A comparison between different error modeling of MEMS applied to GPS/INS integrated systems

Area of Science:

• Navigation Systems Engineering
• Sensor Technology
• Signal Processing

Background:

• Micro-electromechanical systems (MEMS) accelerometers and gyroscopes are crucial for low-cost navigation systems.
• MEMS sensor errors degrade navigation accuracy over time.
• Accurate error modeling is essential for improving navigation system performance.

Purpose of the Study:

• To analyze and model stochastic errors affecting MEMS inertial sensors.
• To develop a compensation method for short-term and long-term error components.
• To assess the proposed error modeling technique within a GPS/INS integration framework.

Main Methods:

• Comparison of autocorrelation, Allan variance (AV), and power spectral density (PSD) techniques for error analysis.
• Development of an inertial sensor error model combining autoregressive (AR) filters and wavelet de-noising.
• Integration of Allan variance, wavelet de-noising, and decomposition level selection for error compensation.
• Augmentation of the Extended Kalman Filter (EKF) in a GPS/INS loosely-coupled strategy with novel error states.

Main Results:

• The proposed method effectively models and compensates for MEMS sensor errors.
• The combined Allan variance and wavelet de-noising approach addresses both high- and low-frequency error components.
• The enhanced EKF demonstrated improved navigation accuracy compared to traditional models during GPS signal blockages.
• Real-world data from urban roadways validated the efficacy of the developed techniques.

Conclusions:

• Accurate stochastic error modeling is vital for low-cost MEMS-based INS.
• The proposed integrated approach of Allan variance and wavelet de-noising offers a robust solution for sensor error compensation.
• This methodology significantly enhances navigation system resilience and accuracy, particularly in challenging environments with intermittent GPS availability.