Self-Alignment MEMS IMU Method Based on the Rotation Modulation Technique on a Swing Base

Haifeng Xing, Zhiyong Chen, Haotian Yang, Chengbin Wang, Zhihui Lin, Meifeng Guo
2018 Sensors  
The micro-electro-mechanical-system (MEMS) inertial measurement unit (IMU) has been widely used in the field of inertial navigation due to its small size, low cost, and light weight, but aligning MEMS IMUs remains a challenge for researchers. MEMS IMUs have been conventionally aligned on a static base, requiring other sensors, such as magnetometers or satellites, to provide auxiliary information, which limits its application range to some extent. Therefore, improving the alignment accuracy of
more » ... MS IMU as much as possible under swing conditions is of considerable value. This paper proposes an alignment method based on the rotation modulation technique (RMT), which is completely self-aligned, unlike the existing alignment techniques. The effect of the inertial sensor errors is mitigated by rotating the IMU. Then, inertial frame-based alignment using the rotation modulation technique (RMT-IFBA) achieved coarse alignment on the swing base. The strong tracking filter (STF) further improved the alignment accuracy. The performance of the proposed method was validated with a physical experiment, and the results of the alignment showed that the standard deviations of pitch, roll, and heading angle were 0.0140 • , 0.0097 • , and 0.91 • , respectively, which verified the practicality and efficacy of the proposed method for the self-alignment of the MEMS IMU on a swing base. relationship between the body frame and the navigation frame, meaning the transformation matrix from the body frame to navigation frame must be solved. For rotary INSs, the alignment determines the attitude relationship between the IMU frame and the navigation frame, and then the transformation matrix from the body frame to the navigation frame can be calculated according to the rotation angle. Only when the alignment is completed can the horizontal attitude and heading angle be obtained, and then navigation can be performed. Therefore, the alignment is the prerequisite for the INS to determine attitude and navigate. High end INSs can self-align without other sensors in a less disturbed environment by applying conventional analytical coarse alignment and using the Kalman filter for fine alignment [8, 9] . However, static-base alignment has some limitations. When the INS has to be aligned when on a swing base, the measurement of the rate of earth rotation and gravity information are seriously affected. A swing condition means that the base of the INS is not fixed; instead, it exists in a disturbing circumstance, such as a ship during mooring, or a vehicle affected by a gust of wind or engine vibration. In the swing condition, the base generally does not move or the velocity changes very slowly, but the attitude changes more obviously. Since the earth's rotation rate is too small, and the magnitude of the angular rate caused by the disturbance may be several orders higher than the rotation rate of the earth [10], the conventional analytic alignment method cannot be used. As such, researchers are focusing on alignment under swing base conditions by separating the disturbance information [11] [12] [13] . Under swing conditions, the inertial frame-based alignment (IFBA) algorithm is often used. This algorithm introduces a body inertial frame and uses the earth's rotation rate, alignment time, and the direction change of the local gravity in inertial space to calculate the north of the earth [14, 15] . Previous alignment methods [11-15] used high end INS, such as fiber optic gyro (FOG) strapdown inertial navigation system (SINS), attaining high alignment accuracy. However, the alignment of MEMS IMUs cannot be directly achieved using these methods because MEMS inertial sensors have a relatively large constant bias, which can seriously affect the alignment accuracy. Directly separating the earth's rotation rate from the outputs is difficult because of the constant drift and noise of MEMS gyros, so the alignment of MEMS INS is always based on information from high end INS, which is the transfer alignment [16] [17] [18] , or by using external aids such as magnetometers or satellite receivers to obtain the heading angle in the static condition [19] [20] [21] [22] . The drawback of the latter method is that the magnetometer is prone to disturbances from electrical and metal surroundings, and satellite signals are not available in urban canyons, or indoor or underwater environments. Some studies researched the alignment of MEMS IMU for UAVs or AUVs assisted by global positioning systems (GPS) or the Doppler velocity log (DVL). Then, a non-linear filter method, such as an unscented Kalman filter (UKF) or a particle filter (PF), was applied to estimate the attitude errors. This method requires a large amount of computation, and the heading accuracy is generally only within a few degrees [23] [24] [25] . Some articles have been published about using MEMS inertial sensors to find north using single-axis rotation, which is the principle upon which the MEMS gyro north finder is based [26] [27] [28] . Applying RMT to the MEMS gyro north finder provides a new method for the initial alignment of the MEMS IMU. The MEMS gyro north finder can only find north with a static base; when the base is swinging or vibrating, the error is large and the outputs are not available because the MEMS gyro north finder usually has only one single-axis gyro and one or two accelerometers [4] . A single axis gyro cannot track the changes in the attitude angles of the swing base, but the MEMS IMU has three gyros and three accelerometers, so theoretically, the attitude change can be tracked. The MEMS gyro has a relatively large constant drift, but the equivalent drift of the east gyro has the greatest impact on heading accuracy according to the alignment accuracy limit formula [29, 30] . The RMT mitigates or even eliminates the equivalent east or north error of the inertial sensors, so aligning the MEMS IMU under swing conditions is possible. By rotating the IMU using a pre-defined rotation strategy, rather than using external assisting information, the errors of the inertial sensors can be mitigated, eventually improving the IMU performance [5] . For example, the inertial sensor bias can be modulated into a periodic signal by
doi:10.3390/s18041178 pmid:29649150 pmcid:PMC5948697 fatcat:pff3jjmnhva7rmdnxky2p5pe3i