Synthesis And Analysis Of Data-driven Robust Learning Control System

A series of Data-Driven Iterative Learning Control(ILC)schemes are proposed in this paper to deal with the problems of incomplete information,variable initial conditions and random disturbances encountered in ILC.The robustness of the system is discussed and analyzed.The strict mathematical demonstration and the simulation experiments are provided.The main innovations of this paper are summarized as follows:First,two data-driven optimal iterative learning control(DDOILC)algorithm based on lifted iteration dynamic linearization(lifted IDL-based)and non-lifted iteration dynamic linearization(non-lifted IDL-based)are proposed respectively for non-linear non-affine systems with measurement data loss.Bernoulli random variable is used to describe the random data loss.The DDOILC scheme based on non-lifted IDL uses more control information and thus has a better control performance than the lifted one.The effectiveness and applicability of the two control methods are verified by theoretical analysis and simulation results,and one can conclude that the data loss will not affect the robustness of the system.Second,two compensatory data-driven iterative learning control(cDDILC)methods are given for linear and non-linear repeatable systems with the random data loss on both input and output sides.Two Bernoulli random variables are used to describe the random data loss at the input and output sides.The virtual linear data model obtained by linearization is used to predict the output of future iterations to compensate the lost output data.The input data at the same sampling time of the previous iteration is used to compensate the lost one.The proposed cDDILC algorithm can compensate the negative impact of data loss on the system and improve the convergence speed of the system.The effectiveness of cDDILC algorithm is verified by theoretical analysis and simulation.Third,an adaptive estimation-based TILC protocol is designed to solve the consensus control problem within finite time of nonlinear discrete-time multi-agent systems with output data loss under directed graph.Similarly,Bernoulli random variable is used to describe the random data loss.The learning control law includes constant learning gain and iterative-time-varying learning gain.The I/O data of the system is used to estimate the partial derivatives,which can further accurately determine the selection range of constant gain.The convergence speed of the system is improved because only the output of the terminal time is considered.In the simulation,the initial conditions of the system varied with the iterations.Theoretical analysis and simulation results verify the effectiveness of the proposed control protocol.