Constrained Adaptive Fuzzy Iterative Learning Control

In the practical application, the dynamic characteristics of the controlled systems have many kinds of uncertainties, and using the approximation characteristic of the fuzzy system is a way of solving the related control problems. This paper studies the iterative learning control method for the controlled systems under repetitive operation. While considering initial error for repetitive systems as well as the position, orientation and other constraints in real systems, it is essential for controller to manage initial conditions and constraints. Based on an improved logarithmic barrier Lyapunov function, the error tracking approach is used to design a controller with the performance of constraint. The uniform boundedness of all variables in the closed-loop system is guaranteed, and the constrained system error function is achieved.The main results and achievements of this thesis are as follows:1. Iterative learning control is presented for a class of uncertain systems. The nonparametric uncertainty of system dynamics is handled by combining the robust method that compensates the boundary of uncertain and learning strategy for estimation. It is shown that the system error trajectory can track the expected error trajectory over the predetermined time interval as iteration increases.2. Adaptive fuzzy control is presented for a class of uncertain systems. When the uncertain systems have repetitive operation characteristics, iterative learning method is used to design time-varying fuzzy systems. Direct and indirect adaptive fuzzy controllers are structured. The time-varying fuzzy systems allow parameter estimation time-varying which makes the approximation better. It is shown that the practical tracking error trajectory is ensured to converge to the desired error trajectory as the iteration increases, and kept in the pre-specified region all the time.3. An improved logarithmic barrier Lyapunov function is proposed to design a constrained control algorithm. This BLF is used for all controllers in this paper and it makes sure that the system error trajectory can track the expected error trajectory with the error function constrained during control process. The improved BLF has generally characteristics and the controller is easy to realize.4. Three methods are discussed to deal with the problem of initial error for repetitive systems. Comparing with time-varying boundary layer method and resiving reference trajectory method, the error tracking approach method in this paper is better from the simulation results of the inverted pendulum system. The derivative of the access point in the revised reference trajectory doesn’t match with the derivative of the reference trajectory. The error function of time-varying boundary layer method only can be asymptotically convergent. However, the error tracking approach method doesn’t have those questions, displays the better performance.