Adaptive Recursive Terminal Sliding Mode Control for MEMS Gyros Using Improved Neural Network With Constrained Input Mapping

To drive the gyroscope into a command vibration mode, and to simultaneously estimate the angular velocity, this paper proposes an improved fully adjusted neural network-based adaptive recursive terminal sliding mode control for Z-axis MEMS gyroscopes. Firstly, a recursive terminal sliding mode control method is proposed to drive the gyroscope into a desired vibration mode, and the gyro tracking error can converge to 0 in a finite time. Then, the time-varying angular velocity in the gyro system is estimated using a modified neural network, in which a constrained input mapping mechanism is introduced to map the neural network input signal into a predefined range. The constrained input mapping can help to guide the neural network parameter initialization to improve the neural network parameter adjustment efficiency. Meanwhile, the Taylor expansion technique is integrated into the design process to generate adaptive algorithms to tune the center and width in the neural network to enhance the adaptation capacity of the neural network. The performance of the proposed method is evaluated on a Z-axis MEMS gyroscope system through numerical simulations, and results validate the effectiveness of the proposed method.