Hybrid Intelligent Modeling and Composite Disturbance Filtering for Refined Error Processing of MEMS Gyroscopes

Although microelectromechanical system (MEMS) gyroscopes offer advantages such as miniaturization, low cost, low power consumption, and high integration, their measurement errors significantly limit their performance, particularly in high-precision applications. In this article, we address the measurement errors of MEMS gyroscopes by developing a hybrid intelligent modeling and composite disturbance filtering (CDF) strategy. Specifically, a wavelet transform method is proposed to decompose the error signals into stationary components and nonstationary residuals. On this basis, a hybrid autoregressive moving average/long short-term memory model is established and incorporated into the vehicle dynamics. Since the dynamic and stochastic properties of MEMS gyroscope errors have been sufficiently characterized using this hybrid model, a CDF method is developed that leverages the heterogeneous error characteristics for refined error processing. Moreover, a real-time parameter update law of the error model is embedded into the CDF for better adaptiveness. By virtue of the proposed modeling and filtering scheme, the MEMS gyroscope errors can be effectively represented and rejected, enabling accurate and reliable inertial navigation without the need for additional sensors. Finally, the effectiveness of our method is verified via both static and semiphysical dynamic experiments.