Research On Fractional Order Iterative Learning Identification And Control Algorithms

As a hot field of intelligent control,iterative learning control(ILC)is a method of tracking control for systems that work in a repetitive mode,which is the com-bination of biological motion system and automatic control.For dynamic systems running on a finite time interval,iterative learning identification methods can be used to identify systems' time-varying and time-invariant parameters.With the increased system complexity,the introduction of fractional calculus into iterative learning i-dentification and control is an effective method to solve the problem of the key pa-rameters identification of complex systems.On the other hand,it can also be used to improve the convergence speed,robustness and adaptive ability of iterative learning control algorithms.In this dissertation,novel iterative learning identification methods are studied to deal with the identification of fractional order and initialization response.A novel iterative learning control against initial shifts is considered for tracking control of a class of discrete nonlinear system.Meanwhile,fractional iterative learning control strategies are also studied for linear time-varying systems.Main works and contri-butions in the dissertation are summarized as follows.1.Parameters identification problem is studied for a class of fractional nonlin-ear systems.Based on commensurate fractional order Hammerstein transfer func-tion,an iterative identification framework is presented in time domain.In order to identify each part of the identification framework,an adaptive iterative learning i-dentification method is applied to fractional order identification.Then least square method or instrumental variable method can be used to identify systems linear and nonlinear parameters.Numerical simulations validate the proposed algorithms.2.There are few results addressed the problem of fractional order system with non-zero history function.The fractional order system with non-zero history func-tion belongs to the initialization issue of fractional order differential equations.In order to solve the identification problem of initialization function,this dissertation presents a fractional order iterative learning identification method based on initial-ized fractional order linear time invariant system state-space model.And then,the problem of fractional order initialization function identification is drawn with partial system information before present moment.This study fills the blank of some aspect of truly fractional order system identification.Numerical simulations are performed for linear time invariant systems,and the results validate the algorithm.3.Considering the tracking control of a class of time-varying systems,we present two novel fractional order iterative learning control schemes.By rigorous analysis,a fractional-order DQ-type ILC updating law with initial state learning and high-pass property is proposed.We also show that the proposed learning scheme can be applied to the motion control of robot manipulators under some reasonable conditions.Moreover,a novel P D?-type iterative learning tracking strategy has been presented to enhance the convergence speed.Finally,the efficiency of the methods are verified by numerical simulation.4.Widely utilized in civil and military fields,nonholonomic mobile robot-s(NMR)have gained significantly increasing research interests over the past few decades.Considering the precise trajectory tracking control of NMR,a novel feedback-aided discrete-time iterative learning control framework is presented.Its main fea-ture is its ability to track the desired trajectory with biased initial condition.The convergence of the proposed algorithm for achieving the desired trajectory over a specified interval is proven theoretically.Compared with traditional ILC method,simulation results validate the effectiveness of the proposed scheme.