Currently, positioning, navigation, and timing information is becoming more and more\nvital for both civil and military applications. Integration of the global navigation satellite system\nand /inertial navigation system is the most popular solution for various carriers or vehicle\npositioning. As is well-known, the global navigation satellite system positioning accuracy will\ndegrade in signal challenging environments. Under this condition, the integration system will fade\nto a standalone inertial navigation system outputting navigation solutions. However, without\nouter aiding, positioning errors of the inertial navigation system diverge quickly due to the noise\ncontained in the raw data of the inertial measurement unit. In particular, the micromechanics\nsystem inertial measurement unit experiences more complex errors due to the manufacturing\ntechnology. To improve the navigation accuracy of inertial navigation systems, one effective\napproach is to model the raw signal noise and suppress it. Commonly, an inertial measurement\nunit is composed of three gyroscopes and three accelerometers, among them, the gyroscopes play\nan important role in the accuracy of the inertial navigation systemâ??s navigation solutions.\nMotivated by this problem, in this paper, an advanced deep recurrent neural network was\nemployed and evaluated in noise modeling of a micromechanics system gyroscope. Specifically, a\ndeep long short term memory recurrent neural network and a deep gated recurrent unitâ??recurrent\nneural network were combined together to construct a two-layer recurrent neural network for\nnoise modeling. In this method, the gyroscope data were treated as a time series, and a real dataset\nfrom a micromechanics system inertial measurement unit was employed in the experiments. The\nresults showed that, compared to the two-layer long short term memory, the three-axis attitude\nerrors of the mixed long short term memoryâ??gated recurrent unit decreased by 7.8%, 20.0%, and\n5.1%. When compared with the two-layer gated recurrent unit, the proposed method showed\n15.9%, 14.3%, and 10.5% improvement. These results supported a positive conclusion on the\nperformance of designed method, specifically, the mixed deep recurrent neural networks\noutperformed than the two-layer gated recurrent unit and the two-layer long short term memory\nrecurrent neural networks.
Loading....