| Summary: | 博士 === 元智大學 === 電機工程學系 === 106 === This dissertation presents an intelligent control system design using self-evolving and particle swarm optimization (PSO) algorithms. The wavelet cerebellar model articulation controller (WCMAC), type-2 fuzzy neural network (T2FNN) and type-2 brain emotional learning controller (T2BELC) are three kinds of intelligent controller proposed in this disseration. Since determine a network size for a neural network is very important, and it is often difficult to obtain the most suitable network structure. This study develops a self-evolving algorithm for the intelligent control system, that autonomously construct the rule base from the initial empty. To guarantee system stability, adaptive laws for adjusting the parameters of the intelligent controller based on the gradient descent method are proposed. However, in control design, the learning rates of adaptive law are very important and they significantly affect control performance. The PSO method is therefore applied to find the optimal learning rates for the weights in reduction layer and also for the means, the variances of the Gaussian functions in the input membership functions. The stability of system is guaranteed using Lyapunov function approach. Finally, the performance of the proposed system is verified using the numerical simulations of the antilock braking system (ABS), the magnetic levitation system, the inverted pendulum, the chaotic system, nonlinear system identification and the control of time-varying plants.
|
|---|
