Abstract:

A method is presented for controlling unknown nonlinear MIMO systems by using diagonal recurrent neural network (DRNN). In this controlling scheme, unknown nonlinear plant is identified by a diagonal recurrent neuroidentifier (DRNI), and the sensitivity information of the plant is provided to a diagonal recurrent neurocontroller(NRNC). A dynamic backpropagation algorithm(DBP) is used to train DRNC and DRNI. The proposed approach is applied to a flight control system and the simulation results are included.

Key words: neural network, adaptive control system, dynamic backpropagation algorithm