数据驱动的终端迭代学习控制方法及不确定性问题研究

发布时间：2018-10-26 11:50

【摘要】：本论文主要研究了离散时间系统的几种新型数据驱动终端迭代学习控制(TILC)方法及其不确定性问题,论文的主要工作和创新点如下:第一,针对一类输入为常数的单输入单输出(SISO)非线性离散时间系统,提出了数据驱动最优终端迭代学习控制的高阶算法(H-DDOTILC),可利用系统先前多次重复运行过程的I/O数据构造新的控制输入,提高了系统沿迭代轴的收敛性能和瞬态响应。同时,将该方法推广到了控制输入随时间变化的多输入多输出(MIMO)非线性离散时间系统。理论分析和仿真研究均说明了方法有效性。第二,为了解决终端迭代学习控制方法在实际应用中,被控系统可能出现的初始条件随迭代次数随机变化和期望参考点在迭代过程中变化的问题,本论文提出了基于RBF神经网络初始状态学习的终端迭代学习控制方法,在系统初始条件随机变化和期望参考点迭代变化的情况下,可以保证终端跟踪误差沿迭代轴收敛到任意指定的区域。值得说明的是,尽管所提出方法的设计和分析过程是针对线性系统的,但控制器中不显含任何关于被控系统模型的信息,而是仅依赖系统的输入输出数据,因此所提出的方法是数据驱动的,可以直接应用于一般非线性系统,且不需要建模过程。第三,当系统的初始状态不可测时,进一步提出了初始条件迭代动态补偿的数据驱动的最优终端迭代学习控制方法,当存在初始状态扰动和系统外部干扰时,可以提高系统的鲁棒性和自适应性。该方法放宽了传统ILC需要相同初始条件假设的限制,使控制器设计更具有一般性;鲁棒性分析和仿真研究均证明了控制系统抗干扰的能力。
[Abstract]:In this paper, several new data driven terminal iterative learning control (TILC) methods for discrete time systems and their uncertainties are studied. The main work and innovations of this paper are as follows: first, For a class of single-input single-output (SISO) nonlinear discrete-time systems with constant input, a high-order algorithm for data-driven optimal terminal iterative learning control (H-DDOTILC) is proposed. A new control input can be constructed by using the I / O data of the previous repeated operation process of the system, which improves the convergence performance and transient response of the system along the iterative axis. At the same time, the method is extended to the multi-input multi-output (MIMO) nonlinear discrete-time systems which control the change of input with time. Theoretical analysis and simulation results show the effectiveness of the method. Secondly, in order to solve the problem that the initial conditions of the controlled system change randomly with the iteration times and the expected reference points change in the iterative process in the practical application of the terminal iterative learning control method, In this paper, a terminal iterative learning control method based on initial state learning of RBF neural network is proposed. The terminal tracking error can be guaranteed to converge to any specified region along the iterative axis. It is worth noting that although the design and analysis process of the proposed method is directed at linear systems, the controller does not contain any information about the model of the controlled system, but relies only on the input and output data of the system. Therefore, the proposed method is data-driven and can be directly applied to general nonlinear systems without the need of modeling process. Thirdly, when the initial state of the system is undetectable, a data-driven optimal terminal iterative learning control method with initial condition iterative dynamic compensation is proposed. When there are initial state disturbances and external disturbances of the system, It can improve the robustness and adaptability of the system. This method relaxes the limitation that the traditional ILC needs the same initial condition assumption, and makes the controller design more general, and the robustness analysis and simulation research prove the anti-jamming ability of the control system.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP13

【相似文献】