Improved BP neural network based active disturbance rejection control for magnetic sensitivity calibration system

Abstract

In the magnetic sensitivity calibration system, the calibration accuracy of inertial sensor is directly related to the control accuracy of the magnetic induction intensity. Since the helmholtz coils in the calibration system have large parameter uncertainties and the magnetic field sensor has some time-delay, the traditional PID controller cannot satisfy the accuracy requirement of the magnetic induction intensity. Therefore, an improved neural network based active disturbance rejection controller (ADRC) is proposed, which utilizes the conjugate gradient algorithm and Fletcher-Reeves linear search method to adjust the parameters of ADRC for achieving the optimal control efforts. Moreover, the extended state observer of ADRC can compensate for the parameter uncertainties and time-delay exactly such that the control accuracy of the magnetic induction intensity can be largely improved. The simulations are conducted to show the effectiveness and superiority of the proposed control algorithm.