A novel fuzzy B-spline neural network and applied into mobile satcom antenna

This work proposed a novel fuzzy B-Spline neural network (FBSNN), and applied to the control system of mobile satcom antenna (MSA). To communicate continuously, the antenna must always points to the desired satellite during vehicle motion. Therefore, the gimbals pointing must be controlled accurately. The control block diagram is established based on the dynamic model. Due to disturbance and coupling effects that exist in system, adaptive control strategy should be adopted. The proposed FBSNN is applied to approximate nonlinear element and disturbance. Compared with conventional FBSNN, the proposed method not only adjust the output weights, but also has ability to adjust shapes and positions of membership function. The adjustment method is obtained by reconstruction B-Spline function, and the training method is also derived based on the property of B-Spline. Simulations show that the proposed method performs better than fuzzy radial basis function (RBF) neural network and conventional PID control method.