A Combined Method for MEMS Gyroscope Error Compensation Using a Long Short-Term Memory Network and Kalman Filter in...

Area of Science:

• Engineering
• Signal Processing
• Machine Learning

Background:

• Micro-electro-mechanical-system (MEMS) gyroscopes are crucial for navigation and attitude control but suffer performance degradation due to random vibrations.
• Existing error compensation methods often fall short in mitigating the complex effects of vibration on gyroscope data.

Purpose of the Study:

• To develop and validate an advanced error compensation method for MEMS gyroscopes operating in random vibration environments.
• To enhance the accuracy and reliability of MEMS gyroscope measurements through a hybrid approach.

Main Methods:

• A novel combination of a Long Short-Term Memory (LSTM) network and a Kalman Filter (KF) was proposed for MEMS gyroscope error compensation.
• Kalman filter parameters were iteratively optimized using a Kalman smoother and the Expectation-Maximization (EM) algorithm.
• A linear random vibration test was conducted to collect MEMS gyroscope data for validation.

Main Results:

• The proposed LSTM-KF method significantly reduced the standard deviation (STD) of gyroscope error compared to BiLSTM and EM-KF methods (up to 51.58% and 31.92% for x-axis data, respectively).
• Performance improvements were also observed for z-axis data, with STD reductions of 29.19% and 12.75% compared to BiLSTM and EM-KF.
• Compared to the BiLSTM-ARMA-KF method, the proposed approach reduced STD by 46.54% (x-axis) and 22.30% (z-axis), yielding smoother output.

Conclusions:

• The combined LSTM-KF method effectively compensates for MEMS gyroscope errors caused by random vibrations.
• The iterative optimization of KF parameters using EM algorithm and Kalman smoother enhances the robustness and accuracy of the compensation.
• This approach offers a superior solution for improving MEMS gyroscope performance in challenging vibration conditions.