Iterative Learning Control Method And Analysis For Nonlinear Switched Systems

In this thesis,a series of data-driven ILC strategies are presented for nonlinear nonaffine discrete-time switched systems,and the problems of non-repetitive noise disturbance and random packet loss generated by network channels in ILC are discussed.Now the key research contents of this thesis are exhibited as follows:First,three ILC methods,P-type,PD-type and PID-type,are designed for singleinput single-output(SISO)discrete-time nonlinear non-affine switched systems.This method only utilizes the offline or online input/output(I/O)data of the controlled object,and does not require a mechanism model or an accurate nominal model of the controlled system.Based on the SISO switching system and the proposed ILC method,a rigorous theoretical analysis is carried out using mathematical tools such as state transition technology,lifting matrix technology,and compression mapping.Consistent convergence along the iterative axis can be achieved under the continuous excitation of the proposed ILC strategy.Second,many practical production processes are essentially multiple-input multiple-output(MIMO)nonlinear non-affine switched systems.For such systems,there is a strong coupling between the I/O data,so the control is more complex and more difficult.Under the framework of data-driven design and analysis,the research on the ILC strategy of SISO nonlinear non-affine discrete-time switched system is extended to more complex and general MIMO nonlinear switched systems.The nonlinear switched system is transformed into a mathematical model based on incremental I/O data through dynamic linearization,and then a strict stability analysis is given.The simulations prove that the proposed ILC methods can make the nonlinear switched system achieve the control effect obtained by the theoretical analysis.Third,compared with the conventional control system,the wireless network remote control system has more flexible system design,less wiring,and low power consumption.However,because of the wireless network,the data may have the problem of channel noise.Therefore,The D-type iterative learning control method is further applied to nonlinear non-affine discrete-time switched systems with network channel noise.Fortunately,for network nonlinear switched systems that perform repetitive tasks periodically at fixed time intervals,the iterative learning control method can effectively improve the tracking performance of such systems.Theoretical analysis shows that,in the sense of mathematical expectation,the output error of the switched system can converge to zero along the iterative direction.Fourth,for a nonlinear non-affine discrete-time switched system with data packet loss at both the input and output terminals,consider the network P-type iterative learning control scheme,where the random data packet loss of the system obeys the 0-1 Bernoulli distribution.The control scheme replaces the communication packet loss data with the data successfully captured at the concurrent sampling moment of the latest iteration.Based on the mathematical expectation,the control effect of the network iterative learning control(NILC)algorithm is analyzed using statistical techniques.Under the appropriate learning gain assumption,the analysis shows that the mathematical expectation of the tracking error of the switched system can fully converge asymptotically.The theoretical values and innovations of this thesis are summarized as follows:(1)This thesis conducts a series of studies on nonlinear non-affine switched systems,which are more general than linear systems and nonlinear affine systems;(2)The convergence proof process of the iterative learning control method for discrete-time nonlinear non-affine switched systems does not use any precise model information;(3)The ?-norm is not used in the convergence analysis process,so the theoretical results do not depend on the system running time,which can ensure the good transient performance of the system;(4)Considering the status quo of the rapid development of network remote control technology,this thesis focuses on the problems of channel noise and data packet loss caused by network channels.