Design of a new kind of RBF neural network based on differential reconstruction

A new kind of RBF neural network based on Fourier progression was studied, and the principium of its approximating unknown function was analyzed. Then it was used in a class of high order system with all unknown control function matrices. The adaptive RBF robust neural controller was designed by using back stepping method. And by adopting the trigonometric function as basis function, the input needn´t be forced to between -1 and 1, and there is no need to choose the centre of basis function. Furthermore, it is possible to make the network more stable and make the selection of simulation parameter more easy due to the introduction of differential reconstruction which increased the damp of the system. Finally, simulation study showed the effectiveness of the proposed method