Performance Analysis and Comparison of Two Deep RNNs in MEMS Gyroscope Raw Signals Processing

Zhenshu Zhu ； Yuming Bo ； Qingyuan Xia

《Journal of Aeronautics, Astronautics and Aviation》 51卷4期 Vol.51 No.4 (2019/12) Pp. 437-448

https://doi.org/10.6125/JoAAA.201912_51(4).09

摘要

Global Navigation Satellite System (GNSS) has been a feasible and flexible apparatus for providing Positioning, Navigation and Timing (PNT) information globally. In GNSS, the navigation satellites broadcast the signals to the earth, and the user receives the signals for PNT information determination. However, a standalone GNSS is not sufficient to construct a seamless navigation system, especially in some signal challenging environments. Thus, GNSS is always integrated with an Inertial Navigation System (INS) for providing reliable PNT information, since the INS is able to provide moderate PNT information during short time. Micro-Electro-Mechanical IMU (MEMS IMU) is popular in the navigation community, due to its low cost, smaller volume and less power consumption. However, the MEMS IMU experiences complicated noise, which contributes the dramatically errors divergence in navigation solutions. For solving the problem and improving the navigation solution accuracy, this paper introduced Artificial Intelligence methods for addressing this issue. Specifically, deep recurrent neural networks (DRNN) gained excellent performance in processing time series. Inspired by this, this paper firstly employed a deep Gated Recurrent Unit - Recurrent Neural Networks (GRU-RNN) to model the noise with the aim to improve the accuracy of the navigation solutions. Two different MEMS gyroscopes (MSI3200, STIM300) from two different companies were employed in the experiments for testing the evaluating the proposed GRU-RNN, and a Long Short Term Memory Recurrent Neural Networks (LSTM-RNN) was also employed for comparing with the GRU-RNN. Following conclusions were drawn according to the results: 1) the employed GRU-RNN and LSTM-RNN were both effective for MSI3200 and STIM300 gyroscope raw signals processing. The results showed that the standard deviation (STD) of the noise decreased by 27.1%, 36.1% and 51.1% in MSI3200, and 37.5%, 60.0% and 57.1% in STIM300 dataset after processed by the GRU-RNN. The corresponding three-axis attitude errors decreased by 11.4%, 21.0% and 25.7% in MSI3200 dataset, and 60.0%, 36.8%, and 34.7% in STIM300. 2) Furtherly, GRU-RNN and LSTM-RNN obtained similar performance in both MSI3200 and STIM300 gyroscopes de-nosing. However, they both obtained better performance in STIM300 gyroscopes de-noising.

關鍵字

Global Navigation Satellite System ； MEMS Gyroscope ； Deep Neural Networks ； Gated Recurrent Unit ； Long Short Term Memory ； Position Navigation and Timing