Neural network sliding mode control method based on error conversion function

Abstract

The invention discloses a neural network sliding mode control method based on an error conversion function. The neural network sliding mode control method comprises the following steps of: acquiring various parameter matrixes of a micro gyroscope and a designed sliding mode surface; adopting a hyperbolic tangent function as input of an RBF neural network to select a center and a base width on thebasis of a tracking error and the sliding mode surface, then estimating an interference upper bound, and designing a micro-gyroscope control law according to the sliding mode surface and an estimatedinterference upper bound parameter matrix; and finally realizing accurate estimation of spring parameters. The neural network sliding mode control method can guarantee that the input of the RBF neuralnetwork is within a determined range, then the proper center and base width of the network are selected, estimation of an interference upper bound parameter matrix is completed, self-adaptive adjustment of the weight is completed by designing a neural network weight self-adaptive rule, the stability of the system is guaranteed, and the measurement precision of the gyroscope is improved.

Description

technical field [0001] The invention relates to the field of automatic control systems, in particular to a neural network sliding mode control method based on an error conversion function. Background technique [0002] MEMS gyroscopes are commonly used sensors for measuring angular velocity. Mainly used in navigation, mobile phones, quadcopters and other occasions. The working principle of the gyroscope is based on the inertial effect of the proof mass caused by the Coriolis force. When there is an angular velocity input, a Coriolis force that is proportional to the input angular velocity and perpendicular to the angular velocity direction and the initial motion direction will be generated on the micro gyroscope. By detecting the displacement caused by the Coriolis force, and after a series of processing such as demodulation, amplification, and filtering, the required angular velocity signal can be obtained. [0003] Usually, in the gyro control system, due to the uncerta...

AI Technical Summary

Problems solved by technology

However, the approximation performance of the neural network is affected by the center and base width, and the values ​​of the center and base width need to be within the effective input mapping range. Usually, the RBF neural network uses the error signal as the input of the network, but due to the error signal The range is uncertain, so the values ​​of the base width and center may be inappropriate, affecting the approximation effect of the RBF network

Images