行了研究,主要研究了两种控制方法:第一种方法是单神经元PID反馈补偿控制结构。该方法属于智能控制的范围,选取了单神经元PID作为主体控制器,加入了单神经元PID辨识器,二者分工协作,由单神经元PID辨识器根据实施情况进行在线学习,将每一个周期学习得到的参数传递给单神经元PID控制器,单神经元PID控制器采用该参数作为权值,运用到内部算法之中。由此,既保证了智能控制的优点,又可以避免离线学习。此外,在整个控制结构之中,加入了传统的PID控制结构。PID控制器和单神经元PID控制器处在并行的结构,在控制的初期,单神经元PID结构需要时间进行学习,而此时单神经元的参数是不适用于系统的,此时主要以PID控制器为主要控制结构,有效地避免了人工神经网络控制前期的紊乱。同时,在系统受到干扰的时候,该结构也可以增加系统整体的抵抗性,增加鲁棒性。并且使用MATLAB的Simulink模块进行仿真,使用该方法对一个欠驱动系统进行控制,和其他控制方法进行仿真比较,对比各方法的结果,分析优越性。第二种方法是直接参数反馈线性化方法。该方法运用了在处理非线性系统时常用的反馈线性化方法,通过构造控制器,将输出信号通过一定改动反馈回系统之中,将原系统中的非线性成分抵消掉,从而使得系统从非线性转换为线性,从而降低控制难度,可以使用线性控制的相关理论进行控制,从而控制欠驱动系统。二阶动力学系统是常见的系统结构,时常应用于各类机械系统之中。本方法针对一类二阶非线性欠驱动系统进行研究,使用了直接参数化方法,使得得到的算法能够直接针对原系统参数矩阵进行操作,降低了系统维数,减少了运算量,增加了数值稳定性。
[Abstract]:Underactuated system refers to a system with less drive dimension than a degree of freedom, which is different from a full drive system with equal drive dimensions. In control theory, both of them are important classification of control system. Underactuated systems have many advantages, such as saving energy, saving costs and increasing degrees of freedom, while some mechanical devices that work in a particular working environment or have a specific function need to use underactuated structures. After the failure of some full drive mechanism, the full drive system may become underactuated system, so the application of underactuated system is more and more extensive, and the research results are more and more. However, due to its underactuated characteristics, it is relatively difficult in the control process. In the control process, it is necessary to let the full drive part be controlled and the underactuated part to keep the balance all the time, which increases the difficulty of the controller design. As a result, there has been a lot of research and attention in this area. In order to extend the research results of underactuated system and to develop more control methods for underactuated system, two control methods are studied in this paper. The first method is single neuron PID feedback compensation control structure. This method belongs to the scope of intelligent control. The single neuron PID is selected as the main controller, and the single neuron PID identifier is added. The parameters of each cycle are transferred to the single neuron PID controller, and the single neuron PID controller uses this parameter as the weight value and applies it to the internal algorithm. This not only ensures the advantages of intelligent control, but also avoids off-line learning. In addition, in the whole control structure, the traditional PID control structure. Pid controller and single neuron PID controller are in parallel structure. In the early stage of the control, the single neuron PID structure needs time to learn. At this time, the parameters of single neuron are not suitable for the system. In this case, the PID controller is the main control structure, which effectively avoids the disturbance in the early stage of artificial neural network control. At the same time, when the system is disturbed, the structure can increase the overall resistance and robustness of the system. The Simulink module of MATLAB is used to simulate, and the method is used to control an underactuated system. Compared with other control methods, the results of each method are compared and the advantages are analyzed. The second method is direct parameter feedback linearization. In this method, the feedback linearization method is used in dealing with nonlinear systems. By constructing a controller, the output signal is fed back to the system by certain changes, and the nonlinear components in the original system are cancelled out. Thus, the system can be transformed from nonlinear to linear, thus reducing the difficulty of control. The theory of linear control can be used to control the underactuated system. Second-order dynamical system is a common system structure, which is often used in various mechanical systems. In this method, a class of second order nonlinear underactuated systems is studied, and a direct parameterization method is used, which enables the algorithm to operate directly on the original system parameter matrix, thus reducing the system dimension and the computational complexity. Numerical stability is increased.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】