Optimal PID position controller of multi-rocket launcher using improved Elman network

Abstract

Considering the atrocious load property when the rocket is launched, including the huge variation of load inertia, unbalanced torque and the disturbance from the air current impulsion, an optimal PID position controller based on improved Elman network is presented. The context neurons of output layer are added to the original Elman network, and self-feedback gain coefficients are trained as connective weighting value, which could strengthen the adaptive ability of Elman network to the time-varying system. According to the load property of multi-rocket launcher, improved Elman network could adjust the PID parameters online to minify the influence of the change of system parameters and external disturbance. The neural network is trained in online phases and a back-propagation training composes this neural network. Since the total number of nodes is only ten, this system is realized easily by the general microprocessor. During normal operation, the input-output response is sampled and the weighting value is trained multi-times by a back-propagation method at each sample period to accommodate the possible variations in the parameters or load torque. A digital signal processor TMS320F2812 realizes this method. The basic DSP software is used to write C-program, which is compiled by using ANSI-C style function prototypes. Simulation and experimental results show that this method could improve the stability and anti-disturbance ability of multi-rocket launcher position servo system effectively.