Iterative Learning Identification And Control For Non-Affine Nonlinear Systems

Iterative learning control (ILC) algorithm is suitable for repetitive systems over a finite interval. In order to follow the desired trajectory completely, the system input is updated from the previous input experiences. The major feature of ILC algorithms is that the exact system model is not necessary. Most of ILC researches focused on linear systems and affine nonlinear systems. On the base of contraction mapping analysis, this thesis presents nonlinear ILC algorithms and iterative learning identification (ILI) algorithms for non-affine nonlinear systems. This thesis mainly covers the following aspects.Firstly, two nonlinear ILC algorithms are presented for the non-affine nonlinear continuous systems, namely the Newton-type ILC algorithm and the Secant-type ILC algorithm. Sufficient conditions are given for the convergence analysis of the two algorithms.Secondly, the thesis extends the Newton-type and the Secant-type ILC algorithms to non-affine nonlinear discrete time systems and non-affine nonlinear systems with time delay. The fully-saturated learning algorithm is proposed to enhance the system performance, of which the sufficient conditions for the convergence analysis are derived.Thirdly, the problem of initial repositioning is discussed for non-affine nonlinear continuous systems. The existing researches often assumed that the initial state is reset to the desired one without repositioning errors or with known errors at the beginning of each trail. In fact, it is difficult to achieve the accurate repositoning. In this thesis, initial state learning method is proposed which relaxes the initial repositioning condition. The sufficient conditions for the convergence analysis are obtained for open-loop ILC algorithm and closed-loop ILC algorithm and open-closed-loop ILC algorithm.Forthly, the ILI algorithm is discussed for non-affine nonlinear continuous system with unknown parameters. Using ILC synthesis methods, this thesis present the basic P-type ILI and nonlinear ILI, as well as the sufficient condition for their convergence ananlysis respectively. The ILI scheme is applied to recover the masked information signal for chaotic secure communication systems carrying the information signal over a finite time interval. The nonlinear N-shift cipher is established to improve the communication security.