A Hybrid Model for Random Drift Compensation of 3-axis MEMS Gyroscope

In this paper, a hybrid method of random drift compensation is proposed for 3-axis MEMS gyroscope based on auto regressive (AR) model and long short-term memory neural networks (LSTM). AR was used to model the linear portion of the random drift data, while the LSTM model was chosen to process the nonlinear data in the gyro. According to experimental findings, the hybrid model has better noise reduction after filtering than it did before, and it can successfully reduce drift error by up to 87%.